MIME-Version: 1.0 Content-Type: multipart/related; boundary="----=_NextPart_01C9D02C.2AB56840" This document is a Single File Web Page, also known as a Web Archive file. If you are seeing this message, your browser or editor doesn't support Web Archive files. Please download a browser that supports Web Archive, such as Windows® Internet Explorer®. ------=_NextPart_01C9D02C.2AB56840 Content-Location: file:///C:/EA89CAB4/CRITFinalReport.htm Content-Transfer-Encoding: quoted-printable Content-Type: text/html; charset="us-ascii" Collator Replacement Integration Team

 

 

Collator Replacement Integratio= n Team

Schedule Optimizer Module<= /o:p>

Final Report<= /b>

 

 

 

 

George Mason University=

Dr. Thomas Speller

SYST 798/OR 680

Spring 2009=


Contents

Contents. 2<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Figures. 2<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Tables. 3<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Executive Summary. 3<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Introduction. 3<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Project Development Plan. 4<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Team Roles and Responsibilities. 5<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Project Intent, Expected Results. 6<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Problem Definition, Mission and Goals. 10<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Business Strategy / Business Approach. 30<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Recommendations. 33<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Conclusions. 33<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Bibliography. 34<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Appendix. 34<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

 

Figures

Figure 1 Production system schematic diagram.. 4<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Figure 2 Value stream map. 9<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Figure 3 Project schedule. 11<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Figure 4 top-level functional requirements. 13<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Figure 5 detailed functional requirements= . 13<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Figure 6 detailed non-functional requirements. 14<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Figure 7 alternative evaluation process. 14<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Figure 8 functional decomposition (initial). 15<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Figure 9 functional decomposition (final)= . 16<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Figure 10 Collator Production Mode Options. 17<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Figure 11 Algorithm Illustration. 20<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Figure 12 Sample Scheduling Results for March 16= , 2008. 29<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Figure 13 Sample Assignment Results for March 16= , 2008. 30<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Figure 14 Weekly sales trend. 31<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Tables

Table 1 Roles and responsibilities. 6<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Table 2 Prioritized stakeholder wants and needs<= /span>. 7<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Table 3 Definition of terms and acronyms<= span style=3D'color:windowtext;display:none;mso-hide:screen;mso-no-proof:yes; text-decoration:none;text-underline:none'>. 8<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Table 4 List of constraints. 12<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Table 5 Hopper Configuration Worksheet. 18<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Table 6 Feeder Positions Calculation. 19<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Table 7 Cost calculations. 32<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Table 8 Financing cost and break even time. 33<= span style=3D'mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin; mso-fareast-font-family:"Times New Roman";mso-fareast-theme-font:minor-fare= ast; mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-fon= t-family: "Times New Roman";mso-bidi-theme-font:minor-bidi;mso-no-proof:yes'>

Executive Summary

Scheduling, production specification and collat= ion of Sunday newspaper advertisement inserts is manually calculated and produced = with inefficient equipment. The customer needs the programming logic for an automated solution that minimizes total production cost. Our greedy heurist= ic algorithm solution will calculate repeatable schedules and assign advertise= ments to collator hoppers. Cost savings for best, nominal and worst case scenario= s are calculated for new equipment to support an investment decision of several million dollars. The solution is documented with functional decomposition diagrams, functional architecture diagrams, requirements, mathematical notation, operational use diagrams, value stream mapping, and a business ca= se analysis.

Introduction

We are a team of two systems engineeri= ng and two operations research majors in the Master’s degree program at Maso= n. Our capstone course customer is the production department of a major national d= aily newspaper. They currently sell and distribute approximately 920,000 advertisement insert packages with every Sunday edition of the newspaper. Physical collation, bundling and palletizing advertisements for delivery to= as many as 550 zones is accomplished with humans and machinery as shown in  REF _Ref229036086 \h Figure 1. The machines have been in continuous operation for thirty years. Scheduling= the sequence of advertisements to be collated into newspaper insert books is currently accomplished manually. The scheduling and hopper assignment proce= sses are based on a penalty matrix of extensive knowledge of what works best. The scheduling subject matter experts have decades of experience with this proc= ess, and replacements will be difficult to train when retirement occurs. In addition, the changing nature of the newspaper advertising business has necessitated increased complexity. The environment has increased from a lim= ited number of large advertisers and wide distribution to greater numbers of sma= ller sales within subdivided distribution areas. Finally, the existing equipment= is reaching the end of its useful service life. The combination of these facto= rs has resulted in reduced margin that should be improved with new technology solutions.

= Figure 1 Production system schematic diagram

The goal of the project is to produce = an algorithm that creates practical schedules and material handling equipment = configurations responsive to physical conditions on the shop floor, advertising and circulation requirements.  The primary objectives of this project are to define:

·         A consistent, accurate, and repeatable sched= ule algorithm

·         Production schedules optimized for cycle tim= e

·         Optimally-sequenced schedules based on edito= rial templates and machine capacity for the current equipment and configuration<= /p>

·         Hopper configurations, sequenced schedules a= nd hopper advertisement assignments optimized based on a combination of human knowledge and experience and dynamically-calculated mathematical analysis f= or the future equipment and configuration

·         An investment decision recommendation suppor= ted by business case analysis

Project Development Plan

The project development plan is to define and b= uild a system that meets requirements (customer needs) without violating constrain= ts. We will use GEIA-632 as an informative standard for systems engineering pro= cess management requirements because it provides a balanced mix of the entire spectrum of management and technical processes. Team roles and responsibili= ties as aligned with this standard are defined in Table 1. A spiral development process of requirements elicitation, definition, model= ing, prototyping, and validation will be continuously iterated from start to fin= ish. This development model is necessary and appropriate to ensure that we deliv= er a useful product that the customer can rely on for production scheduling and a capital investment decision.

1.   &n= bsp;   Strategy

a.   &n= bsp;    Use mind ma= pping to brainstorm functional requirements, and a combination of deductive, inductive and abductive reasoning to identify possible solutions=

<= ![if !supportLists]>   &nbs= p;            &= nbsp;           &nbs= p;            &= nbsp;           &nbs= p;            &= nbsp;        i.   &nb= sp;  Elicit requ= irements

<= ![if !supportLists]>   &nbs= p;            &= nbsp;           &nbs= p;            &= nbsp;           &nbs= p;            &= nbsp;       ii.   &n= bsp;  Define requirements through functional decomposition

<= ![if !supportLists]>   &nbs= p;            &= nbsp;           &nbs= p;            &= nbsp;           &nbs= p;            &= nbsp;     iii.   &= nbsp;  Prioritize requirements

b.   &n= bsp;   Identify in= tent specification – use progression of principles development =

<= ![if !supportLists]>   &nbs= p;            &= nbsp;           &nbs= p;            &= nbsp;           &nbs= p;            &= nbsp;        i.   &nb= sp;  Collect dat= a

<= ![if !supportLists]>   &nbs= p;            &= nbsp;           &nbs= p;            &= nbsp;           &nbs= p;            &= nbsp;       ii.   &n= bsp;  Make observ= ations from the data based on analysis

<= ![if !supportLists]>   &nbs= p;            &= nbsp;           &nbs= p;            &= nbsp;           &nbs= p;            &= nbsp;     iii.   &= nbsp;  Empirically verify observations

<= ![if !supportLists]>   &nbs= p;            &= nbsp;           &nbs= p;            &= nbsp;           &nbs= p;            &= nbsp;     iv.   &n= bsp;  Describe observations

c.   &n= bsp;    Define the problem

<= ![if !supportLists]>   &nbs= p;            &= nbsp;           &nbs= p;            &= nbsp;           &nbs= p;            &= nbsp;        i.   &nb= sp;  Gather data=

<= ![if !supportLists]>   &nbs= p;            &= nbsp;           &nbs= p;            &= nbsp;           &nbs= p;            &= nbsp;       ii.   &n= bsp;  Interpret d= ata

<= ![if !supportLists]>   &nbs= p;            &= nbsp;           &nbs= p;            &= nbsp;           &nbs= p;            &= nbsp;     iii.   &= nbsp;  Summarize concepts

2.   &n= bsp;   Concept implementation

a.   &n= bsp;    Follow a cr= eative process of discovery based on application of scientific method-controlled e= xperiments

b.   &n= bsp;   Evaluate scenarios

c.   &n= bsp;    Allocate functions to forms, identify interface requirements and resolve conflicts

d.   &n= bsp;   Aggregate f= orms and functions, specify interface requirements and resolve conflicts

e.   &n= bsp;   Evaluate sy= stem architectures suitability and ability to meet requirements

f.   &n= bsp;     Select best= system physical architecture utilizing morphological analysis with cross-consisten= cy checking to identify best instantiations

g.   &n= bsp;    Build proto= type of best solution, resolve conflicts

h.   &n= bsp;   Test, resol= ve conflicts, try a different solution if initial is deemed unsuitable

i.   &n= bsp;     Validate performance against customer requirements; maintain bi-directional requirem= ents traceability throughout development process to facilitate stakeholder value mapping

j.   &n= bsp;     Invite cust= omer to test the prototype solution, solicit feedback

k.   &n= bsp;    Build final= solution

l.   &n= bsp;     Deliver to customer

m.   &n= bsp;  Provide ini= tial training

3.   &n= bsp;   Most critic= al function identification & risk mitigation

a.   &n= bsp;    Prioritizat= ion of IPC scheduling and hopper assignment/configuration designation to minimize = production time

4.   &n= bsp;   Final presentation

a.   &n= bsp;    Document pr= ocess, rationale and results

b.   &n= bsp;   Publish doc= uments and models on web site           =

Team Roles and Responsibilities

 

Team Member & Role

 

Philip Coady

Joseph Cremaldi

John Deas

Steven Escaravage

EIA-632 Processes

Configuration and data management

Support (S)=

Lead (L)

S

S

 

Decision analysis

L

S

S

S

 

Design Solution (logical & physi= cal architecture form)

S

L

S

S

 

Implementation

S

S

S

L

 

Integration

S

S

L

S

 

Interface management

S

L

S

S

 

Logical analysis (functional archite= cture)

S

S

L

S

 

Requirements definition &am= p; management

S

L

S

S

 

Risk management

S

S

L

S

 

Technical assessment

L

S

S

S

 

Technical planning

L

S

S

S

 

Transition

S

S

S

L

 

Verification & Validation

S

S

S

L

 

Table 1 Roles and responsibilities

Project Intent, Expected Results

The customer is faced with the challenges of decreasing advertising volume, increasing complexity due to microzoning, an= d reduced productivity due to difficult-to-handle materials and aging collator machin= ery. In response to these challenges, we were asked to provide recommendations regarding how to optimize the current Sunday advertising scheduling and insertion process. In addition, we were asked to prepare a business case for replacing the current collator with new technology, and optimized schedulin= g to eliminate or reduce many of the current constraints.

Stakeholder and Needs Identification

The stakeholders and their prioritized requirem= ents are listed in Table 2.

Level

Stakeholder<= /p>

Wants/Needs<= /p>

Value*

Primary stakeholders=

Sponsoring Company

Increased profitability

100

Production manager

More efficient process, workfor= ce savings

75

Data manager<= /p>

Greater automation, less variat= ion, workforce savings

85

Schedulers

Greater automation, less variat= ion

70

Workers

More efficient work day, Job se= curity

20

Secondary stakeholders

Advertisers

Less wait time, increased opportunities for targeted, smaller purchases

10

Collator manufacturer

Proof of design, requirements v= alidation

90

Trucking companies

Opportun= ity for more efficient scheduling

70

FSI printers<= /p>

Increased printing time & customer base

50

GMU SEOR Department<= /span>

Potential for more projects, educational feedback to the faculty

50

Table <= /span>2<= span style=3D'mso-bookmark:_Toc229574950'> Prioritized stakeholder wants and n= eeds

* 0 – 100 point scale, 100 being the greatest gain<= /span>

 

Industry-specific terminology and customer-spec= ific acronyms used throughout this report are defined in Table 3.

Term

Definition<= o:p>

AIMS

Accurate Insert Management Syste= m

Book

A package of collated and wrappe= d FSIs assembled to an IPC specification. Also referred to as a package.

Bundle

A labeled package of strapped bo= oks

Bundle count

The height restriction in MOGS t= hat limits the number of products per bundle for the bundler and palletizer

BPM

Books Per Minute

Collator

Production system consisting of hoppers, conveyor belt, controls, meters that collects designated FSIs and stacks them into books or packages according to IPC specifications

Collator state=

FSI assignments by hopper for an= IPC

Distribution zones

Geographic distribution areas wi= thin editorial zones (e.g. Arlington, Fairfax, Montgomery)

Dual mode

Running two sides of collator independently (applicable to new equipment only)

Editorial zones

Broad geographic distribution ar= eas (e.g. Virginia, DC, Maryland)

Feeder

The Utility mailer role of loadi= ng advertisements into hoppers

FSI

Free Standing Insert, referred t= o as an advertisement

Hard changeover (current)

Physical movement of an FSI out = of and into a hopper at any time. Requires a stoppage in production.<= /span>

Hard changeover (future)

Physical movement of an FSI out = of and into a hopper with no opportunity for idle time replacement. Requires a stoppage in production.

Head

Component of collator where FSIs= are laid into the raceway

Helper

Lowest-skill workers<= /span>

Hopper

Component of collator where FSIs= are stacked

House book

A FSI that is used in every IPC = (i.e. TV Week, comics, Magazine, Parade)

IPC

Individual Product Code, referre= d to as a job or order

IPC Book

Individual product code specific= ation consisting of designated FSIs in no particular order other than end piece= s

ISS

Insert Scheduling System, the sy= stem of systems that provides end-to-end sales, production, and delivery of in= sert advertisements

Mailers

Highest-skill workers who operat= e the feeders and collator

Micro zone

Small geographic distribution ar= eas within distribution zones

MOGS

Manufacturing Order Generating S= ystem; Specifies the FSI composition of an IPC.

Package

The book of collated FSIs produc= ed for every IPC

Penalty

Time cost associated with change= overs

PPFM

Pages per feeder minute

Product run

Collation of all required FSIs i= n an IPC

Production Zone

IPC

Raceway

Collator conveyor belt

Ripple start

Collator restart after a hard changeover

Ripple stop

Hard changeover indication when = collator stops for next IPC

Sharing opportunity

A hopper with more than one FSI assigned for separate IPCs

Single mode

Running two sides of collator in series (applicable to new equipment only)

Soft changeover (current)

Switch on-off of a FSI within a = hopper or insertion of a FSI into an open hopper

Soft changeover (future)

Current capability plus the abil= ity to clear excess FSIs from a hopper while the collator is running (reduces ha= rd changeovers)

SOM

Schedule Optimizer Module

TV zone

Large geo= graphic distribution areas

Utility Mailers

Moderatel= y-skilled workers who feed stacks of FSIs on hoppers and clear excess FSIs from hop= pers on completion of product run. May also operate forklifts.

Table 3 Definition of terms and acronyms

Stakeholder Needs Analysis and Value Map

The problem is that the current system is ineff= icient, labor intensive, and does not account for new equipment constraints and capabilities. The production process value stream is shown in Figure 2.

3D"Sunday-only Figure 2 Value str= eam map

Future production operations will be d= riven by the need to serve an increasing number of advertisers in up to 550 micro= zones. If ten advertising customers in each microzone placed discrete orders for a given Sunday edition, the production schedule would increase from the curre= nt limit of approximately 300 schedules to over 550, and the average volume of= a production run could increase.

Scheduling must be accomplished in one = hour processing time or less using hardware and software of our design or specification. Ideally, the solution will run on the customer’s system under Windows/Office 97.

  • Scheduling is on= ly partially automated
  • Requires manual hopper selection, assembly li= ne mapping, and job assignments
  • Requires long lead-times to produce schedules=
  • Collator operation has multiple inefficiencie= s
  • Feeding and conditioning FSIs into collator hopp= ers is labor intensive
  • Changing hopper contents for the next IPC is = time (3-5 minutes) and labor intensive
  • Line must stop during change-over causing labo= r loss
  • Usually 250 – 300 IPCs consisting of 25-30 FSIs per Sunday edi= tion

Stakeholder Needs: 

  • Develop a feasible, repeatable solution
  • Automate the scheduling process
  • Provide scheduling algorithms

Goals:

  • Introduction of = new collator technology creates an opportunity for process improvement
  • Reduced number of hard and soft changeovers =
  • Schedule should facilitate increased pace of new machine
  • Minimize the effects of any required hard changeo= vers
  • New collator provides physical performance enhanc= ements to minimize conditioning, thereby increasing feed rate per worker from= 8 to 16 thousand pages per feeder minute (PPFM)
  • Hopper changeovers can be accomplished with= out stopping the line, therefore a goal is to schedule the entire producti= on run without mandating a hard changeover
  • Analysis should = result in identification of other su= pply chain optimization opportunities

Design Schedule Optimizer Module:

  • Develop algorithms to optimizes job schedul= ing and hopper assignment for Sunday inserts
  • Minimize instances of hard changeovers
  • Minimize total production time
  • Minimize resource requirements <= /span>
  • Meet all production deadlines

Secondary goals

  • Identify required computer hardware
  • Define software interfaces
  • Document functional and physical architectu= res, software specifications, and provide mathematical solution<= /span>
  • Verify system design using modeling tools

Other tasks

  • Develop business case analysis to determine options and viability
  • Recommend replacement collator’s phys= ical attributes (hopper arrangement)

 

Quality Function Deployment (QFD)

We used a proprietary spreadsheet implementatio= n of the QFD House of Quality to prioritize customer needs and wants. Appendix E is the results document (spreadsheets).

Problem Definition, Mission and Goals

Problem Definition

The current Sunday advertising insert production operation requires 31 labor shifts of seven hours each on three collators that produce packages up to 60% of the available time. Hopper assignments are developed based on experienced scheduler knowledge and a ShivaSoft™ algorithm [Neiss]. This arrangement= was acceptable when hopper assignments remained relatively consistent throughout large wee= kly production runs. It is insufficient to meet the challenges of increasing complexity, materials, and projected sales that will increase the number of discrete ads and the volume of small orders.

The project schedule is shown in Figure 3.

3D"clip_image001"

Figure 3 Project schedule

Customer constraints are listed in Table 4.

Constraint

Impact

Solution must be generated in less than one hour

Prevents possibility of true optimization by invalidating a solut= ion

Utility mailers will handle 16,000 pages per minute while loading hoppers

Restricts distribution of workload; determines number of workers required per shift

Accuracy must meet or exceed 98.5%

Limits solution by restricting losses

Pallets of FSIs must not be used in disparate hoppers or collator= s

Limits solution by preventing use of FSIs without regard to hopper location

IPCs within TV zones should be produced contiguously

Reduces optimization of production run time by forcing a limited number of hard changeovers

Excess FSIs should not be reloaded onto pallets after placement in hoppers

Requires complete consumption of a pallet, thereby limiting the number of FSIs available for loading

The first and last positions in hopper assignments are reserved

Requires hard coding of designated FSIs to certain hoppers

Physical limit on number of hoppers

Cannot avoid single-mode hopper configurat= ion for a small number of IPCs with a large number of FSIs

Table 4 List of constraints

Systems Engineering Methodology [Buede]

1.      = Requirements elicitation was continuously employed throughout the development process.

2.      = Requirements definition was used to refine and manage the originating and derived requirements.

3.      = Functional decomposition was accomplished using mind mapping tools & techniques. T= he mind map was then translated into a hierarchical representation which became the baseline document.

4.      = Functional architecture development was based on the functional decomposition. Derived requirements were validated against customer needs and wants throughout the development process.

5.      = Morphological analysis was used to identify the best available solution forms for each functional requirement. Cross consistency checking methods were used to sco= re and rank the alternative solutions and select the best combination.

6.      = Physical architecture development involved designing the logic automation algorithms that would satisfy the validated functional requirements.

7.      = Scheduling and hopper assignment algorithm designs were implemented in MATLAB.

8.      = Verification of design to specifications was conducted to ensure that the solution built= is consistent with our stated understanding of the requirements.

9.      = Implementation of solution uses a combination of MS Access reports, MS Excel data tables, = and MATLAB computations. The solution accepts MS Access input and provides outp= ut in MS Access-readable format.

10.<= span style=3D'font:7.0pt "Times New Roman"'>    Integration of hardware and software involved ensuring that our solution would run on Office 97 applications in less than one hour.

11.<= span style=3D'font:7.0pt "Times New Roman"'>    Validation of solution to the customer’s needs and wants was conducted prior to delivery to ensure that we addressed the requirements in a manner that was feasible, useful and repeatable.

12.<= span style=3D'font:7.0pt "Times New Roman"'>    Delivery consisted of providing electronic and hard copy of the algorithms and associated routines, and instructions for their use.

13.<= span style=3D'font:7.0pt "Times New Roman"'>    Documentation was provided electronically and in hard copy formats.

Scope and Context Definition

The scope of this project is limited to optimiz= ing schedules. The input is a list of FSIs that comprise each IPC, and the outp= ut is the sequence of IPCs and hopper assignments to be assembled on the colla= tor. We recognize that optimizing one part of the production process may result = in perturbations in other parts, however, we have attempted to identify and minimize those effects to the maximum extent possible.

Operational Concept

The input consists of a database table= of IPC specifications of FSIs for the weekly production run.

The solution will be used to generate sequenced hopper assignments for the entire production run. If changes are required during production, the solution will be able to recalculate the remaining hopper assignment sequence from within any point in the schedule. With few exceptions, the production run is to be completed without any hard changeovers.

The output is a set of printable databa= se tables of IPC specifications with hopper assignments for each FSI in sequen= ce.

Functional Decomposition

A group mind mapping process was used to decomp= ose required functions and performance as shown in Figure 3 top-level functional requirements, Figure 4 detailed functional requirements and Figure 5 detailed non-functional requirements 08D0C9EA79F9BACE118C8200AA004BA90B02000000080000000E0000005F005200= 650066003200320039003000330039003900350037000000 . 3D"ISS

Figure 4 top-level functi= onal requirements

3D"ISS

Figure 5 detailed functio= nal requirements

 = ;

3D"ISS

Figure 6 detailed non-fun= ctional requirements

Technology Strategy

<= ![endif]>

Figure 7 alternative evaluation process=

Alternative Solution S= pace is a factorial combination of potential forms for each function

Constraints include non-functional requirements of the proposed system, operational and organizational constraints of the client, and operational constraints of th= e delivery team

In some cases, common sen= se was applied to further reduce the alternative space (i.e., unrealistic options - substantial data entry)

An Effectiveness Ratin= g or relative evaluation was used to differentiate alternatives in terms of performance against functional requirements

The Preferred Alternat= ive was identified through application of an additive value function of the normali= zed effectiveness ratings and HOQ weights

Throughout the evaluation process, the resulting alternative instantiations were evaluated against ne= eds and wants to mitigate risks

Architecture Development

Our initial functional decomposition d= iagram is shown in Figure 8. The identified functions were mapped to a number of possible forms. The for= ms were considered and analyzed using morphological analysis to determine the = best options. As the design matured, functions were reallocated and refined.

3D"ISS

Figure 8 functional decomposition (initial)

The = as-built functional decomposition is shown as Figure 9= , from which the functional architecture [Appendix B] diagrams were developed. The = two best system solutions were selected for prototype development and testing. = The two prototype solution models were then tested in simulation against various problem sets given one year of historic data and our projections of future sales. The option that met all validated customer requirements was selected= for the physical architecture and implementation. Derived requirements were verified to originating requirements as they were identified.

 = ;

3D"ISS

Figure 9 functional decomposition (final)

Assumptions

Delivery of FSI pallets to the hopper staging area is not a constraint

The new equipment will handle at leas= t 90% of the Sunday workload

Spatial orientation of FSIs in books = is not a requirement

Hard changeovers will take one minute= or less on the new equipment

Utility mailers will work as a team assigned to a group of hoppers vs. individually by specific hoppers

The new equipment will stop and resta= rt without penalty, therefore overlapping shifts are not necessary to optimize production

FSI conditioning is not required on t= he new equipment, thereby freeing workers to load up to 16,000 pages per minute instead of the current standard of 8,000

An arbitrary dispatch schedule is acc= eptable assuming the production run is completed by Friday

MOGS-generated IPCs will be impacted = by incomplete FSI deliveries prior to commencement of production, therefore scheduling module must be responsive to changes

The new equipment will produce 225 pa= ckages per minute and operate at 90% availability

The greatest workload occurs during h= oliday seasons and is used for worst case analysis. The median workload can be accommodated without penalty using the worst case solution provided.

Utility mailer skill level, plus supervision, will suffice as the entire workforce on the new equipment, the= reby reducing the average labor rate.

Utility mailers are multi-skilled and c= an be used to operate forklifts to stage FSIs when hopper feeder demands are redu= ced.

Algorithm Process and Results

The following sections describe the algorithm developed [Albright] to find the IPC sequencing schedule. Individual sections outline the initial data pre-processing steps, the scheduling modules, and the assignment modules.

Data Pre-Processing

The Sunday Edition insert sched= ule is impacted by variations in the quantity and make-up of the weekly advertisement order and the associated collator configuration and staffing levels to best support those variations.  Here, the collator configuration r= efers to the number of hoppers positioned on the two raceways (conveyer belts), w= here the total on the long side must be less than or equal to 50 hoppers, and the total on both sides must be less than 80 hoppers.  An advantage of the new collator i= s the ability to run the machine in dual mode – where the two sides of the collator operate independently essentially doubling production capacity = 211; or single mode – where the two sides snake together to increase the hopper capacity while reducing the total production capacity.  These two modes are depicted Figure 10 in below. 

Figure 10 Collator Production Mode Options

To take advantage of this operation mo= de flexibility, the machine must be configured in such a way to approximately minimize total production time by balancing the workload on the short and l= ong sides of the machine, while considering the trade-off of single mode and du= al mode operations.  For example,= if the machine were equally balanced with 40 hoppers on both sides, individual jobs with less than 40 FSIs (advertisements) could be produced on either si= de of the machine.  Under this configuration, all jobs with more than 40 FSIs would need to be produced in single mode.  As an alternative example, if the collator where configured with 50 hoppers on the long side = and 30 hoppers on the short side, the increased capacity of the long side would= reduce the need for single mode operations.  However, if less than 50% of the job groupings can be run on the sho= rt side, the total production time will increase due to the unbalanced product= ion load on the long side. 

As a result, the proposed algorithmic a= pproach requires the pre-processing of the job make-up for the given week to determ= ine the preferred hopper configuration.  The inputs to this pre-processing are the maximum IPC size for each = job grouping, the total packet demand for each job, a target machine speed, and= a target uptime percentage.  The inputs are used to determine the total runtime and hopper requirements for = each job group.  Then, starting wit= h the extreme configuration case of 30/50 (short side/long side), all hopper configuration possibilities are enumerated to determine the potential production runtime for the short side, long side, and single mode requirements.  For cases where= the short side potential runtime is less than 50% of the total runtime, the sho= rt side production time estimate is set equal to the short side potential runtime.  If the short side potential runtime exceeds 50% of the total runtime, the short side and long side are set equal to approximately half the total production time estimate excluding any jobs that must be produced in single mode.  A sample calculation for editorial= week 3/16/2008 is presented in Table 5 below, with the resulting preferred configuration of 35/45 hoppers.

Table 5 Hopper Configuration Worksheet=

Similarly, required staffing levels ar= e a function of the product make-up and target machine speed.  Staff members responsible for “feeding” the advertisements (FSIs) onto the hoppers are design= ated “Feeders.”  For th= is effort, we assumed a single constraint governed the required number of Feed= ers per job, namely the pages per feeder minute constraint.  The Sunday Insert Production Job F= oremen target an average of 16,000 pages per Feeder minute (PPFM).  Simply stated, a single Feeder sho= uld not be responsible for loading advertisements onto hoppers at a rate of gre= ater than 16,000 pages per minute on average.&n= bsp; Individual advertisements or circulars with a large number of pages,= for example 40 pages, require more frequent Feeder loading into the hoppers giv= en that each packet removes 40 pages from the stack as compared to a two page advertisement, where each packet would only remove 2 pages from the stack.<= span style=3D'mso-spacerun:yes'>  In practice, the term “tab pages” is used to represent the number of sides per advertisement (FSI).  As a result, the small= est number of pages (“tab pages”) encountered by the algorithm is 2 (for a single sheet of paper).  The resulting PPFM constraint is presented below for a single job, although the constraint is traditionally applied as an average across all jobs.  Here, the pages collection holds t= he tab page size of each advertisement (FSI); speed indicates the target packets p= er min (e.g., 225); the third condition includes only the advertisements (FSIs) assigned to the Feeder position under interest.

 

The proposed solution, given an aggress= ive schedule, requires the predetermination of the average number of Feeders per shift to provide the assignment algorithm with a non-variable number of Fee= der positions for assignment.  Fut= ure iterations of the algorithm might incorporate sensitivity analysis and cost-effectiveness constraints to determine the number of hopper positions = to feed or treat the number of positions as a variable to be determined in-line during execution of the heuristic.  Given this limitation, the algorithm uses the integer ceiling (round= up) of the average number of Feeders required per shift.  Here, the number of Feeders per sh= ift would be calculated as the production time weighted average of the number of Feeders required per job (IPC)[1].  For jobs that require more than the resulting number of Feeders, the machine speed would be reduced as to not violate the PPFM constraint.  = For jobs that require less than the number of Feeders, forecasted changeovers or other activities could be planned to consume the surplus labor.  A sample Feeder position calculati= on is provided in the figure below.

Table 6 Feeder Positions Calculation

Potential Algorithmic Approaches

The objective of the algorithm is to p= roduce a feasible job schedule and hopper assignment that minimizes the downtime of the collating machines due to product change-over.  Here, the use of the term “f= easible schedule” is to assume the resulting algorithm will not increase over= all production time in an attempt to minimize product change-over time (e.g., r= un the entire editorial week in single mode on one collator).  Given the minimization goal, a num= ber of optimization and heuristic approaches were evaluated for a proof of concept implementation. 

Exact optimization methods were resea= rched to ensure the size of the problem was reasonable in terms of the number of potential “nodes” in the job “network.”  Given the number of jobs per week = ranges between a lower bound of 200 and an upper bound potential of 550, all brute force enumeration and many progressive improvement algorithms were eliminat= ed from consideration based simply on size (i.e., many progressive algorithms = support up to only 200-225 nodes).  Ho= wever, some advanced cutting plan solutions were found to easily handle the potent= ial network size, and could theoretically handle up to 10,000 node tours.  In the end, exact optimization met= hods were down-selected due to (1) computational complexity and the lack of availability of a computational infrastructure to support customer deployme= nt testing, (2) the need for the resulting algorithm to be deployed as a module within the current software environment, and thus be implementable in stand= ard software development packages (3) to keep the scope manageable in terms of constraint derivation and implementation, and (4) realizing that in practic= e, the results of the algorithm will be used as a starting point by Client Experts, and thus transparency outweighs the need to prove optimality.

In addition, constructive and improve= ment heuristics were evaluated given the relative ease of implementation and accuracy of results.  Construc= tive heuristics generate a path through greedy- or insertion- methods (e.g., nea= rest neighbor, best insertion) and generally produce fairly good results estimat= ed from 1.10x to 2x optimal tours when the problem is structured as a traveling salesman problem (TSP).  In th= e case of the Sunday Insert problem, the week can end on any job compared to TSP constraint of returning home, thus estimates are potentially even larger th= an would be found in practice.  However, in some and not necessarily rare cases, constructive heuris= tics will produce far from optimal spanning paths (e.g., 2x optimal).

Lastly, the improvement heuristics we= re researched.  Improvement heuri= stics start with an initial sequence or schedule developed through a constructive approach, and attempts to improve the initial schedule through removal or reconstruction of a subset of transitions.=   For example, k-opt appro= aches attempt to improve spanning paths and tours through selection and evaluatio= n of k transitions.  Other approaches use similar “swapping” or “switching” approaches paired with ar= tificial intelligence components to control the number of improvement attempts (e.g., simulated annealing, genetic algorithms).&= nbsp;

In the end, a greedy constructive heuri= stic was selected to ensure prototype testing and analysis would be possible wit= hin the provided period of performance.  We elaborate on future considerations and extensions to the algorith= m in Recommendations 08D0C9EA79F9BACE118C8200AA004BA90B02000000080000000E0000005F005200= 650066003200320039003400390031003400340032000000 .

Introduction – the Scheduling and Assignment Algorithm

A greedy scheduling and assignment heu= ristic was derived and formulated as the key deliverable of the project.  The heuristic first iterates throu= gh the order detail (Product Table), and at each step determines the next best job= to place at the end of the current schedule as to minimize the number of hard changeovers and soft change-over respectively given the current assignment = of the collator.  Subsequently, t= he algorithm re-orders the initial hopper assignment to balance the workload across the predetermined number of Feeder positions.   

Figure 11 presents a high-level overview of the inputs and results of the algorithm, which will be elaborated in the following sections.

Figure 11 Algorithm Illustration

Step 1: Declaration and Initialization

This= section defines and initializes the data structures and constants used throughout t= he modules in the algorithm.  The following sections will use the actual variables names based on the mapping below:

Acro= nym        &= nbsp;       Reference Name      &nb= sp;            =             &nb= sp;  Generic Name
FSI      &nbs= p;            &= nbsp;       Free Standing Insert    &nbs= p;            &= nbsp;          Advertis= ement, Ad, Insert, Circular
IPC      &nbs= p;            &= nbsp;      Integrated Product Code     &= nbsp;           &nbs= p;  Job, Product

The first set of declarations and initializations establishes the static values that describe the current configuration of the new collator, the product make-up in terms of number of FSIs and number of = IPCs.&nb= sp; Throughout the remainder of the formulation, any side specific notat= ion will be listed for the long side of the collator only.  In actual implementation, two inst= ances of the algorithm would be implemented and called as subroutines or procedur= es on the remaining set of IPCs to be scheduled.

Let numFSIs   &= nbsp;           &nbs= p;       =3D the total number of FSIs in a given edition week
Let numIPCs     &n= bsp;            = ;    =3D the total number of IPCs in given edition week
Let numhopper<= /a>sLong      =3D the num= ber of hoppers on the long side of the new collator

The next declarations establish the primary data structures us= ed to store the product and product make-up tables from the Insert Scheduling System Database.  Also, the following indexing applies to all data structures:

i   =3D 1..numIPCs
j   =3D 1..numFSIs
k  =3D 30…numhoppersLong<= br> m =3D 1..numIPCs
n  =3D 1..numFSIs

Let IPCi =3D the ith integrated product code for the edition week
Let FSIj =3D the jth free standing insert for the edi= tion week

Let IPCij     &n= bsp;            = ;        =3D      =

Let collatorLongk    =       =3D

Let jobLongm     = ;            = =3D  

Let scheduleLongm,n    =3D   

Let sizeIPCi    =             &nb= sp;   =3D the number of FSIs in IPCi

Let productionTimei   = ; =3D the production time in hours at 90% uptime for IPCi

Let productionLimit     =3D the max runtime for the current side of the collator as determined through preprocessing

Let totalProdTime     =    =3D the total production time for all IPCs in the current schedule

The next set of initializations and declarations sets the valu= es for known production constraints.  The first statement declares and initializes the set of FSIs referen= cing the TV Books.  Similarly, the = FSI identifiers for the weekly comics, newspaper magazine, and Parade magazine = are declared and initialized.

Let tvBooks             =       =3D {identifies a TV Book}
Let comics     &nb= sp;            =    =3D {identifies the Comics}
Let magazine     &= nbsp;           =3D {identifies the Magazine}
Let parade     &nb= sp;            =    =3D {identifies the Parade Magazine}
<= /p>

Lastly, a set or collection is establishe= d to store all remaining IPC identifiers that have not been assigned a schedule sequence, remainingIPCs, and a set to store all IPCs that exceed the side capacity currently under run that must be reinserted following initial sche= dule generation.  The final stateme= nt removes all IPCs that exceed the current side number of hoppers from the current scheduling set.

Let remainingIPCs =        =3D {1, 2 ... numIPCs}
Let singleModeIPCs     = =3D {> the numhoppersLong or numhoppersShort (depend= ing on current schedule)}
remainingIPCs =3D remainingIPCs - singleModeIPCs

Step 2: Select Lead IPC and Initialize Hopper Positions Sequentially

Given the dual mode capability of the new collator, and the gr= eedy nature of the algorithm, care must be taken to ensure jobs in groupings wit= h a relatively large number of FSIs are reserved for the long side of the colla= tor while jobs with a relatively small number of FSIs are reserved for the small side of the collator.  As a re= sult, we require an initial pre-processing step to determine which side of the collator to prioritize scheduling of each job. 

Let sidePreferencei     =3D      

Given this preference for each job grouping, and ultimately ea= ch IPC, we first develop the schedule for the long side of the collator, and t= hus prioritize selection to those job groups with large size preference.  Within this group, the algorithm identifies the IPC with the largest sum total of FSI frequencies as the lead IPC.  Stated in another way, t= he algorithm looks to identify the IPC with the largest number of frequently u= sed FSIs. 

For all IPCs with sidePreferencei =3D 1:

Let freqj             =             = =3D the frequency with which an FSIj occurs across all IPCs

 &nb= sp;            =             &nb= sp;            = =3D

Next

Let jobLong= 1        &= nbsp;         =3D IPCi with the highest summed total of FSI frequency with sizeIPC= i < numHoppersLong + 1

        &= nbsp;           &nbs= p;            &= nbsp;      =3D

The algorithm then initializes the long side of the collator and the schedule f= or the lead IPC.  Here, FSIs are assigned sequentially to the first j hoppers positions, where j is equal to= the number of FSIs in the lead IPC. Lastly, the IPC number of the selected IPC = is removed from the remainingIPC set.

Let collatorLongk    =       =3D sequential order of FSIs in jobLong1 (lead IPC)

        =3D

Let scheduleLong1n     = ;    =3D collatorLongk   =   for n=3Dk, for k =3D 1..numh= oppersLong

Let remainingIPCs     =    =3D remainingIPCs – {jobLong1}

Step 3: Iteratively Find the Next Job

The next st= ep in the algorithm is to recursively select the next job that introduces the minimum number of hard and soft changeovers from within the current active = job grouping.  In the event of a t= ie break scenario, the job with the highest summed total within the set with the min= imum number of hard and soft change-over is selected. Again, given the first set= of schedule assignments is focused on the long side of the hopper, only jobs w= ith a large side preference are evaluated.

This module first sets an index for the current job number to be scheduled, m, and decl= ares two arrays to capture the number of hard and soft change-overs required to transition the collator from the current IPC to the next IPC.   Next, the module begins an iterative “While Loop” that cycles through all IPCs not current= ly assigned to a schedule sequence.  The next set of statements determine the set of unassigned IPCs that should be considered for the next job based on the following priority: (1) = all IPCs in the current TV book zone, (2) all IPCs not in the current TV book z= one but with a long side preference, (3) all remaining IPCs.  The resulting set of IPCs is set e= qual to a temporary holding set or collection, potentialIPCs. This process is terminated if the total production time for the current side schedule excee= ds the predetermined runtime upper limit, and all IPCs within the current TV B= ook zone are included in the schedule.

Let m =3D 1  

Let soft_changeoveri =3D the number of soft change-overs required to= run IPCi next on collator1

Let hard_changeoveri =3D the number of hard change-overs required to= run IPCi next on collator1

While {

IF ()
       &= nbsp;           &nbs= p;            Let potentialIPCs =3D {}

Else IF ()

        &= nbsp;       remainingIPCs =3D {null}

Else IF ()

Let potentialIPCs =3D {}
Else

        &= nbsp;           &nbs= p;           Let potentialIPCs =3D remainingIPCs

End IF

For the set of IPCs selected for evaluation, the algorithm next loops through a= ll potential next jobs and determines the FSIs in the potential next IPC that would need to be put on the collator (transitionOn), as well as the FSIs on= the collator and not in the potential next IPC that could be changed out (transitionOff).  The last sta= tement creates a temporary array with the same structure and contents of the colla= tor array for use in determination of the number of change-overs that would be experienced if the IPC under evaluation was sequenced next in the schedule (tempCollatorLong). 

        &= nbsp;       For each {

Let i =3D i= ndex of next <= !--[if gte vml 1]> potentialIPCs&nbs= p;

        &= nbsp;           &nbs= p;           Let transitionOn }

Let transit= ionOff }

Let tempCollatorLongik =3D collatorLongk 

The next set of statements execute an IPC change-over on the mirror copy of the collator state (tempCollatorLong) to determine the number of hard and soft change-overs that would result from this potential IPC sequencing.  The order in which FSIs in the tra= nsitionOff asset are selected for change-over according to the IF clauses is as follow= s: (1) open hoppers are filled first, (2) FSIs in the transitionOff set that a= re not used in the previous job, resulting in soft change-over, and (3) FSIs in the transitionOff set that are used in the previous job, resulting in hard change-overs.  Given a tie bre= ak scenario for either hard of soft change-overs, the FSI that is used least frequently in the unscheduled IPCs is selected for change-over.<= /span>

        &= nbsp;       For transitionOn {

        &= nbsp;               &= nbsp;       Let n =3D index of next transitionOn
       &= nbsp;           &nbs= p;            &= nbsp;           &nbs= p;   Let j =3D index of next transitionOff

        &= nbsp;           &nbs= p;           //Replace empty hoppers first

        &= nbsp;           &nbs= p;           IF tempCollatorLongik =3D 0 for some k Then {

 &nb= sp;            =   tempCollatorLongik =3D n       <= /span>//assume FSIn moves into open hopper

 &nb= sp;            =   transitionOn =3D transitionOn – {FSIn}

 &nb= sp;            =   transitionOff =3D transitionOff – {FSIj}

}//end If

//If no hoppers are empty, conduct soft change-over on least used FSI in remaini= ng IPCs

Else If  {

 &nb= sp;            =   Let tempFSI =3D    =  

 &nb= sp;            =   Find ktempCollatorLongik where { tempCollator= Longik =3D tempFSI} 

 &nb= sp;            =   Let tempCollatorLongik =3D n                =             &nb= sp;    //put FSIn in hopper k  &nbs= p;            &= nbsp;          soft_cha= ngeoveri =3D soft_changeoveri + 1 &n= bsp;        //add one soft changeover for IPCi

 &nb= sp;            =   transitionOn =3D transitionOn  - {FSIn= }

 &nb= sp;            =   transitionOff =3D transitionOff  - {FSI= j}

}= // end Else If

Else {&nb= sp;      //only hard change-overs remaining

 &nb= sp;            =   Let tempFSI =3D

 &nb= sp;            =   Find k tempCollatorLongi where { tempCollatorLongik =3D tempFSI}&= nbsp;      

 &nb= sp;            =   Let tempCollatorLongik =3D n                =             &nb= sp;    //put FSIn in hopper k

 &nb= sp;            =   hard_changeoveri =3D hard_changeoveri + 1    //add one hard changeo= ver for IPCi

transitionO= n =3D transitionOn  - {FSIn}

transitionO= ff =3D transitionOff  - {FSIj}

        &= nbsp;           &nbs= p;            &= nbsp;           &nbs= p;  }//end Else

 &nb= sp;            =             &nb= sp;     }// end For FSI transition= On
       &= nbsp;        }//end For each IPC
potentialI= PCs

 

The algorithm then selects the next IPC for scheduling with the minimum number = of hard change-overs, then soft change-overs, and has the largest sum total of= FSI frequencies in the remaining unscheduled IPCs.

Let tempIPC= s =3D

Let tempIPC= s =3D

Let   =            //recal= culate FSI frequency

Let nextIPC= =3D

Let jobLongn =3D nextI= PC        &= nbsp;           &nbs= p;            &= nbsp;           &nbs= p;                         = ; //set the next IPC in the schedule

Let collato= rLongk =3D tempCollatorLongik                 | i = =3D nextIPC    //configur= e next collator state

Let scheduleLongi n =3D tempCollatorLongik        &= nbsp;    | i =3D nextIPC    // r= ecord schedule & assignment

totalProdTime =3D totalProdTime + productionTimei  &n= bsp;         | i =3D nextIPC

remainingIPCs =3D remainingIPCs – {IPCi}             &= nbsp;    | i =3D nextIPC    // r= emove nextIPC from list

}// end For IPC remainingI= PCs

 

Step 4: Single Mode IPC Scheduling

Following Step 3,= all IPCs within the capacity constraints of the current side of the collator (i= .e., long, short) are included in each side schedule, thus we assume Step 3 has = been executed once for each side.  = At this point, there are a number of options for augmenting the current schedu= les with the single mode IPCs stored in the singleModeIPC set.  First, the single mode IPCs could = be inserted as a set at the front or tail end of the schedule, with the single= mode IPCs schedule optimized through the procedures outlined in Step 3.  Secondly, the single mode IPCs cou= ld be run independently on one of the existing legacy collators with sufficient s= ize (e.g., Collator 4).  Although = simple from a planning perspective, either approach would introduce some complexity into the delivery process as jobs within given delivery zones could be run = at vastly different times in the production week, thus requiring multiple inventory movements and staging.  Alternatively, the single mode IPCs could be re-inserted into the si= ngle side schedules by TV book zone (which would be running on one and only one side).  In this case, the sing= le mode collators would be inserted into both side schedules at the point that resulted in the minimum number of hard and soft change-overs, and also resu= lted in the minimum amount of downtime waiting due to the unsynchronized starting and stopping of previous jobs on the two sides of the collator.  Here, the total production time on= the collator would be used to determine exactly which IPCs on either side of the collator would terminate immediately before the insertion point for the sin= gle mode IPC.  <= /h3>

=  

Due t= o time constraints, this insertion module of the scheduling algorithm was not implemented in the prototype or formulated.  As a result, the current algorithm= only allows front or backend augmentation of the single mode IPCs.  However, the extension described a= bove is highly recommended for future iterations and testing. =

Step 5: Hopper Assignment=

The first sets of statements declare and initialize the key st= atic variables, arrays, and collections for use in the assignment heuristic.

Let numHoppersMax =3D numHoppersLong || numHoppersShort based = on current side under assignment

Let numFeeders =3D the predetermined number of Feeder positions allocated to this side of the collator

Let maxHoppPerFeeder =3D the maximum number of hoppers that could be assigned to a single Feeder

Let virtHoppersk =3D the average FSI tab page count= in hopper position k resulting from the scheduling module

Let maxPPFM =3D the maximum number of pages per Feeder minute = that could be assigned to a single Feeder

Let machineSpeed =3D the target machine speed in terms of book= s per minute

= //The hopperAssign array is dimensioned for f<=3D numFeeder= s, k<=3D maxHoppPerFeeder

= Let hopperAssignf, k         =3D  

Let hopperMappingk  &n= bsp;    =3D =

Let feederWorkloadf =             &nb= sp;  =3D the tab pages per book assigned to Feeder position f   //running total <= /span>

Let feederNumHoppersf     =3D the number of hopper assigned to Feeder position f   //number of items=

The next set of statements introduce hard hopper mapping constraints for the top of stack and bottom of stack FSIs (i.e., TV book, Parade are set to the bottom of the pile; magazine and comics are set to the top of the pile).  Here, this = is a departure from current operations where TV books are the last item added on= the top of the stack.  This change= is to enable the new collator to operate in single mode while ensuring the TV boo= k is inserted on one side of the stack.

'Initialize starting conditions - e.g,. tvbooks, magazine, com= ics, and parade

Let hopperMapping1         &= nbsp;           &nbs= p;     =3D {mapped to hopper position k by the initial schedul= e}
Let hopperMapping2         &= nbsp;           &nbs= p;     =3D {mapped to hopper position k by the initial schedul= e}
Let hopperMappingnumHoppersMax=   &n= bsp;       =3D {mapped to hopper position k by the initial schedul= e}
Let hopperMappingnumHoppersMax= - 1      =3D<= span style=3D'font-size:10.0pt;mso-ascii-font-family:Calibri;mso-ascii-theme-fon= t: minor-latin;mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin'= > {mapped to hopper position k by the initial schedul= e}

The next section creates a data structure that will be used to order the greedy assignment process.  Here, the data structure, sortedVirtHoppers, is a set of pairs of elements, where the first element h= olds the virtual hopper index from the initial schedule, and second element holds the average tab page count per virtual hopper position across all IPCs.  The data structure must be sorted = from largest to smallest value of the second element, and must exclude the virtu= al hoppers assigned to the TV Book, Parade, magazine, and comics). =

Let sortedVirtHoppersk, n =3D = {(s, pages_s), (x, pages_x), (y, pages_y) …} a largest to smallest sorted = two dimensional array with the virtual hopper index (s, x, y, z) as the first element, sorted by the average tab page size in the hopper across all jobs = in the initial schedule as the second element.  Hopper positions associated with t= vbook, parade, magazine, and comics are excluded.=  

//Sample sortedVirtHoppersk, n= =3D {(2, 48), (9, 40), (3, 36), (18, 32)} where the first element is virtual ho= pper 2 and //had an average tab page size of 48 pages

The next section executes the initial greedy assignment of vir= tual hoppers to Feeder positions based on the current workload assigned to each feeder and the size of the next virtual hopper in terms of average tab page count.  In each case, the algo= rithm attempts to add the next virtual hopper to the Feeder position with the min= imum current workload.  One special= case requires consideration, specifically the case where a single FSI exceeds the pages per Feeder minute constraint of the Feeders (e.g., 80 pages).  These special cases are assigned to positions on the front of either side of the collator to be post-processed = by Client Scheduling Experts as needed. 

<= span style=3D'font-size:10.0pt;mso-bidi-font-size:11.0pt;line-height:115%;mso-as= cii-font-family: Calibri;mso-ascii-theme-font:minor-latin;mso-hansi-font-family:Calibri; mso-hansi-theme-font:minor-latin'>For i =3D 1 To numHoppersMax – 4  //four hopper position already res= erved for tvbook, parade, magazine, comics

 &nb= sp;            =   min =3D 999 //initializes a min value holder for the iterations – the min= imum workload by feeder

                For j =3D 1 To numFeeders

 &nb= sp;            =             &nb= sp;     If (feederWorkloadf  += sortedVirtHoppersi <=3D maxPPFM/machineSpeed Then {

 &nb= sp;            =             &nb= sp;            =          If feederWorkloadf < min Then

 &nb= sp;            =             &nb= sp;            =             &nb= sp;            min =3D feederWorkloadf
       &= nbsp;           &nbs= p;            &= nbsp;           &nbs= p;            &= nbsp;      minIndex =3D j

 &nb= sp;            =             &nb= sp;            =          End IF }

 &nb= sp;            =             &nb= sp;     //to handle single FSIs that are greater than the tab page per book feeder limit=

 &nb= sp;            =             &nb= sp;     Else IF sortedVirtHoppersi > maxPPFM/machineSpeed Then =

 &nb= sp;            =             &nb= sp;            =          minIndex =3D 1

 &nb= sp;            =             &nb= sp;     End IF

 &nb= sp;            =   Next //end For 1 to numFeeders

         &= nbsp;      //Assign hopper i to feeder j with minimum workload

 &nb= sp;            =   hopperAssignf, k    =3D sortedVirtH= oppersi, 1  where f =3D minIndex,= k =3D feederNumHoppersf + 1

         &= nbsp;      feederWorkloadf        &= nbsp;        =3D feederWorkloadf + sortedVirtHoppersi, 2

 &nb= sp;            =   feederNumHoppersf        &= nbsp;  =3D feederNumHoppersf + 1

Next //end For 1 to numHoppersMax-4<= /o:p>

 

 

Lastly, the assignment algorithm looks to order the hoppers to centrally concentrate the workload of each Feeder, resulting in larger FSIs being placed in middle positions and smaller FSIs being placed to the outside.  A new array is intro= duced, tempAssign, to hold the sorted array as each Feeder position is processed.<= span style=3D'mso-spacerun:yes'>  This array is eventually used to r= ecord the final virtual hopper to physical hopper mapping.

Let tempAssignz =3D s if hoppe= r s is assigned to hopper position z

n =3D 3 //index for final physical hopper mapping, four hoppers already reserved for TV Book, parade, …

For i =3D 1 To numFeeders

 &nb= sp;            =   Let midposition =3D Int(feederNumHoppersi / 2) + 1=

 &nb= sp;            =   tempAssignmidposition =3D hopperAssigni, 1

 &nb= sp;            =   j =3D 1

 &nb= sp;            =   Do While (j <=3D Int(feederNumHoppersi))

                &= nbsp;           &nbs= p;   If (midposition - j > 0) Then

        =               =             &nb= sp;            = tempAssignmidposition - j =3D hopperAssigni, 2*j

        =          =             &nb= sp;  End If

        =          =             &nb= sp;   If (midposition + j <=3D feederN= umHoppersi) Then

              =             &nb= sp;           &= nbsp;         tempAssignmidposition + j =3D hopperAssigni, (2*j) + 1

        =          =             &nb= sp;  End If

        =           =             &nb= sp; j =3D j + 1

                Loop //end while loop

 &nb= sp;            =   'Reset the hopper Assignment with the sorted values

 &nb= sp;            =   For k =3D 1 To feederNumHoppersi

 &nb= sp;            =             &nb= sp;     hopperAssigni, k =3D tempAssignk

 &nb= sp;            =             &nb= sp;     hopperMappingn =3D tempAssignk

        =           =             &nb= sp; n =3D n + 1

                Next

    Next=

The resulting hopper assignment capturing the final mapping fr= om virtual hoppers output in the initial schedule to physical hoppers is store= d in the hopperMapping array.  This= array is then post-processed against to restructure the initial schedule with the balanced workload assignment.


 

Sample Results

The = graphic below presents a snapshot of the results of the  initial scheduling and assignment heuristic for the first 20 jobs and first 20 hoppers of the March 16, 2008 edition week.  Here, the green shading indicates an FSI is included in the current job, gray shading indic= ates an FSI is loaded on the collator in that hopper position but not used in the current job, and a maroon border identifies a soft change-over (note: no ha= rd change-overs are depicted in this sample).

Figure 12 Sample Scheduling Results for March 1= 6, 2008

#

1

88

104

52

0

0

47

32

91

0

24

0

0

0

94

0

46

95

14

0

0

75

2

63

104

52

0

0

47

32

91

0

24

0

0

0

94

0

46

95

14

0

76

75

3

69

104

52

0

0

47

32

91

0

24

0

0

0

94

0

46

95

14

0

76

75

4

70

104

52

0

0

47

32

91

0

24

0

0

0

94

0

46

95

14

0

76

75

5

71

104

52

0

0

47

32

91

0

24

0

0

0

94

0

46

95

14

0

76

75

6

75

104

52

0

0

47

32

91

0

24

0

0

0

94

0

46

95

14

0

76

75

7

73

104

52

0

0

47

32

91

0

24

0

0

0

94

0

46

95

14

0

76

75

8

74

104

52

0

0

47

32

91

0

24

0

0

0

94

65

46

95

14

0

76

75

9

76

104

52

0

0

47

32

91

0

24

0

0

0

94

65

46

95

14

0

76

75

10

77

104

52

0

0

47

32

91

0

24

0

0

0

94

65

46

95

14

0

76

75

11

67

104

52

0

0

47

32

91

0

24

0

135

0

94

65

46

95

14

0

76

75

12

68

104

52

0

0

47

32

91

0

24

0

135

0

94

65

46

95

14

0

76

75

13

72

104

52

0

0

47

32

91

0

24

0

135

0

94

65

46

95

14

0

76

75

14

85

104

52

0

0

47

32

91

0

24

0

135

0

94

65

46

95

14

0

76

75

15

65

104

52

0

0

47

32

91

0

24

0

135

0

94

65

46

95

14

0

76

75

16

80

104

52

0

0

47

32

91

0

24

0

135

0

94

65

46

95

14

0

76

75

17

86

104

52

0

0

47

32

91

0

24

0

135

0

94

65

46

95

14

0

76

75

18

89

104

52

0

0

47

32

91

0

24

0

135

0

94

65

46

95

14

0

76

75

19

94

104

52

0

0

47

32

91

0

24

0

135

0

94

65

46

95

14

0

76

75

20

101

104

52

0

0

47

32

91

0

24

0

135

0

94

65

46

95

14

0

76

75

 

The resulting scheduled provides t= he Sunday Insert Scheduling division with a feasible schedule that can be used= to plan the sequencing of FSI (advertisements) onto the mailroom production fl= oor from the warehouse, as well as to inform the collating machines as to which hoppers need to be active for each production job.  Lastly, the resulting visualization provides the mailroom Foreman with a visualization of upcoming forecasts to enable ad hoc change-over forecasting.

 

A second key output is depicted in= the table below, which presents the average workload per physical hopper positi= ons calculated in terms of average tab pages per book.  Ideally, the resulting hopper assi= gnment would provide discrete Feeder position workloads that are equally balanced = and equally distributed across the physical hoppers barring constraints.  The sample data below for the Marc= h 16, 2008 edition week presents a properly assigned and balanced assignment as confirmed by client experts.

 

 

 

 

Figure 13 Sample Assignment Results for March 1= 6, 2008

 

Business Strategy / Business Approach

The decision to invest several million dollars for = new equipment must be based on sound financial analysis in addition to the appa= rent requirement to replace aging equipment. Our approach is based on the calcul= ated production times, notional labor costs fully burdened with overhead, and la= bor productivity rates as specified by contract. Sales trends are shown in Figure 12 and used for projections.

1.  Current cost calculation

Existing equipment, run times, availability rates a= nd labor skill mix are used to calculate current production costs for a single week.

·         Identify number of shifts worked per week

·         Identify number of workers per shift

·         Estimate average fully burdened labor rate

·         Identify hours worked per shift

·         Calculate the weekly operating cost

2.  Simple projection cost calculation

This projection uses promised (manufacturer cla= imed) run times, feeder capacity and equipment availability rates with the current labor skill mix to determine expected costs to operate the new collator.

·         Use the estimated number of shifts to be worked per week<= /span>

·         Estimate the number of workers required per shift<= /p>

·         Reuse average fully burdened labor rate

·         Reuse hours worked per shift

·         Calculate the simple weekly operating cost

·         Calculate the simple weekly operating cost savings=

 

<= ![endif]>

Figure 14 Weekly sales trend

3.  Optimistic (best case)= projection cost calculation

Decreased labor costs due to lower required ski= ll levels are combined with minimum calculated production times (at maximum ra= te and availability) to determine the best case scenario. This calculation does not consider the effects of worker learning.

·         Lowest calculation of number of shifts to be worked per week

·         Identify number of workers needed per shift

·         Constant fully burdened labor rate to reflect lower average required skill mix under higher demand

·         Reuse hours worked per shift

·         Calculate the minimum weekly operating cost

·         Calculate the maximum weekly operating cost savings

 

4.  Pessimistic (conservat= ive) projection cost calculation

This calculation provides a cost projection bas= ed on production run time least improvement and worst-case labor cost, and assumes there will be a 90% learning curve with the new, more automated equipment.<= /p>

·         High estimate of number of shifts worked per week<= /p>

·         Identify number of workers per shift

·         Constant fully burdened labor rate to reflect lower average required skill mix under higher demand

·         Reuse hours worked per shift

·         Calculate the maximum weekly operating cost

·         Calculate the minimum weekly operating cost savings

=  

=  

 

Collator shifts/
 week
<= /p>

Workers/ collator shift

Fully-burdened $/
 worker-hour

Hours/
shift

Operating cost $/
week

Savings/ week

Current conditions

31<= o:p>

10<= o:p>

 $45.00

7

 $97,650

 

Simple projection

15<= o:p>

10<= o:p>

 $45.00

7

 $47,250

 $50,400

Conservative projection

20<= o:p>

10<= o:p>

 $45.00

7

 $63,000

 $34,650

Best-case projection

12<= o:p>

10<= o:p>

 $45.00

7

 $37,800

 $59,850

Table 7 Cost calculations

*Simple projection - no change to current labor = cost, lower bound of required shifts per week

**Optimistic Projection - increased labor costs,= lower bound of required shifts per week

***Pessimistic Projection - increased labor cost, upper bound of required shifts per week

5.  Calculate financing co= st and payback time

This calculation determines the amount of time = to realize a return on investment for each projection without financing. If the purchased were financed, at a 3.3 million dollar cost (given a nominal principal amount of ten million dollars to be financed at six percent for t= en years) the return on investment is calculated for the pessimistic projectio= n.

·         Estimate capital investment cost

·         Estimate financing rate and term

·         Calculate monthly payment

·         Calculate cost of capital (total payment over full term)<= /span>

·         Calculate break even time for each operating cost calculation

 

Table 8 Financing cost and break even time

Recommendations

Future project possibilities include l= eaning the entire production and distribution process. Opportunities identified du= ring this project are:

·         Pre-staging FSI pallets by splitting invento= ry according to hopper collator assignments, containerizing distribution palle= ts to increase distribution truck flexibility

·         Web-enabling the order placement and fulfill= ment process to reduce customer wait time by an order of magnitude

·         Specifying requirements for FSI printers to deliver advertisements when needed and not before

·         Eliminating the use of collator 4 by alterin= g TV zone constraints that would force hard undesirable changeovers on the new equipment

·         C= alculate effects of new sales model. What happens to the business case if:

·         Advertising sales increase or decline?

·         Number of customers increases or declines?

·         If more cus= tomers buy smaller orders?

Conclusions

The proof of concept was confirmed thr= ough implementation of a prototype algorithm that will increase automation of the planning process. The result of our analysis indicates that the best hopper configuration would be to run the new machine 100% in dual mode, and the existing collator #4 only to cover those IPCs too large for one side of the= new machine.  This solution assume= s that TV book zones will govern production scheduling. Under this model, the enti= re average production run could be completed in a little as 12 shifts (worst case using 3/12/08 data is 13.5 shifts). With increased automation of the new equipmen= t, an optimal operating labor cost savings of 61% will result. At this savings= , a $10 million investment break even time is only three years two months.

We verified that the new equipment sh= ould consist of 80 hoppers to achieve optimal results.

We verified that the best hopper configuration consists of approximately 35 hoppers on one side and 45 on the other. An exact balance by run time is achieved with a 36/44 split. This ve= rification was tested using a calculation of balancing the opportunity for hopper load= ing in comparison of single and dual modes to minimize the total run time.

The solution is sub optimized to allo= w for TV zone, pallet splitting & distribution constraints.

We used a weighted average of workers n= eeded per shift given a target line speed of 225 books per minute and a physical constraint of 16,000 pages per feeder minute. Ideally, the resulting schedu= le would optimize production time for maximum line speed for a minimum number = of workers.

Bibliography

GEIA-632, Processes for Engineering a = System, Information Technology Association of America (GEIA Group), 2003

Albright, S. C= hristian, VBA for Modelers : Developing Decision Support Systems Using Microsoft® Excel, South-Western College Pub, 2000

Buede, Dennis, The Engineering Design of Systems: Models and Methods, Wiley, 2000

Neiss, Alan M., Ins= ert Scheduling System synopsis, January 20, 2009

Appendix

= A.      = Shiva Scheduling Sequencing Engine

<= ![if !supportLists]>B.      = IDEF0 diagram ISS functional architecture

C.      = Project schedule

D.      = Business case analysi= s calculations

E.      =   QFD-HOQ

F.      =   Customer input documents and reports

G.      = Morphological analysis

 

 



[1] Alternatively, if under staffing or overstaffing concerns question the use = of the weighted average, the use of weighted percentile is recommend (e.g., nu= mber feeder positions =3D 85th percentile of job requirements)

------=_NextPart_01C9D02C.2AB56840 Content-Location: file:///C:/EA89CAB4/CRITFinalReport_files/item0001.xml Content-Transfer-Encoding: quoted-printable Content-Type: text/xml ------=_NextPart_01C9D02C.2AB56840 Content-Location: file:///C:/EA89CAB4/CRITFinalReport_files/props0002.xml Content-Transfer-Encoding: quoted-printable Content-Type: text/xml ------=_NextPart_01C9D02C.2AB56840 Content-Location: file:///C:/EA89CAB4/CRITFinalReport_files/themedata.thmx Content-Transfer-Encoding: base64 Content-Type: application/vnd.ms-officetheme UEsDBBQABgAIAAAAIQCCirwT+gAAABwCAAATAAAAW0NvbnRlbnRfVHlwZXNdLnhtbKyRy2rDMBBF 94X+g9C22HK6KKXYzqJJd30s0g8Y5LEtao+ENAnJ33fsuFC6CC10IxBizpl7Va6P46AOGJPzVOlV XmiFZH3jqKv0++4pu9cqMVADgyes9AmTXtfXV+XuFDApmaZU6Z45PBiTbI8jpNwHJHlpfRyB5Ro7 E8B+QIfmtijujPXESJzxxNB1+SoLRNegeoPILzCKx7Cg8Pv5DCSAmAtYq8czYVqi0hDC4CywRDAH an7oM9+2zmLj7X4UaT6DF9jNBDO/XGD1P+ov5wZb2A+stkfp4lx/xCH9LdtSay6Tc/7Uu5AuGC6X t7Rh5r+tPwEAAP//AwBQSwMEFAAGAAgAAAAhAKXWp+fAAAAANgEAAAsAAABfcmVscy8ucmVsc4SP z2rDMAyH74W9g9F9UdLDGCV2L6WQQy+jfQDhKH9oIhvbG+vbT8cGCrsIhKTv96k9/q6L+eGU5yAW mqoGw+JDP8to4XY9v3+CyYWkpyUIW3hwhqN727VfvFDRozzNMRulSLYwlRIPiNlPvFKuQmTRyRDS SkXbNGIkf6eRcV/XH5ieGeA2TNP1FlLXN2Cuj6jJ/7PDMMyeT8F/ryzlRQRuN5RMaeRioagv41O9 kKhlqtQe0LW4+db9AQAA//8DAFBLAwQUAAYACAAAACEAa3mWFoMAAACKAAAAHAAAAHRoZW1lL3Ro ZW1lL3RoZW1lTWFuYWdlci54bWwMzE0KwyAQQOF9oXeQ2TdjuyhFYrLLrrv2AEOcGkHHoNKf29fl 44M3zt8U1ZtLDVksnAcNimXNLoi38Hwspxuo2kgcxSxs4ccV5ul4GMm0jRPfSchzUX0j1ZCFrbXd INa1K9Uh7yzdXrkkaj2LR1fo0/cp4kXrKyYKAjj9AQAA//8DAFBLAwQUAAYACAAAACEAlrWt4pYG AABQGwAAFgAAAHRoZW1lL3RoZW1lL3RoZW1lMS54bWzsWU9v2zYUvw/YdyB0b2MndhoHdYrYsZst TRvEboceaYmW2FCiQNJJfRva44ABw7phhxXYbYdhW4EW2KX7NNk6bB3Qr7BHUpLFWF6SNtiKrT4k Evnj+/8eH6mr1+7HDB0SISlP2l79cs1DJPF5QJOw7d0e9i+teUgqnASY8YS0vSmR3rWN99+7itdV RGKCYH0i13Hbi5RK15eWpA/DWF7mKUlgbsxFjBW8inApEPgI6MZsablWW12KMU08lOAYyN4aj6lP 0FCT9DZy4j0Gr4mSesBnYqBJE2eFwQYHdY2QU9llAh1i1vaAT8CPhuS+8hDDUsFE26uZn7e0cXUJ r2eLmFqwtrSub37ZumxBcLBseIpwVDCt9xutK1sFfQNgah7X6/W6vXpBzwCw74OmVpYyzUZ/rd7J aZZA9nGedrfWrDVcfIn+ypzMrU6n02xlsliiBmQfG3P4tdpqY3PZwRuQxTfn8I3OZre76uANyOJX 5/D9K63Vhos3oIjR5GAOrR3a72fUC8iYs+1K+BrA12oZfIaCaCiiS7MY80QtirUY3+OiDwANZFjR BKlpSsbYhyju4ngkKNYM8DrBpRk75Mu5Ic0LSV/QVLW9D1MMGTGj9+r596+eP0XHD54dP/jp+OHD 4wc/WkLOqm2chOVVL7/97M/HH6M/nn7z8tEX1XhZxv/6wye//Px5NRDSZybOiy+f/PbsyYuvPv39 u0cV8E2BR2X4kMZEopvkCO3zGBQzVnElJyNxvhXDCNPyis0klDjBmksF/Z6KHPTNKWaZdxw5OsS1 4B0B5aMKeH1yzxF4EImJohWcd6LYAe5yzjpcVFphR/MqmXk4ScJq5mJSxu1jfFjFu4sTx7+9SQp1 Mw9LR/FuRBwx9xhOFA5JQhTSc/yAkArt7lLq2HWX+oJLPlboLkUdTCtNMqQjJ5pmi7ZpDH6ZVukM /nZss3sHdTir0nqLHLpIyArMKoQfEuaY8TqeKBxXkRzimJUNfgOrqErIwVT4ZVxPKvB0SBhHvYBI WbXmlgB9S07fwVCxKt2+y6axixSKHlTRvIE5LyO3+EE3wnFahR3QJCpjP5AHEKIY7XFVBd/lbobo d/ADTha6+w4ljrtPrwa3aeiINAsQPTMR2pdQqp0KHNPk78oxo1CPbQxcXDmGAvji68cVkfW2FuJN 2JOqMmH7RPldhDtZdLtcBPTtr7lbeJLsEQjz+Y3nXcl9V3K9/3zJXZTPZy20s9oKZVf3DbYpNi1y vLBDHlPGBmrKyA1pmmQJ+0TQh0G9zpwOSXFiSiN4zOq6gwsFNmuQ4OojqqJBhFNosOueJhLKjHQo UcolHOzMcCVtjYcmXdljYVMfGGw9kFjt8sAOr+jh/FxQkDG7TWgOnzmjFU3grMxWrmREQe3XYVbX Qp2ZW92IZkqdw61QGXw4rxoMFtaEBgRB2wJWXoXzuWYNBxPMSKDtbvfe3C3GCxfpIhnhgGQ+0nrP +6hunJTHirkJgNip8JE+5J1itRK3lib7BtzO4qQyu8YCdrn33sRLeQTPvKTz9kQ6sqScnCxBR22v 1VxuesjHadsbw5kWHuMUvC51z4dZCBdDvhI27E9NZpPlM2+2csXcJKjDNYW1+5zCTh1IhVRbWEY2 NMxUFgIs0Zys/MtNMOtFKWAj/TWkWFmDYPjXpAA7uq4l4zHxVdnZpRFtO/ualVI+UUQMouAIjdhE 7GNwvw5V0CegEq4mTEXQL3CPpq1tptzinCVd+fbK4Ow4ZmmEs3KrUzTPZAs3eVzIYN5K4oFulbIb 5c6vikn5C1KlHMb/M1X0fgI3BSuB9oAP17gCI52vbY8LFXGoQmlE/b6AxsHUDogWuIuFaQgquEw2 /wU51P9tzlkaJq3hwKf2aYgEhf1IRYKQPShLJvpOIVbP9i5LkmWETESVxJWpFXtEDgkb6hq4qvd2 D0UQ6qaaZGXA4E7Gn/ueZdAo1E1OOd+cGlLsvTYH/unOxyYzKOXWYdPQ5PYvRKzYVe16szzfe8uK 6IlZm9XIswKYlbaCVpb2rynCObdaW7HmNF5u5sKBF+c1hsGiIUrhvgfpP7D/UeEz+2VCb6hDvg+1 FcGHBk0Mwgai+pJtPJAukHZwBI2THbTBpElZ02atk7ZavllfcKdb8D1hbC3ZWfx9TmMXzZnLzsnF izR2ZmHH1nZsoanBsydTFIbG+UHGOMZ80ip/deKje+DoLbjfnzAlTTDBNyWBofUcmDyA5LcczdKN vwAAAP//AwBQSwMEFAAGAAgAAAAhAA3RkJ+2AAAAGwEAACcAAAB0aGVtZS90aGVtZS9fcmVscy90 aGVtZU1hbmFnZXIueG1sLnJlbHOEj00KwjAUhPeCdwhvb9O6EJEm3YjQrdQDhOQ1DTY/JFHs7Q2u LAguh2G+mWm7l53JE2My3jFoqhoIOumVcZrBbbjsjkBSFk6J2TtksGCCjm837RVnkUsoTSYkUigu MZhyDidKk5zQilT5gK44o49W5CKjpkHIu9BI93V9oPGbAXzFJL1iEHvVABmWUJr/s/04GolnLx8W Xf5RQXPZhQUoosbM4CObqkwEylu6usTfAAAA//8DAFBLAQItABQABgAIAAAAIQCCirwT+gAAABwC AAATAAAAAAAAAAAAAAAAAAAAAABbQ29udGVudF9UeXBlc10ueG1sUEsBAi0AFAAGAAgAAAAhAKXW p+fAAAAANgEAAAsAAAAAAAAAAAAAAAAAKwEAAF9yZWxzLy5yZWxzUEsBAi0AFAAGAAgAAAAhAGt5 lhaDAAAAigAAABwAAAAAAAAAAAAAAAAAFAIAAHRoZW1lL3RoZW1lL3RoZW1lTWFuYWdlci54bWxQ SwECLQAUAAYACAAAACEAlrWt4pYGAABQGwAAFgAAAAAAAAAAAAAAAADRAgAAdGhlbWUvdGhlbWUv dGhlbWUxLnhtbFBLAQItABQABgAIAAAAIQAN0ZCftgAAABsBAAAnAAAAAAAAAAAAAAAAAJsJAAB0 aGVtZS90aGVtZS9fcmVscy90aGVtZU1hbmFnZXIueG1sLnJlbHNQSwUGAAAAAAUABQBdAQAAlgoA AAAA ------=_NextPart_01C9D02C.2AB56840 Content-Location: file:///C:/EA89CAB4/CRITFinalReport_files/colorschememapping.xml Content-Transfer-Encoding: quoted-printable Content-Type: text/xml ------=_NextPart_01C9D02C.2AB56840 Content-Location: file:///C:/EA89CAB4/CRITFinalReport_files/image001.png Content-Transfer-Encoding: base64 Content-Type: image/png iVBORw0KGgoAAAANSUhEUgAAAuQAAAFBCAIAAACw/y3yAAAAAXNSR0IArs4c6QAAAAlwSFlzAAAO xAAADsQBlSsOGwAAqbpJREFUeF7tnQeAFNX9x/f2bq9XOHo7BBEVe+8gJoIYNQZjr/iPIsYSIWg0 ahITQyBqNEo0EDUmUSN2CWhEiL03QKlydI6D631v9/6fmbc3t31nd2fv9u5+E3LOzr73e7/3ndmZ 7/zaS2lra7PJJggIAoKAICAICAKCQLIiYE9WxUQvQUAQEAQEAUFAEBAENASErMh1IAgIAoKAICAI CAJJjYCQlaQ+PaKcICAICAKCgCAgCAhZkWtAEBAEBAFBQBAQBJIaASErSX16RDlBQBAQBAQBQUAQ ELIi14AgIAgIAoKAICAIJDUCQlaS+vSIcoKAICAICAKCgCAgZEWuAUFAEBAEBAFBQBBIagSErCT1 6RHlBAFBQBAQBAQBQUDIilwDgoAgIAgIAoKAIJDUCAhZSerTI8oJAoKAICAICAKCgJAVuQYEAUFA EBAEBAFBIKkRELKS1KdHlBMEBAFBQBAQBAQBIStyDQgCgoAgIAgIAoJAUiMgZCWpT48oJwgIAoKA ICAICAJCVuQaEAQEAUFAEBAEBIGkRkDISlKfHlFOEBAEBAFBQBAQBISsyDUgCAgCgoAgIAgIAkmN gJCVpD49opwgIAgIAoKAICAICFmRa6CHILBiVoqxTV5YaumsShdO1mTPWhG/VKWm1QpGrZc3Wu2w WaOUhVhps1KKWoG8gVHi5h71aZAOgoAgYBIBISsmgZJmSY0Az58J8zo0XHr1SGuevEk96cjKAUsU j/mkQS06tSPDYKJF0szdhK7SRBDojQgIWemNZ73HzXnFYo2pTFqwqY1t04JJfFi6aJm11pXuBppm 4fAmcMH0n7lcA0zfkgU1f7XHz9W0mzvecviTcO6Wz1EECgI9BwEhKz3nXPb6mXj4Scm0Jdrzbcm0 Eg0RX7eLj4+i/asVysnj724wvAX+RhqPEJ8u/t4KU96LsAoEG0U/xd5fGHYTz3gLPVOZPHnyyKuX aq3nTTDndCqZOFXneGs0iucnrX0Ub/+Jr8UmBFZhwA86E6bmr3ZwYP3PVoRTGeGnEd/cg56PEKfJ 97C3eyvk6e71P2sBQBBQCBhvVrIjCHRfBJbP7PhBe70yMyH1jcfo4rEf2FQT705Gd0/v0N8FfqN3 UYfbx/b95I2rtz6hB1F2Dp9NTSHwi2BzmTRzpnf/9tl76RGgoEewauqrmH4siEa+k/VT1wuUoOAH QX/mct9B9H7eiobWIdypDLiqEzt3feIhTlOo0xrydHffX6RoLghYjIBYVoS19gQExs/tuN9rtoSo IljVQ9cjYNVGTAulC+foETCex7HPs0T5JdSmHnt6l/FTtP15i1doR5RbauaU8eawDVRA9fMhGbql aMXDmr3E5+nvGdKnx5K5c5d4lNZke4xMgcroUKlNWTQmTZ3IKO2bZxytuxrYF5B5EzTzSjiswkzf x3OngJy3uHRaOLVD62CMEwrJIIpYMXfPiW4/H55ZrIhwmgJPa6jTbe7ykVaCQG9AQMhKbzjLvWGO uvOng1UQMGkytHTSghk6p/D4Arywav8myFceX4x3TIgXW4mOqwRTQCnDJBSTaHdElW5cpenXflwR DC+24ss1oj3tPOl9aI2XNDVwByDTZnfQNH2YcFgF08MjsJ0bmYlMiaiDoUTgqYwIRWxzDzGLkKcp xGn1XGABpzui1tJAEOg9CAhZ6T3nuhfMVIWreNtIrJ+0Co94eJQ2ko/3wWAr0XGV4Br6zESxExPk a9yokqgm7Osx849ijVZaVEN3eeMumXuo0xrb6e5yDEUBQaDzEBCy0nlYy0iJQsCvdknJqHH+I6m4 UVvpskV61GmkTUlYevXDK/SWHv+DLkE5iJSHx/MG3S7Nw1Ym6PYW0z6gcLp4YoUV+8KC4pmZXwxK 8FSZIChEmnbo7/0AaXf9aHQmNFbt4oKB79/LqD4TWu0wOsQ+MRM9o5i7ZxY2dQGGPE3+p7VdiVDH TSgpTQSBno6AkJWefoZ7w/wUSzCs6Cpj1+PEUN/pSTHtgRkmIPHp5Z0B7HmWBhfn6aXJj5Or+NYt U+4eTeT4GRpv6Zhp5IJpZrOBIoCiBvbA6Alwmblcp0khsfL7ypPn4xkoSC9vJ1YwtcPoYOKUxt4k urlrswh1mkKd1lDHY9dZegoCPQ0BISs97Yz2yvkQ8hCQSNIefjF+ruGsmbRgeWCWTXDAkNjh45m5 oKObj7hN+qgdJV08j7W4uYot2IR0ZsDLt89McWaEqkHSQZ08po24Lg3trd8v56p94JBYwVZCg+83 RcwQ6oyFUTuMDnHNLVLn8HP3Ph/tswhxmkKd1pCnO5Jm8r0g0GsQSMHx3msmKxMVBBKMgKdQSBgK kWAFRLwgIAgIAj0RAbGs9MSzKnPqIgTak2vN5ix3kZoyrCAgCAgC3QwBISvd7ISJukmKgF6BVAuW wROQgOLwSTprUUsQEAQEgU5BQNxAnQKzDCIICAKCgCAgCAgCsSIglpVYkZN+goAgIAgIAoKAINAp CAhZ6RSYZRBBQBAQBAQBQUAQiBUBISuxIif9BIHgCBC9Qnl8Smd4lZzVC2mYKEHrJdFXgP9QahAt UMZfqnfNDv/1ok2estKFsxaWmmwbvplvARGjKozSPv4Rgi8DHR65kKNaN+v4JyYSBAFBwA8BISty SQgCliKw4uE1s5dMXLZ4bEecrVbwdebMmfqCh2Y3avaHKyzHKjNTR5Xwd+bYEl+GM8Gmr8uor7M4 zvQKSV4yShdOv9oWZd3+ELOixK//is968HHpmqVx1/LXiErHXDctWDXBw9sYNIaSfBbO2uwplnaC gCBgHgEhK+axkpaCQGQEeFRSF790zbxFy9qpiZbQPHPKjLE2VfPf1KY95n1oiE8vLBPTbRPHr5g1 3TZDr6Tm2bT6/14Paq0k27w50ZowYBJhhjalvaFPcFISfnKmRlgxa8I872o2JdPmL5ikz5UlFWyh kQsp3MJZm5qANBIEBIHoEOhY7l72BAFBwHoEtAKnunlh+UwfK4NX4VN/64NaItFYaS98Sx+NtRqz QaT5iNOL/SrhgZK9qv2isncLQ6yuWns75BhCfJcGVJ2DKeOtjZcGqjuC2/t4T6ZD53atgkpuR86Y RjgQkdkxEXXbDCXU+stCJAoCgkBUCIhlJTpuJ60FgegQ0M0qszXzR8lYW4e1RfM6jFP+mk0LbFdP 9zN/eNkeIrT01Wb8XE0aSzSn+MaElIydZDih0GicvqpPMMl6fXztka3Vvtfr8bZrOXXRSD08RrPe zJuweIpiFqzhM90237O/WPvea8Na4buMkXLUGJPT5C+aqpMiUFg1QdMZTT0SVixeNcnYb9dZfaeZ QViCJ9iZ8FZP00/XyXsgzTmmw62tatk+kXaO1L5GQ3TnWFoLAoJA4hGIitpIY0FAEIgGAcOs4jFS dBgflEkhrHnAM1D4lsG18RgsDPE+xhSfo/46eFlDvIwcXup7H/Wy/wSxonhbh7zUNFr6ym+3t7T3 Wj5TN3voiPlJ8jWz+CLgLTR4u/YWfi3FqBLNhS1tBYHORkAsK4nngzJCb0VAM10sNdZI1lYd7oix 1RbHa2ubvQYrSGBijE9QR9iWIaDV+3iZbLTForWx0WjR1PntYS5BJGvmBk/sKyG+3us7o74WyqKp 1m7U8A4CDoyZDRWZ0t6S3t7mEc0k0rERk2ybMl6ZWdhftWDG+PBXkSfByFs9zQCjInh8kpIoM6zN 0Lul16x767Uq8xYEkhwBIStJfoJEve6LgO4BUq4etS2fqdY/7sg41pfbtV39sOar8HpSEyOqCEGk lt69eCL7+X4mTm13o2jeFcbWPEC6SyqkZN84Uz9rg+4aWuNRzcudo4kLDAgOkfITgsNofEGP69U9 ViuWLRqnFlhatXFhh9Ke6WrUS/l3jI2Z6Xp5qWfoRDDuKi1oSN90a4sm2qeldTHF3fdyFc0FgaRG QMhKUp8eUa77IqDlK0/yNQi0R47oD9v2LJ0gaSgdj/lILb3hGT9l5lJv2qMn47YbQXTTyhwvC0Vw yZp5oz2pWOM37UE2+spHWsAJ5hAj6dibjURvWNH5U7v8FbMw3HhCe9B0zeI1U9tNKYsWBZpVxs8g 9WeCUWEG7aAjmsHIW71gCdK6rUvnRD4T8Zp1973gRHNBoGcj0Nl+JxlPEOgVCGj5KAHpMR2xFwFZ MMGCOgxTgOceFCAvAEkjC0brEZh85CsgmA5KQHu7DnEeWd6RKd77QcJTIkWsKFNT+93VS1fvJKCg KHaYSNp7t+vrEzjjpUDHODMXqAwl3xAb31n3iutTJikIdDMEZCHDns1FZXaCgEJAL80yX7Jd5HoQ BASBbomAkJVuedpEaUEgKgSEqkQFlzQWBASBZENAYlaS7YyIPoKAtQhoqTCUMzFygKyVLtIEAUFA EOgEBMSy0gkgyxCCgCAgCAgCgoAgEDsCYlmJHTvpKQgIAoKAICAICAKdgICQlU4AWYYQBAQBQUAQ EAQEgdgRELISO3bSUxAQBAQBQUAQEAQ6AYHuQFZ8amUbhaCiAKdDQHvv0tJS8/3JpNDLYc0KLIse XIing/kR/FuGVE9JDgaI/xz1luZVjl1X6SkICAKCgCAgCCQYge5AVnyqW2mrxUa3UWzb5ql5vtym lw1dMWv6suhkaK3Hz+2kIhWh1Vvx8JrZOgBeJbT0z4Fz1Oc3fu7sNX6V3KOft/QQBAQBQUAQEAS6 FoHuQVb8McJiMFmV/243KHiMHn4f9X7tNc7b+YZWBp2F6zVjiV5EXN/0/v7WCc/n6Yt0OcpMEbzN 5FmzdOuL3+bX2BjO09RXW8+kJk9W6i1s182rqLi2tlvQ68V/jkaj8VMUP5NNEBAEBAFBQBDotgh0 D7LCw9uLU2hgLx03u61tbglLsyqjyew10xeWaiu1en30nBSWlp1vm676aw/+kmmzZ2KZ0C0SE+d7 Kprb9AXmPHXGNy1YxSpphrTZ45b6nF+9trdvm/ljV/m2MTp4N2ZxNqWePnagtvqklizR1Zth82ms yWPlNW1RE336foD4z7FD35Kx7VPrtteoKC4ICAKCgCDQyxHoHmTF8HoYjhh9LTJtpbJ5E3QWMmEe a8r6ffR6Yk9bokjJ2Dk+cScltmW6aYVF1FRjJVZtBjdQ69QbW7A2JV7r2/pcUd6NYUk2pa2uRKC2 YRpr+mxcZSwhFwgIHCzoHFmvbtVGRpNNEBAEBAFBQBDorgh0D7ISCl1oRMfSbHPH+330cI6Fkw2C oq0067URHGLTTSs+y7+1NzBsErCK0Ap47BZh2nj1HT9X50zLx2mLzQbVNlRjjoehHfiXQs3Rm+J0 14tU9BYEBAFBQBDo3Qh0c7Ji2Cp0a0WH6cJjvNDObcm0+VMXebxIKXPGakXHoSF6zMr4KeOUO2Wx bWag+cGQNmeVj2XF+4Ix08Zob4SsTLDN1rTwVb5DrK6eHpSj24z0xvpMQnp0gsyxXZyX76h3X+ky e0FAEBAEBIFui4CU24/r1ME/Hh61hBAUYmUXT1GxKAncVsyaZZsb1SDR90ig+iJaEBAEBAFBQBCI BYHubVmJZcaW9jGsIxNWLZiRYKaC4uNnjJ0TTaGZFbPmjO0EtSyFVIQJAoKAICAICAJ+CIhlRS4J QUAQEAQEAUFAEEhqBMSyktSnR5QTBAQBQUAQEAQEASErcg0IAoKAICAICAKCQFIjIGQlqU+PKCcI CAKCgCAgCAgCQlbkGhAEBAFBQBAQBASBpEZAyEpSnx5RThAQBAQBQUAQEASErMg1IAgIAoKAICAI CAJJjYCQlaQ+PaKcICAICAKCgCAgCAhZkWtAEBAEBAFBQBAQBJIaASErSX16RDlBQBAQBAQBQUAQ MEFWWPamY/MUe+84pg6wRk5AGfhgxwRwQUAQEAQEAUFAEBAEokPABFmx2SYt2NTm2fRV9FbMmmBb rg4st81ZWBpkSNjMyKuXRqeLtBYEBAFBQBAQBAQBQSAAAVNkxb9XydhJqzYqijJ+7pJpJYG4rrBN adu0YJIA3v0QiMWQZvSJZpXF7geNaCwICAKCgCDQNQiYIitLrx6pHEGTlRWlZNqS+bbp6lDwx9P4 8eM7bT64mzx66UOumOX9KawW4qgKCk+0hrTShXOUoW3TglXB7Wyddi3IQIKAICAICAI9EQFTZMV4 enVYUeAruhto09g5prlBgvArmTZ73KJlOovSuMriVVMnliRoqBjFdnM6FcmQxrWgewdtJaPGjRuV ZNjHeMqkmyAgCAgCgkASIWCKrPjp6/3s5fnU9bMZP8VgK+1cpd0xoZiUfzyw5/P0RUp3JuRlOMI0 M5nPFno0kp9O+ZzD6A1pHp44a87YGTprkU0QEAQEAUFAELAQgVjISsm0+VMXeRxDKXPGzg8WtGKh iiZEtbMVzSExG3UMx0Tb7DXTletqpuaowE+xeEXHt7PHaSHApQunL5qqRRBvmrpINV46bnZbmzIW WLQlO53ymWZMhjT43+IpQeOXLIJQxAgCgoAgIAj0WgRMkJUgMbTtTiCe8Or5ZHgCvIAMdixROCsy ULps0bgpGscoXbN03gTdWDJh3tI1Gv+YNFZTU22la2zqE/4N9VE5jkomTrUFNLZI46SnU6HnGdmQ plmmFk+xlt5ZhLuIEQQEAUFAEOgBCJggK91ilhoZePhhD1fRWIhuSdG3ABNJyVjFSTROo1MWmwp5 get4WEwCppz0dCrknCMa0lY8TJK6hxx2dQBTAs6ciBQEBAFBQBDoagR6ClmxQQbmzVN2Fc3QM9um LCtGBpMX0Ma3c1ZplhXjYTxy0dQEurSSnk55EIrekDZ+bjsvbDe0dfVVLeMLAoKAICAI9CgEUnjO 9KgJJfFkCOvQiul5LD3ap3m6tsSILBn18OSNM3Cp4VF5eBTJNZ5vJ02aNG42Hzms19jTmk4rIQRY b5zEcxXVBAFBQBAQBAQByxAQsmIZlCJIEBAEBAFBQBAQBBKBgIesbD7x8uav1z87a7LDkd7a6kxN TXWkO1qam+32NLfbxUen05mRkdHY2JiTndXibOWIy6Udd7laHQ5HRlamu7HR5m6zpaTYbG10djlb 2GfH1tbmanE6aNDmdrW0aHNIsTU5nS8999ozWWNGvPtkImYlMsMjoE53J6OUcfC+cro7GXMZThAQ BASBnoGAh6ysyz9+TM37Zbt3OltacQxlZKS73DZ7Spo9ta21tQVPUZstNTPDwY7bDetwp6bZ3W5b m7vN5W4hLmTAwCHuqupWZ3NamqPV2ZKWkeVsbISlOLKzWp2tba2u9KysVleLq7UVQuNqbWlwu2bf 9avbn9/AoD0Dx+41C073n4cex0nVqSP0Uv+P+muzOZ0u9tPS7SnaedfiUfQmKrxJ9dG+SNGYKf+z t6W4M/LSXU2tXBt8bku1ZWRltLrczgan3ea22xzf1n2zdPNr6hrrXkCJtoKAICAICALJgIBPgG1R Uf7QoYOGDRvev3+xs3Z1TfmKnJzUvsU5DXvfsTvXFxbl5+U50mx78nJtRUUFhYWZOVktgwb17zeg z5a1/9lbt8bRr68z013RtL50y5ImR2X2oMFpBYWZxX1SizJbM9sy+/TL6t8/c0D/vKFDU7IyW5qa LZ5/sEVtfIZQhfhDlOMvLS21WJ/kFpdXl8+//LrC3PoC7V9dQUZlXsbevIzy7H0Gjhw9eGTKlnT7 9gxHWVb23pycqvy86rz8mryCuoJ+Lf37NQ7q2zioqH5gXm2/nIr87PJ8+4b09F15hdXFGWU5gwqG 5zbmFzv79UsZ0N8xaGi/of0c/ZMbDNFOEBAEBAFBIKkR8LGsNDbVYFOxp2Q6W6u/fv+vDfW7s/OG 5uQPrNzxaYvTmVc4tK21pbWpMrtwqD0tt81ZX1+7Mzu3f2pGYV3F1jZ3XXb+8OaGPY7MvIqytQVF g/KKx/AC7myqTNG8RTaXq2H46DOL+o3m7b2qumLGdTfe9Wqpla/a3mGnQUNQ1cH2UFZ/HtNpIasd kbWoQIZ1QGp1GD3Jry4tLSkpifOawsjxeN/TbK3YRGyuFndrM7YU234nDLJlZzfXtfbdpyh3SEFW W2ZGgcORl5qZk5memZ6akWq321uaG1aveddta8prc2BKczU43a0Od7Mtq83hrEuvr2mr29PgbHO2 VDvry+tyivPyBuQVjih679O3//7t38SyEudZk+6CgCAgCPRaBHwsKykpfIRLuFNTM4eMOGbgsGOy svvn9xk9fL9Jw0Z/Lyd/cE7R6IIBBxQW75eemdPqauw/5NCCfmNy8/pn5RUXDjwoM3dA4YBxA4Yf O2zM5MJ+BzhbGmsqtqXYM+0pudk5gx0ZhamOHM1xgPvAnorzKOGgt5fRD5a/7F2Pn4K386gxb2GB /fBT61gpcLktupX/Vsyavswa2JqqWlqaW9syUvsfPODoq4754YM/PPOs7PPHfnfZggu2fFW2z6Ej xv/ykBNvOPjEK0848sdHHHzWQQeefsD+3xu7/8R9q/euaxu+uqnw7bYhn7QMfaOh8JPWQR+kjn4v bezyE38+9Ob/3nDNk9MueOi8M355+kGT90vPTW11OfVIJtkEAUFAEBAEBIEYEfAhK7w6azEHNheB CANGnDB63Dn7H37e4KGHDxl56j77n7HfIRfud+h5ow+7dMi+p48+6MfDD7qiT8mUfiNOG7DPxP4j zxg4+pyh+5818qAfFQ87btRB5+xz0I/HHHHZQcfPGHvklfscdO6oA88+8PBp+YUDNOEaW9G2GFUO 3c1Y1IYUYa3qvm3ifLXa4gJPFTijq289ftu02TMhEJYW2I9mcl1Bqg47/6CL/nreNc9def0rPzn/ gR+dcsVx/Q/Yx/bvf7b99b7c3Ox1u9956Ztf7qnZ6QlSaZ9Na5NtsP0A54bsut191q5s/maVY9vm uvodtRVbKrZt2VlVvefb0rW1GbUZo1L7fb9430tHnXbnqSdddnx2cUY0cEhbQUAQEAQEAUHABwE/ y4qHQ+g8ohUTS4oWUNnaZnPq/zjitKe429pa4Bw5ufnpjhS7nRSgtryCfumpKanwHLgI8bTuFndb iz0lJT09Cwnpjky3p7uKz0wUWemwWCjfSoltmb5CoV6ixGcLrMffmddFMpCqy/5y8RFnH15yxIiM nEw1d8cxp7hOm9T2+MKT7Z/bi5qPHn71oPz9/WBJy0wbedzgAfkD126u+m57fX29o8WdsqXcWdWY mpme5WptKy8r37ShdMt328q3l1dWV489ar+DJx142jXa2ZBNEBAEBAFBQBCIDQE/sqKRCVWOVBen Zf9oiR/6Qe0lW08O4SP5zGmp9owMR1paajqpy5np6ZkZGGb42rCYtJE11ObSZbk1CTZojqe7LtJ6 y4ofBPhMbLppZflMf3DC1+OPDUrzvZKQVLWR1WXbXnDH1Y5ZF48+tnnyIYfsk1rc1lzmN6nW1qaV X73+7sdv2Rsrx/XPGpyTnu5uLUy3Nza4nS5XU4Nz+LBhffv2GTBwAFtRYVFDXQMSsvrmmgdHWgoC goAgIAgIAn4IhCm3rwWX+FEKncJoriI99ERPYyZ9uQ3Pjs4+UrRwXS2nVfMltW/6bkobPEbrq//T WnbCmaDA/dX64tCLbTNXbSz1HtG/Hj/rA3VizEoSkipMZS3f3GNfc1be0Utyvv9B5qZLWt490Vn1 eQCnTN1VVTco1X16/4xTM5sOTW/Oc7cVOtxZDVW1ZfW1tfXl5XsaGhrKy8uxutTU1BQUFGjXiVNj q7IJAoKAICAICAKxIeBDVjCLaFK057vGKlJsqSnaX30/RdvR/sJIaEJ0SwrpIWl2LVRWHW//Z8PC ojXThOgRu7oE7amnH9RtN7r5xjF8UP2bH8amd5BegYvatC9aM3fuXK04vWrgadaxoI12UF9Guqti VjqZVDV+sgrkvQHUTF/2TMcBs9vyT3RXlblrthF5a9vnxtR+pym7mrFRLCfb3nLUsMHpzc6K8pq2 vXsOy2w4uCj98H0GDS7MTse+lpnBiU5zpLW6WrOys6giqPeVJR0su8xFkCAgCAgCvRABn9TlVndT KmXgPBYVlRnk8Qfp0KTqHz1hJ+12F3XE7IKIrjYnwS119bXXTb/+iYfm77zyzsyjDux727ReCH2X TLlq4Yt1LywbMP92b77i1kq32etqKUzclFr/tcNeZ8sdm5IxPFDDir07nvzN9ccMHDAkw1FRvsvZ luJqS91VWZGRkVrQpyDv+B+OPH5Kxa7dbW5bTlZOvyH9nC4nBY5fffXVH/zgB5K63CVnXAYVBAQB QaAHIOBDVpytjZSgxRCybdvaP/7xgaZWlz3Fnmu312EGsadiLVFVTrG/ZGNXsaXUtjrTU9OyszLr Gxtys7Pq6psoapuWlqa9qeuUB0MKUS3p6WmNjc18GDa4+CfX3pyRnl1bX3f99Ouf/PtTILj33oWN 734x+F+/txdIZEMCryh3dV3ZzX9ILcjrd++NKZnp3iPhyVv5+sr6yrohB4zILCrM75eDGa2lrq6q vCqvD5npuam2NBVitGf3rsd+f9O4opxiu73AkUGcS1ll/WebdhBr2y8r6+wrr00ZOGpsyZiG+sZV X6+qaa45/YzTMbcIWUngeRXRgoAgIAj0AgR8yEqzsyHdkY4Vf/Xq90849fz6xpaRGenFqfYK0nvs KXXulMJUfDxa0o+9TVtAqCm3YEj/fjvLduflZWdnZdXX1Tc0NlVUVRUV5Dtb3WksHWQnCDfDkWrb sbNsT9nOQw/b99333snJyquprbl+xo1/18kKW8M7n+++eS6v+1lHjesFmHfBFFvWbS6b/tvC6T/O m3pa4PC1e2vfWfD+IeeM2/LRVneqq2JHlb3N7nI56/c2HX7eYaOP3IczqXrV1Va+/OQDfesrXQ11 dpLDHBnfbd7uaKmzZaXvbHEeOemi7XXuYcMHb926bWD/AW8uW3bnnXcVFxcLWemCUy5DCgKCgCDQ gxDwcd80NTc3tFAFv4W1DDMcNofdlmYnHbW1ILWtv92dYWsptLsK09zFjtYCvkq1DSwucKS1NdRX p9kxprgGDew7cECf3OyMPkX5/fv1GTZ0IOvGpKURuZIycGA/4ix5O29lRURWD2ppJjjXgDH7pMOH Lv7znlsfrHzkWW9sDzvsMFWRJWk3NEz+i6F20Zs7r7oTLhiUqaA/2cs5/bPX/m/99jU7d64ua9jZ QNj0wFEDcoqy+w8rNpgKLdMzskYOHZreuDuvsTbD5mzYscVeV/ndnqr1u+sqqhrgpieceOLA/oO2 bN567HEn3HzTz4qKipIfH9FQEBAEBAFBIMkR8LGsbNu1LdWRlp6aWrFn53/fWNzS4uKFmjhZN1XZ 2+AjbQ7esFPsrjY3/yEPyKWtaehqTwLSAmdxFpEhpLmLXKxx6Ib0YInRcpjdmGJayWb94XmXpjsy qmuqZ8689d/PPueHTvmtf3Ju2Tlw/h3KJaQkJzOCYTSEx3z55ZedrPyhhx76xRdfGIO2NbWU3/Yn V3XtwEfu8HP9+CnW3NhUua0yt1+BjQUIm1rTsx2p6faWhtbMHEd6tqcQC11cDZXVf59Ws2l1oyvL ndHPWdPc1tKYX5Tx9w2uhoamy396w7jJV27duvXRRx/92c9+RiqQIjpiWenky0CGEwQEAUGghyHg Q1a+2/wdKT5wjjSHoyA/P0UrCqcFnmAD0Swwbe76+ib4A2VVNDpCpEObGy7idumMBMuJy4Xjh29I BsrKSNfSmt0QG7iLFqtrT7VRiqO2tiE1xV5RXXX7bb98/vkXAtGse+3tyvv/0f/+WRkH79utyUqX KO89KLQP10/uuRMLp/0wEOcmVysBSXoFHBtnSMv70q1serC0JzVd9VJ8UWWbN1SVbf7Xr4rTHbZ+ /VwO0sOctS2pbS2uNEebu7mx6ODJRfufWlFR8fnnn48YMSKPpYEGDKCjkJUedteQ6QgCgoAg0MkI +JCVNRvW2FMdim2kpmm1arUHFJYV/SEG9SjfXe5yuTKzcmAhzpZml8udzdJ3LU0wEocjHfuJw5HW UF9PjG2/4r5ZWdkujC9ut16YBfuKlrGMJHKd91ZV3nXn3S++8FLQ2aoAi/yrzi685Mzua1npWrJC Tvieu+fD+UKFAW3ZsiWdUGhYBnFIlO7DY+eGm/LP1traunXLVk4csbF47fiLlUjZSLZu3v7MH244 85DiDf0Gf1VRVvnFh4cOLMxIzeDs2xoqssee23f0BJezGe9PWVkZl8qQIUP23XffF194+Yfnni3Z QJ3825bhBAFBQBDoMQj41lkhjkSvVAv50Hw8mvUEcwnenFbKZsAb+vUvHjh4YEFBXt8+RQP69x8y dHBhIeEp/YYMHkzR0sGDBxFNOXzE8CFDBuPraW11u1pdLrpqlEUrZau9pLu0yJW2sO6d9DEjCGFp evfLPw08HEdGj8G60yYCTal65N/DFj8cLmD58xWVf1tQu+hfmWs+bHr2mYw9u3LyWW8yJ7ug0JGe XldbR4kUrZy+wzF69GgjbCUrJyv30h89NeaAd4rz64YMTcsbsGXDltrqLSR45RcXZ+dltzS3cNIx rgwbNuyYY44ZPlzPf9Zsc7IJAoKAICAICAIxIuBDVngVVmVUNMJCxAlxJ9AKSnCwaYVpKaWveXZ2 7ti+Y8e28m3b6rbtaNq7t37bttrtO2q2bqvdtr1+R1lLbb0WykJvtxMpyo2gJTPzxMJWw1s8qzpr rqVwDzACLAgIPe/2W7ZNuR5Di/fkzETdqrjXiC29w2Ojahwb2BGHwBhjUnOjpZ8mxSkOELPn5w55 4b7wqeCbP/rvhq8/3vXlB98999+N776/+/U3bC5XmrM+pWxlmrOZE67ld2VlNTURdd1MLVqIC1dF blrGyMGnpReck5N2ev+sSUMmzth24o3rxl5Qsd9l5Qdf2NxnGKlinO2Wlpba2lr6kgum2dKc1bEh Jr0EAUFAEBAEBAHtnVf5WZSJ/qvVq1jnRwtYIKpWC2GAXpDOk6ZXttU9OPriP7qlhPrs0BgbEbVY XkhmbnM609IztPJimg9JcxlAbqi6gi0FzoMojfBowS9tvKbX1dX/4rY7XnnplYjnoPnr9WQ1F918 Se6ZJ6vGZtwrqk3Elt4NompsqB2mV+BXEYcwZme+pTeAlKb9+sJbDnr8XrKrIgL778nn9B+cXzyo /7elu5x1e23ZucdefUn97q3fffzRmImTd2T0zy8sICaas6Y8Qfn5+WPHjq2rqFmzaX1tWmZbaro7 JT01N3vg4H6Nrbbd1a1ry3ftn9k6wlnPZcDVw1nOzc3t27cvmvzrz3+66PobxQ0U8aRIA0FAEBAE BIGgCPiQlZXffMNDRtXU1x+c2r/KykpcQXAOncBodVM0gtLqptYtTh5YQYuzGZtLViYsJ4VKKxms beemyAqSqNXBLjEumbyja3G2BMPYMdW0NTc13/rzO158IUiAbaCWVDPbNf0eKq72+/2NxuP8gQce WLduXWDjMWPG3HTTTQZZuf/++yM2MziQ+cbxkBWTmkO2zLQ0NCHru+75ZYP+dW/aAI0fRNxevGVm cWt9s921u76paMhg54BhBznc5Y1VzSmprryiuj7D8/JzyQw7+uijdLaqMT9k1pTt3rJlg6P/yPSW XRmplWVffTho5P79Ru23a8e2VStXjhhz3D5jj3KkadHWOtn1bP+c/+jF068RshLxpEgDQUAQEAQE gchkZdWab+2pado6hNqm2zDcbdXV1XqELEQjFXuKI9WhLROkuYTgIgSyaKnLvIJnZmXo6UF6YZU2 6qm0apyHaE2XK4OatblZWqSuzn+I4WxqbISsvGCOrCi9jYexY2BxxKjbXmVZ8SNzJi/0995+q5UK OZVVFd9tLt9bM6hkaN6OVXsra9Mcuf2POOqIH5zrSHPALDGoeAus2rVzx+olBcV93JWr9tRs2bax rG6ve99Rjsyi7OyRx+YPPqu4X5Ai/f/4+98vuewyISsmT400EwQEge6KQOnCySOvXmqbtGCTtuic bBYi4GNZWb1mTWoa2UAaTcGMr3ECzUYCR2mFrBC/Uldbn+HIdLpbHamYTuxE3aalpcI8cPlQUx+j S3NzC084Kr9lZpAMlN5GhRZMK2xtrbT3LHBotzfUN9x66+3z+x9BXEXa8IEZB+2befCYiOX2cXPg Evrxhy991FptxurALMwbS6JqbJyAGNxAJjU3aVkJdJOFuTgAkAAg5/otrrKKVwpL3fkj6xpacjds qKyrs43er3jkoL3rNm7ftumMKaedduUMgpS0ACWXC6uYIbNy+66KTZ+ku9aWl767eWdTelZGW4Or qLgFJ5/bMfjwM3+R12d0oAIP/fEPP73l50JWLPzdiihBQBDQEYAdTLfN15mBThTGLdeXpPU6HBmo FbMmb5xhCbkoXThr2cS502zqP8JWImNvvoVv6vK6DZAQLRyWgvqa1d/eBk2xwVogK1q1t4rKKoiI W4tBSclwsL5uisNhb24hBLOFpGX+pdrTampqsZ+QK5Sbm433ALJCsK3Ge7RABs3YgluBaM3bbr39 9//diWendcuu5pXrm75eZ8/MIA+IdQ3Thg1kJ2jgRWvZ3jXn3jj4R6f3+dmlYSbZSywrlQteqP3X EiKRgSsQDSwuoNr0yWrYSdPKddAakoNo6dh3OOlC750/cO3O1PL6+qLUlEZ32zEjB7kO2Ddz0NC8 vFxyu/YZNaalxYn/jiBZMryUD4itpqxy+84t2799qfyjl4uKXTvKimwDMvs4avJzcu3uxgMn31Y4 8vvkfzU0NOzdu5ekZdXrsfv/8JObhayY/1VKS0FAEDCLQActWTFr1mLbPNsUja2smJWyWN8xs1lH VtRopStW2MaPF6piBnvzbXzIyroN3+kuHvw5OlnRImg1kkH+sh6EoFlGoB64c4hI4bjm1CEOhdQe PSsZKqLta+U62HQ5ei4QziFVXE4PZdBsNpCVn//8F/e+vp2oXkNXiAjv/TxcW7fuYocFg5r75BUd uG/WiYel7zsibcRAIxGXZ23p62//88D85nTPmjWGkF4Ss3Ljtdc9WnL8uRee770qIXQE3FrWb2Zh SHbcTc3YqyB/sJPMg8ZQZM9ACSPHqvuO/nZHxtbSPUXVLVtqdh82bkx6cZ8du8qHjR1z9LEn7j9m zF8fW1AycuT2bdtGjR49afIk3HkscdlQVf3xB//76KP3jj+4eGBRS/XmZldObkprdVp6dW2LreTE a194acWZZ0zBxvb+B+831NUddfTRX3z5lb2h6pJrfyqWFfM/S2kpCAgCZhFo5yVwFdvcKYt1I4lt 4eSHRy2ZW6KcMmy6X6Z01uQ5q5YuxfgyZXHKhHnth0sgK4tttnnzltpm6nYZjy/H48xppzIc9ZGp 2mq0yEuU/nnOWHECmT175tv5kJXSLaU4a3TDiVZvRTEM3UqhkRV9XUK4COk9WvE3StJCUbSoFrgI 1cSoC5fuoDqtZ2VmbCoeewwdNDJDDpEq4oKFBbLyi1vv/N0bO7zJSqDSQ1Mz1614X3v0rt/cunkX XgwMA40DCvufeKRrd0X9fz8c+PivgpYS6dmWFVU0L/O803IOHI1RCsOJ4nbQkTc3rD7nxukavRsz IkykLbzhg/fO7Js6Mqeub+Nud1Nba/mmDbs++GBvs+ugKaefOuHU1JTUOb//PWaVzPSMIcOG3nHn 7a98sHLZ6s03jB/XJ81tb65Oz8p0NTXbWmqwprVUllGIx5aWXdGWfd8/X5521RUYY+rq65cve2vy lMnP/fv5fn0yr/rJDUJWzP8spaUgIAiYRkCxiVEPa38nLtO8QvNt03XfkK20tKQEC4eHZ9g85hbP R51pqK6KXyiKM2Ojx7HksdkowdNKVC++1SgLffXNT9TsNSN15iJBK6bPnumGPmRl587ten6xHsCq 8RPNFqI/+LXEYxV3q4JbceVQQE4zw7RpVd84yN/aqsqK3TtSUzMdWVkO1pZJz0jHS2RPSUtN0zKd XU7aw3fS0lJqauvuuuue376xKzxZCYwI4ak85cDDn/n1Hzb85y3HnuqssqqFZ+/fkKXF2aitx1tW Ru1umPLeFnt9kyvNnnfcoYT7YDgxvGZmcp5BCd7wwMNX9c/MOjAvb1hBcVnZzo1LX04r2zHwzPNP Pvvc9LTU8l273njjDchKQWHBnvLy2bfd+uhr7y58d+XCC79fu+kj2GlaXkFRn6JMdz2Uxd1cm2Vr y8SL12/Eq59t+PiDD3980YUnnnD8k088MXnKGU89+dSA4sJLrpgmZMX0r1IaCgKCgHkEdMIwf+xi PUxE/zDbNkcFobTbSGy6zcTmiU1pN4boQ/CFxxzjoSPtn9qZiO+3c8f7mFJKPXaVdlEmHU/mJyct 2xHwISu7y3fqthPIB5YSVbRNc+3omataPrMeutCWztpA5B+nEH3ZqtVdsaVoOzZbc0Nd+a4yzQuk uYfwFKWk2tqamxtbnC6KohIAobhLVm4BcS133f3r3y6NYFkJ+ug18zzuwZYVfD2tu/fCTtJHDA7M ijIDjiIrHzy9oH+/nEG5NgcRJlSrLdtud7Ucdvo5mY7UPZVVH3/0UX1dw6GHHLBq5eo+fQopSby3 oW31um3fGzei/PWHKxprv9mdm5ph65OV25jdpzkvNzsrOzMn58D99/v6668POOCAww4/lBWGPv7w 03H7H/DlV1+Vbt58yaUXCFmR244gIAgkAgHCWh9eYxs7Qw9p1dwwqyaNm431w4hF8XiK2slKhzlE aePr6AllWfELg1GdsOR4G1oSMTuRqRDwISsVlWV6frKWg0zRdMwm2EH0GrQaR4GB8F/M/lu2bnvg /kdysrN+dssN+QX5WrFbtxsnEYwkLS2dcFutyL4TzwAVbFmUt2HnttKKPbv0zBL33soKd2rOfmP3 e+ih+ff+N2rLiqaxbt0xmVPT47OB/K5j82SlZOu/a+t3N9U3OtIL8goHf7zsBVdLfV1Ndd90e0Y6 Jz3FlpE/8uizq+sb+hT1yczUTGUktVdWVdlamlhMyO1iOSCHFoHb4Fz9xQdV5VuxxJEclN/UVulu zrLb81Iz3LnZDc2Ndkf6lxu23DD7NiErctMRBASBhCCgGTtsKoREBZGsUqnD7UaQmTNnrhrb7iny MBrlrtH9Nb6OnrnjfWNW2qVMmqRRIKiMJwzGL2ZFXD8JObUdQn3ISvmeXdriydhU8N1QCoxHEiXY 0xza8kB6rKyKUfn5rb9Y9OzzRx91+GN/fVQjKKQApaVR53TD+o0tTU1j9h+TlZ2N66eiorKwqBBb DH4fbSXnlBRna0tLa0uzLuuO227/9X+2RusGMshKGFh27do1aNCgXljB1gw4Cjd4Q/4X9zS76zGY pdkztn+zavXSp9ramtvczr6ZGTUNTQQW5RUPPuSC+7dV1RTk5w0dOiwnOycjM+OFR+/ftf7bxmaX s9U9+ohjzrt6ekN94ztPz9+8+qN8l91W35K3vfxTV83QkhEHuPIJY1o9ON1V11iemnXn/X8TspLg 37KIFwQEAUGgxyLgszbQ5198/e036z/69KuPPvn66694gq3dtWvv9p3l1bU4CprJAWKlu1defW35 W28XFRVOOfOMgsLC7du3vrd8xXNP/ePl514o27G9pqqqtrp225atn33y6YZ167dv3/bN6pVrv/22 fO/ejZs3NztbPQYYR2aL05kIUEv1gKrLL788vHCiMUaMGHH66aeb0SGqxmYEhmoDzTLZ3XzLUALr K3c2lu2q3rJ57fLXdny4JDM9vSi/aGhxcWpGutOW4XK29RvQJw3bSUZmq9P19Vdfb9y40dncUr9l 06avVleX71nz5Rc1u3YgnOBp17rN+W+vatyxM6+0rK26Yb/9D8gqLnI563cXte27qeKQ1TtzW7SQ bdkEAUFAEBAEBIHYEPAhK+W79+zZvTM9ze52ttTXVjc01MM2CF/48L0PiF1Yt3b9Z5999pe//JVM ZkJnqa6/af361uaWouI+g4cNy83NKMzLys/J3LDmm61btpTt2lVVuefT9975bt26PeW713zzzber V3319dffrFm/eeO67zasaaivj03jwsJCPXom+DZy5Mjzzz//iSeeQHiYlhQ623///V966SVDh6ga x6Z5+F4maRZCzLcMM+IXi59Z/9ZLa5e/umPlquqGpj11jc3uFFeKo7Gl1dHmLCjISR8+LjUjb/jw YSwJpK2fPGIEnr38DEeBvbHIUVmQ3pqXnY58UsJaC/Ob7Lbsssr1g1LeHpm5Lb0m87NvGzeU7umT k7mtrDwjpYFEd9kEAUFAEBAEBIFYEfAhK0cccVC/Af1bW5pSU9x4cLKzMqsq9h5y8DgCUxprq1LT bI/Mf5TCG7iCBg/oN2JkCYW/cABRnraxqXno0OGDho1IT0+t3FNuY+mgpoYMhyMrN6e4/4AcFvDN zGhpbmyoqa7aW75lc2ll+a6srMzYdGatIr12S8jtySefVJLDt1y6dGlmZocOUTWOTXNLaBYczZuQ xaYJvTbtqdi4afPe3WX1TVUNNeU57pYMd11FdUV1s23AQUcN+d6FaSUTs3PyBg4YmJPDCczOy81l J8NVP3jwoNR9xw8Y0DfDrkVVo497cJ+qcUOaahr2Vjd8RyXbdzY4Kuv+NyCrfE91rdO5dfTQjQ11 MespHQUBQUAQ6HwE2lNMOn9kGTE4Aj5kZdvWna0u96aNpXt27x02ZBgP1x3bd5SWbsvKyd137AGL l7zx9op3NL9Aq3Ps2APq6ynCkZafn1NRtqNfcVHffgPXrd24e0913/4DKCY2et+RGWmpI0fuU9Sn D1nNJAIdeOCBzY11WRn2wsL8LVt27Nhp1uXRY06dVTQLmmYQspjBGXDYWcUnnZ1/2Mn5I/bJLeyX kt9nt2NQ5v7jT7jqzuGTZmSUTCzuMyItwCKyp6G51m13N7Vk9Bm8e3fF8088tmbV6oa66sa+WV+O HVCandlYXvtVS+uKAkfZ4OJyV92aQpsjvy0j1VMAN2ZtpaMgIAgIAtYiUKdvNTU13FH5S91tbtGE aVKRnRXx2Gpra3khX7t2LTsEA1g7ukiLFgGfANuf/+ymY44/vqhvn6rqmtXfrN+8edvqlV8fOGbU 5FOPS2mpu33Oo7uralnIkLUIH/7zgyeNP7mluYlk5nWrv4HcOFudvHvnFxTs3bPHkZ6RneLO6du3 oaW1qqKysqqS3NeRI4d98cVnvIkPHDRs+87djzz44BOrbTEE2B522GFffvll+HkeeuihX3zxRcSW qpkSFVVjY/So1gaKOARiTWputPTDwXw2UN6b83KHDkvLzyVbh5xzMtApjJObW0AtwMbmZgcVdNw2 ss5zC3ONIfhJf/b68427N7796utpOUW1Tc07d23/0bWzi/vn1dbs3NvgTEnPaGps3ltdl5Xt2H/U 6LY0W6YDl2LTJx+svmHmLyXANtofp7QXBASBxCGwYsUKKAjUZODAgVCTsrIy1m0944wzvvnmGwIe WC2ErBFlwIfE5OXlnXrqqenpmu9bti5BwIesPP30PxoaW0YMH1K9d+fHy15LczfnZGUPHdh3dMmg 5sbGl99e8+o7q6kOl5WRftnFFwwYNuyUU0+heD7pQ3XVdenpjnwsMdu2bli3ZtCQESQ2v7Psv3kF +X37D8zJzty5Y+u2bTtYJXH0fmOPOOowZ2vb1VddPf+TxhjIipnncQ+usxKeJ5kBBwnwhqqXfmVP z3VkZ6ZlZObk5ZYMHxZ4/dVV1mXmYUjTwqL5VtUw3r7qkwdvuDa7X3FTq2vb9t3X3nZP3/32r2+o z83JJv2n1dnicrsHDBgwaNBAQ+CSJUsmT54sZKVLfuEyqCAgCARFgDdV6MiWLVv69euHBQXjCjWi jj766LfeemvTpk3Dhg2jEjc70JTdu3f3799/woQJDkdHAVJBtZMR8CErazeuL/9utaN+W5ajtTjP wXOHmJXWNnuT05mW6ti6s/LRl76oqG7sk+348RnHZQ454KBDDyO6Njsro7xsNzmw2XkFZCnbU9Oo stJQX7d106aRo0YS6/DpJ59u37FzyNDhB447EE5Ddbm6+ppZs26TOiveJ9uv9q7JWjIxW1Y2PndX n6IizoXT2ULsjtPZyq9x6NChe/bsGThgwJatWxvrG/r3HVDfUk+oCr9kQpIzKFHbr//2lZ8svefG hkxHiz2jbveeydN/lTV8ZHV1JcWJqc2Do3fUqFGwGn7n/LBrqqrLyss///yzCy6QonCd/NOW4QQB QSAyAhhUCgoKsJ1wG4SXEIWJPwji0rdvX1xCFRUVhEPAafhKL+MuW5ch4ENW3n75iRz37ow0bVkg 6qm4WpqJpaWYvtvltKU6mlvcKz7b+tr/1hw4vO9N085o7XewOyO3T3FfFgTilZpnW3Zuvvaw0hY1 dONGcKSltbpYtJmSLax6yDu3E6ZC9ALrHTY01N18y62/+2+ZWFaCsg0zBpKYy/sqy0ral/8oKuqj lyROgVCydHZqWlpzU9M777xbWVkBz2hubJr0vclL/vufE048kd/qRx9+RGjt908/fdtXH7586+Up x13gzizYuviRH97yxxFHH1dZuTc9k/Tn9NzsbH7bK5Yvh98UFhSs/HollXjcNveNN9wolpUu+5XL wIJAD0IAWy9xJMpBw52K1zxoBAfXrVunomK1xXT1jR0MJEVFRbSk+AIshPco7mbaWmb6NmTIEIpy sUNUyvbt22kPQaEjtzLaIJZcSL6l+yuvvAJx4QgNDNaCrYUSGDTAAIMEjiut2EEUdAf5fIuPafPm zaovygwfPrwHnY3Om4oPWfnk37+lPCyAp1JTJdXm4qy2uCiIQolae6qDc7B1R81Ti78+dNTgH//g qNTiMYWjj3JDZlopHKfVfUtNTWmsr8/MzdEegC5bq8tJ5X2NvGg1+7VVmykfl5Vmp5h/XV39TT+9 Yc47tTGTFalgGydZCYo8p4rf8+N/W3j66d8vLh6Qnpr+/IvPYSM9/PAjyAbnApg1e+a2lZ++eO/P z7xmWnpe36cefOjkc6898sgjXXXVOInsrLKQm5fWp0+LLWXJf5aU7dxF9vuMG366dMnSs88+W8hK 5/2sZSRBoOciUFVV9eGHHx500EEQgmeeeebqq6+GAUAIODhu3DgOsujHfvvtRw4joSfA8IMf/IBA 2scee4y7EHbif/7zn7fffruiIHfccccvf/lL9t9///0+ffoQs0IJK5ZFu+yyyzj41FNPXXnllZic EY4vG5ncdT///PPRo0fTEsJEyAs1vXhOPvzww9S3xfqCZbq8vBx3EkuYbNu27aSTTuLB99///hed TzzxRL763//+d/3116uVa2SLCgEfsvLe32/F9qFHocAoUqCWrU5tkWWiYlNT05ytru82V77/5daD Rg/83oRDWEkv78DTHZk5+oLKGpPlrGirHbL8D94FCvZrSxzCZVzaMkEam9HWF3Kk2R321Jr6uluu v37Ou/Uxk5Xwk5SYlfD4wBvyvn2RBRQ0ItmWkpmRkZOTa6NMsUZSHZ99+nnfvoOyc7Iy0mxvrVh2 6CGHDh8xnN9/+e7yUyaMr6uqqNu9MT/D3paeXlnjTM8Z0C/Nga3MbdcuBGJs7ekZOcOHrv9u0+4y LeHr2OOOfeW1V3947jlCVqL6ZUpjQUAQCIoAD34sGeQr8O2f//xnXMw4neET3333nTpIMMohhxzC 4wiysnXr1nPOOQe3DsxjxowZfLtw4UL4jZIMWfnNb37DzkcffURyQ0ZGBibh55577tprr1XCf/Sj H2F6UaaRgw8+mINfffUVlIjHJMO99957l156aUtLC0zo//7v/+iObjt37qSIFy0/+eSTww8/nJav v/46bvSTTz4Z68vf//736667TshKDNe2D1l594mbIHyaiUxbIk9fUxl/TgpGEfbt9Y3OneWNq9bv 2WdQwfFHjW5scWUOPzq1334uZ7O2eJC2QLOtydmG0ycr3ZaWkgqFgeRgY8EvhCiNyrC0IcEvdntN XcOtN90w74OmmMmKWFbitKw0zZmS1VBVW9+4Iy27YfCQA/cpdNja+g8c1lDVkl5QYEvf7nRmOnKP dGRmEy3Pewm3g/zCfLezDTNZShrEhpwwmz3d1lTbVFm9t9npJM3ZyUoKdq4fV2N9Xb+BxUMGDE9J TXE1u57+22OXTJ8uZCWG36d0EQQEAT8EIAQkhMJLeEzBPDB+YBSBT6gMUA6uWrUKcy8xKFAQLCUY VGAJtLzooovgExiJb7rpJiXzF7/4xW9/+1t2PvjgAxxGmEZo/9prryk289e//hUuQroAwhkRNsPB Tz/9FBMLLb/99tv169crsvLII49MnTqVg9hOCIKBzWB04dvjjjsOsoKphrsoZAUmtHjxYrGsxHZJ +5CV/y24Hn+fvoQy4ZJpuHV4MqVjVEmzO1vtVTXNW3fWDB9QBOvoW5Tb0koh2z5Z+3+/DWdPmw1K wvmgS5NTW3E5w+bOSE/FpeTW1hNqS7PbqLAC5WmhhJwtFbJyz20zH/i4JWayEn62YlkJjw+8wfHT vrWNabtbUz52pxUOyd8ntXVQTsHAgsLqPRVZ/bKzBn/ZnD8qvf/PN2xnGe1WVnrq07cP6xhg48zM JLsrC+eexmudzQ6b41+LHltS/tl5B448YPgJ1e7v2lxD//326/uuSz33tptLt2w7/oTjFz364NRr bhCyEttPVHoJAoKANwIEw7766qswBngJz/6f/exnuIFgMJAM8o1p+fTTT69cuZLgBFr+/Oc/h0bU 19ffd9990AV8Q3SBUpCuzEGMJTAYuuCdIaAEJxF+HPYxqHCQEuc333yz4QbiBsiTZfXq1ezQkiAY XuxxGOEGmjt3LmlEkBXiWmAkqAFZQT6+IR6XWFY2bNhw1FFH8S2KsaqiWFZiuKR9ycpfr4FQ2NPg HkTW8jTCY5NCoAkhsUTKVtY0bdxcefyhxBOlYifBxdPc3FSw33hH332JZEojzLbV1cJazVS6ZXFm +miRRjY4TVMrTiK91CmuIXcbgbd7Kivn3X3rg5+6YyYrYlmJ07JSdue+tfmj8geOaMrpW1Szw7Zt c/POzQMaqguIjh7Qz3bKaU39hmfl7l/dUIVNhZhZzumuXTtee/V5jGR984saG+uy8/NKRu0/Zp8D nnvzuZfefmTCiMEFI/Zf9cF7KfZRlLIdvLbpx9ecnb7oX0NmzFy6Zeu5l14lZCWG36d0EQR6OQI8 9WEDRKJgosjN1co+wUtYckTZOQgWwQ1EKKu3G+inP/0pHhyF27///e/zzjsPy8o//vEPnDvcNhct WgSxUFGu+HRwGLHz8ccfs6PcQM8///xPfvITDmIvOffcc1UVFsPHRMwKBU5pSTwvvS655BIsKwsW LJg2bZq3G0irSvXZZ1h6lGUFNxDxK95q9PLTGsP0fcjKisf+j7gSyAZ/oCg4cdr4P1zFxoLJ7sqa xlXf7p580n7NLdprNc4dyArxuP2PPNedWZiKywgmgknGkeaysZYv/AZHggNHQKvb1tTUwiqG2k6L k+V7G+ub7rn9pgc+jp2siGUlTrLy+WsT3el9HBWpebVZeU21bVXljordWVtLK8qqmoYMqaNi8SHH DD/qe9mFuVp0tL6t2fDd5rWrm6sqG+tqMwrysgv79i85YJ+BAz9audxevbNPUeU/X/rE2VKWkj6w aXf9jqH239Xl9mna0Fp0+Ns5/S/55a+ErMTw+5QugkAvR4BHPtQEXwzmEwJXue9BVoggUSQDPsFi cIqsEMiiGMwNN9zw0EMPBZKV6dOncxCygpuGNB84EKF4RxxxhCIr9CUJCOMHDQLJiiEcZxNkhZZ4 eXAzGWQFzxEHUYOYFZVDhObIVGQFiw4BtvjTCZ1RnKmXn9YYpu9DVv770BWEyGpeHwBOS9eKpmis BaNJamubq76ueceOSrctLT8nY0BxbmsrSUIud0tj8Zhj7IOPaSFxqJVQ3FTYZQvZQyk2p5MoXZsj HRtLSnNTc0a6HbscgmEtNbW199wx84FPYqlgqx7SYlmJk6z8beFVWc31JdvX2hvsWQWFQ6u2p6Rl f7d3d9Oe8qba+kFDhux39oVDjp6U36dAXVVOZ/PLf5rz3cbSlopdtSkZwwcUa6tw52Wdcv60XWl1 rkZX36zM6rqqj/b+b8XmdVUpjUWpeb/5tPHzQ8cdl5W7/NP1Vz34mJCVGH6f0kUQ6OUI4J1h8y4d i3PnP//5D4SABwH+IAJQVDYQB0ljBi7sHO+++y7+F3Ib7733XuUGuv/++8k0xsJBXg+1oGhJFySf dtppdHn77bf5i3MHDxGVbTHG8PGFF1648cYblRsIwsGibNx1CYhhh5Z4doh8UG6gefPmEZ7i5wYi qvf0009XbiC4DuYcmBCxL7fccouQlRiuah+ysvT+SzCFwDcgKQ68NVpVFNxApPA44B0urF3NraVb agcPLMjNIRwzpbGRxQkbcvqNGnTM1PqGBo3ZwEyItCS6Fr+RlrdM2VM7GUYQl7r6hjwW6nWkUdKj tq7p7ttuvP/j2MlKmKkSIUX8NtdxxGol3g2iamyMHqZX4FcRh0CsyWgbo6UfDmaGoAu84d2Fd7I4 Zdqub23uplEFOazutG3r9ozC/Kya3bb6lpHHHNe07zEFef369i/ykJWmuo9m/Xj3tu2ZBYUtA0bn 1u1JaarOaNhVMv3P+Ycd2+JsIM3L4bDvbiivdzan2lKz3ClZj/75lRMvP+KDxetqdl1w33whKzH8 PqWLINDLEcABBOcgOoTnDxYUnjFYVki0UZYVomUxaagAW+XKUQG2+F/gJRyhMQG2KnUZCkLULRSE eBeMNDhleFiccMIJyCF1GXcPbibCYyE9BO1y8PHHH7/iiitUgC1xtSRLcxA3EESHlmvWrMHAowJs 8ToxikpdptwtwuExEBSEK8sKCmBZgaywj5dKyEoMV7VPST6Kt/G8cdhd6SmutBSnPcWZkoKVpKml td7tbHG38q+xb36qva2lsqqupbW1T356/+IcV0NZY2U53iKtGLudhOdmd6szI9WWlenIz3bkZaUR aJvpsGVmOOpZnbm5BYuNtrKdXr3H8g2DIdFP5L6Hl4wflGI+0F4zCkTV2IzAUG345Zjsbr5lKIFH jcz//rH7n3XZtJ/cfu9pP/3lQRdcf9rNv5pw1S1HXX/PMbc/MHDSxUOGjYRZ0l0PPbKRmDw0z73/ oNx9hxTtn+sc1Cdz8LD+w/umOtJb161dv+7b9atWr/7gg093ri1v3dlaU1rd1pLh6NP/0P/9O79s Y/rI0SbnJc0EAUFAEPBGQEWqvvnmm5QtMaqxQSAweLDBD/T7k7apg/ylNBT0BbKigu3UTQxHDPk+ fEUQ3uDBg2nJI4A2qgFWf6JYOEgb+irhFFMxhNNLHYTTqJYUfDM4BxwIaaq7Es4O1hdjIko3Osrq QjFf3j5kZdDg9MKCtNw8G0U3MjJT0tPIQCaUNtXtwgugCqY48vLSCUd57Z31v3z0nV8+/vGTb6z7 dFXp7k2rOFvpadSTc/MvHdsK4bROp40gFlebq1WLrM3JSM1OS6mqqWlqbHK7WrHcxKY01x/CQ20Y 6HBhQrcRHqYl1yip8AR7GzpE1Tg2zcP3MkmzEGK+ZZgRKxpcZRV1O3bvWblq9cpV3+zZuzfDkZGa lpGemdvaktJGJrKbiGl36aZNFCrAv5qamZ11xKnZhx2XNWpc1rBR+fuOyxyxr/2osxx5fXHuYUjD JUgNQcq2lFfszs7LHjJsSO75Vx2w34jhl12ROnRUIhATmYKAINDjEcDUgQ0DAoGtQhWoZcOkwUbQ Ky4hg8Gwyg/3Rg6ykVdMAxJ2FNvgeYFTBjm0gfRwnG9ZsBDjihKIdYQudKfUG0En7NCAYFuDjlDf loNs2EuUcNxAhnBsP0hT3ZVw2mCPMc4ORxiagxSL6/GnLEET9HEDbV18DbnHLS47JVJaqKjRRrUU ygbbmhrcLU1tuHJIayZ+hVShir211TVNb329a3tt2uQpp48ff1Jhn/4kCUFKcL84Um2NTU7KrmRk ptOd/BI3Bcj0Cv6tTldmlqOquuE3d9/yh3fbYsgGShAQsYmNyg2k8uvCDIT788knn6SB+ZZ+0sy7 gSreuA9LJuFEkFDeABiR2wHOVyq/2VvsO7bv5IVmvwPHVNVX5ublNdTXNzQ37bvfmHR7ql5Px/Mq g8Glvqph1+4dza3NhYVFOIDRhxtKbU3NkKFD8/OLNm/evXvHzjVrv7xi2uXiBortGpNegkBvRoAb Ee9LvEzCIfC/8FYMQSFGBAsHtxoe/3/4wx8wgXBrJSoF+4eiMpg3sGEoRwwZPVAEqtZykNdUXEjE CWAXgalQi1aFpxBWQrAtwmEY7JDkDBHB48NA+J4gOjh6tBum3Q4jYWhaQnpwJ+Ew4r5HcTkWGKIB MmmMcEak2cUXX6xiVsiXRnmUpPGdd94pbqAYLmkfsrLjtek5uQ4CJ0lLrm9xNVO+Vsv7SW112mrq XU11ehAtJAaDicuZnZFKPf41G6rz+g8cftAxRfsc5sjI0IkJVVC1IFhsMs0E09psWelpxK64qcdi t5FVVI1lxem+73ez5n1g71VkJYbTE20X82Sl8Z1HMjOztEx1FiB0aQsmYJTih/b+e++nZ6RXVVfX 19adevKpX3/zNetlszxCYVHeBRddlJ2RbfAU3Tdka6hp3Pjd+srqykMPI5Y+A1HUw/3q66937NhR U129t2IP0qpran8+8+dCVqI9m9JeEBAEIA0EfFBDhYhakmtUNhB2DlJyAOcvf/kL8bOqgi3sQR2k FIraoagJVhaiSSAQFF+55pprOIhBnegTYmzZV3Xk2IHBUAeFHdgGFWyvuuoq9h999NGzzjpLVbCF MJFDxEECYtQOlIgwXugIVhkqzqlUI0wskCpDOHVcCI+g3D7uqmOPPZaWxMGQaiRkJYYL28cN1Njc bLe5cjJT8rLshTmpuVkpWRkpafa2zPSUotzUPgX2gty03Ey7I6XNwTqFKdR5Sx01LD2vbVfZJ8+X fbo4JyMtOysrNyc7G5dQRiYPQuqr9CnIy82hjFhGFkErDkdts7u4ML8gL5P4lhjUlS5WITDukHGj 9hu1z5iRw0cOGz1m1D777MPvJz+/oGTUiH/866mR+4w4+9yz+/Tve9qk0wYM7HfhJT++6sppMBVG x0Sm/qnzZ09P4dRm5mSt/mb1hx9/iE+pwdl4yOGHjN5v5OdffEbp4qumXXXg/tqNQzZBQBAQBKJF AMsKbz4kGPOkZz9Md/xE6ltVmkXtqCMwBpVVxD7N1HGMH+oIG0Ya5aDhLwOpg0T1GqzCWziMRLU0 Ilq8uyvhSFbLJfIVm+pOR8OTFS0O0t7HsjLr4sNzctIJjNUgJtYEY0hbir40EPGwoM9SQSxPaHO2 tDY3t0JriDshe7kFG0yT02VLHXH499J486ZTW5tTuwjaKFmLgYWIWvZZkJnXeBxCUBbSit5c8uJT m/qLZcXaS9C8ZaVt+eOcKOXPwTSam5PLGae2Tqu79aNPPu5X3L9f3775efmc9rXfruVFhAI8xg+b LvpqzdrW0tRa11hLEBIdEYeogsJCbKTUuv322zWc9ZNOPHnx8v+cf/GPxbJi7bkWaYJAb0CA5BrC VjCQ4FXB+IFXhZwasn9xDPHgJ9OYNB+4CDyAAnEEsfL0wVMzdOhQimgQVkJGMWsDEX3y+9//noO4 vKmsjz+IWFfsJZiTMcwAI0k62GMIwlW1/MePH89ByqhQmla5gebPn4+zidHRBFsLLbGs4EXCDQQp wbODGx3zCS0Ri3CCggnmpQoLPiD0V2srogZMiMa94cRZPkcfsmK59PACIz69TD56O1lt7+Giilnp BD1NIhYR+USo2iWDJmIiIlMQEAQ6DQG1jjEV2GAY2Fe4xXGE5BploiAQBL8M7IR4EUgG3h/ICnVp 8RnxyoSXh/ZnnHEGRyiiz/LLMJiXX36ZXiRY4PGBPRCeghyWJCRuDyFwI+gFCckcpIAbedEqdRmK Q0IyBxmFavqQHmJpycqkJVacBx98cPLkyXRHIBIQjikFt9Qpp5wCUyGDGnMLtAmaRREXareIGyiG 68dDVjYM+767ui6G/vF0was0eusbYSREjDONZ3RL+sLuifYKKsokb7BEDUOIyUGT83RbC4VIEwQE gR6AAA94mAokgx2e99huISu4rVUiD+EjhJtw38OyQrqNKr4CNSFYhIOQFQ4Sd0IXiuirtZQpUIup GDuHXwVbglew+nM/pwFLKNMSa8oPf/hDLDrYS7CjqDorCIe10BKzzYcffmhUsKULB2lJzArCaUkF W1ZdXrp0KaNjg0F5zC0QIEJnhKzEcGV6yEoMPaVLeAS6hGmFIU9yvgQBQUAQ6HYIkNFDzApqQ1MI sMURw7OfYlo8+DmIKYWcHawX5P6QmUwDlcUDn1CJP5CYM888k7zit956i1WEID1wEewusBzoDlTm +OOPR074VZfxDWG/gXnQEjMJzMPkqssIJzr4nXfewQ4kqy7Hee1JlGucAIbsDkPXyql17hbKzJOo SYpcQUAQEAQSiQBhJaQfs8EVVEkVIlR+/etfwzlI2/n+97+Ph4VQFawaJOzAV9jwCrHKoKowC0vA jKGsI+Q547uB+kCA2OGjYjxshJKoLthLVGUXNqPOCg4dFmfGecRG7AsUh28JXvGus6K6YARCuNpn RIZWO0hgRIYIHyOcSCC7vezUu+++u9tPQiYgCAgCgoAg0DsQIByEOBLCQTCfwEv4y0satIA4EoJa iKJln1BcasiCB+YTQkxgLXiRfvSjH1EBhQgYDDA4jGjAt4TC0IxAE9buoQotWdBEopDtjBy4BTYY IlHgNxRcgXOo0iy33XYb3xITw3H8TWQeYLnBYaSK2BKHi3xsP7RHTxUHQ8gtbiA4E9zlyCOPFDdQ DJdq97OsUJ2W9ajYYLhmJrz33oWEdvr946CZvtJGEBAEBAFBINkQMLKLIQqYN6AIuIGMRGJlzkZn tYMlhsQcI2fY+NZ7UrAHVQgfP1Egk6CL0V0V0uWIkcys5HiPaBzBFATjgeIomYaGyYZnt9Cnm8Ws 3Hzzzay3bCDLeptULQwPNLwkkJr0vW0a/7rFGRIlBQFBQBAQBAwEsKNw28dGAlMhlITsYiwcGDBY bBlSAiHAjsK3KnWZqBGSgMgcfuaZZzCiqNRlco+xghCJQqAu5hYkU2GW5YfoArHAEIJ3CUZCeArl cUldxrlzww03EKeiF6PKJ12ZgBjcOnAmyo6r1GWa0YC8Ib/UZewrLOlMxAwpQlh3oDiSuhzbxdzN 3ECY2lSlHbWRRXb99ddjjgsz+cZ3v+CfX4Pskw7nX2yQSS9BQBAQBASBrkKAsA+eAnhV8MgQPoLr BwKBMhykSiwHiV/Bp0N8CYW5sW3gyoF5sGQP2cW4hCATJP7gJ8I1w7f4a+gLg7n11ltJYCYmF8ks tgyPIQiGltAgBDIKqyVTmx/PDl3gPVAlyIdyA0GG4DfQEWw8OJIYnVAbmA1/SZkmIpixTjrpJHrh Y0JJcQPFcPF0MzeQX3QSH8OvthMDItJFEBAEBAFBIGkR4EkPI4FMsFGKDT0JicXIoY6wqTIqaodv IROkE5OeqQ5iUyG1mB14hmI5bMZqiOwTWQLJoAHMQ3l/sNYYDbCpwFew0GBKMVw/1IhDGr1UUjR9 GYsR6c4RRlGKEShjrPOctPAmrWLdzLICSyWiykCTOCbocHhwxbKStBefKCYICAKCQLQIUOcelw05 ybh7qKSCowfeAC3A1UJ2D7aWTz/9FPMJBIUdXDPwBqzvLNBDGwwbOGXw9dCSkrX8hfegAEZ61gxS mkBW8CIhB0ML5hB4Cb4b0o9x6GBroXoKfijk0IXG8A/ELlu2DH3gTLilCKFFLKEq2F0gLgjBskLq kDLAIPbEE08Uy0q0J5323SxmBaaCJ0hdJVxSL774Iia18NOWmJUYLgvpIggIAoJAciIAdViyZAlO H1gCrAKyAmuBr1AcVgW9YvBQbABzCIXzYRsEtcASIBNqkSC+5SuKwmH2UEsekm/8/vvva09EVu5N TcV/hBBCWPAcsc9Y0BHICr1ULrRqSUI1jyH2qRQHR1G9lBmGLth18CIRAQN3geLQnoNYX3ASJSew Sa5VNyMrCk3qBuIAUss3RNyErESESBoIAoKAINB9EcDWgu0klLkCdgIvCR/amLi5UzmXLGsCWRI3 RC+R3M1iVrDUEf3ERuoyxj3SmHvJeZJpCgKCgCAgCARFAJOJXyKxdzPCWWjg1xEDCblCuHXYMKvg ssFsTzPicNnHOsJxGmCS4TgHMedT2E2ttxzVdtpppxUUFETVRRoHRaA7kRVK8eADwqyiZoLZjRSy Cy+8UE6tICAICAKCQK9FgIhXVSUl6IY3h1BZv68gIhwhsoQacUS3UNUNVw4EhUAWjkB92CFEkr8w FRWMwkdyhaIFeeXKlWoZI9niRKDbkBU4StAqcGTPi30lzotAugsCgoAg0H0RwPgRpow9TiKCWvxm R0IyMSj4hkgOwk9EXAueGlqyvBrBuXiUsMfwLYk8JPXANqA77MCKokWJmBiCfKPtJe0DEeg2MSvY VIxobb9pEGNLlnyosysxK3LdCwKCgCDQgxGggAWswjv92HuyhLUS4hroi4HfEBLLxrfkHtOM0Fqi W7DE0B3jCh+RieuHOiscpAGNo4WRNCLCeOFA0XaU9n4IdBuyQhC1d9Ky9zTgvzBiIStycQsCgoAg 0OMRwBCCE0eVwIcEUGmNYrL8xZtjlEUxEnYo4EZ7/DsUOIGUQD6MbCCAwppCGhE7WFnIRlbQKSGK qVAcRVlZiGtRqUBGNhA7pAKpQi/Ix/qiUo34qFrylVpAkYMqaIaD2GbUiLJFi0C3cQOFYipMWNax jPasS3tBQBAQBLopAtACyArkA4MHNAUqQPl8SphQVYWDBK9QHZ+psU9ICoEm7LBxUO0sX76cNGZ2 oCCEqigQCEnhI30xqKiW7LOqM/ErfEvkCnnRGFd4MeZbRqTB9u3bly5dyrcYY0j5gYhwkAr9BFNi p4GvsMO3WFao7AK14luE01JWCIrtwus2ZCW26UkvQUAQEAQEgR6GAHYUYk2ona/CQcjfIdCEsBIO YgthOWVqnLAPicGJA43AsIERBSsLZe9ZP4i/bFRsM+rJYhShCw2oEUd3+rIR16IsJfxVvTjIesuM yw6N1dovkA/cQ3SnAYYTSr+wQ41/4mAYmrr+BLuMGjWKLgg3lmDsYWekE6YjZKUTQJYhBAFBQBAQ BCxDQK0XSGVYzOoQArgLtEMdJH6FCrOUYlP7qsgbRAF3DB85iGFGkRvaGNQBIaoLphS6K1EIV3Ew CKGXOlhXV6c4CgzGiJLhiOpujKKyqWlAPOVefVPCxawS80UgZCVm6KSjICAICAKCQGcjwPMePw5O lm+//RYGgDMI8oEvhjooeIU4TswADiC1A8+gPdQBusARGlA6hQAUduhOFIvSHtaiuuDloRc7bHQx Yl8IiKEL2+uvv85CzWedddY111yjirvQBvuNEo4OZB4p4Spj+YwzzsAqowQyhAQtxHy5CFmJGTrp KAgIAoKAINDZCEAOWDXwmGOOYeEeglipg4IGmEbGjRtHJXuSQylyTyF8GrDDcfiBcuvwFQ1w0LCE Mjsnn3yyCo9lwzCjulC8n17ssOEPMiJ28RnRhQ1uRKjKq6++SiSK4QYizhdl+JZEZVasY4cC6yQo EbQLd4FIKeH8xbXU2Xj1lPG6DVkJUyy5q+oo95RrQOYhCAgCgkDyIgAnwLCB8YOnvtISxqB2OEKw Lc4Xdoy6cHyL60dREGqyEUX7xhtvGB4f4ktUSwJmDT8OBEgd5K8hHCGG18ZIWvauPmcU+GdHJTzT 1+iOY4jwXgwqGHJULxV4m7xAJ7dm3SZ1GdMc2/Tp01WItdrgyPPnzyc9jJ1QOEudleS+AkU7QUAQ EARCIkCMCC4VuIhKIWadHUJP/vrXvxLEqvJ3zj//fCjFu+++i8GDNZDZh77wLeyBHRrMnDmTJZcR QgQu/IP0HBVCi5uGqNuzzz6bsVnGGQ8O5AaHDsSISFjkQDV+/etf0+bzzz//17/+hb0EHRYtWvTC Cy/QhQaPPPLIT37yE7gUzRCIKYWMJOJqeR5hzmEs1UVV9EcfnFYExPziF7+QVZdjuOK7Dcsj9BrD mlri0tj4yMEwTCUGRKSLICAICAKCQJIgoIrMUleNxeBgA4olqNwcOAel3li1hxxjYk3IxOEg2TqQ Ev7yLTyDJB3sLvwlM0h1wfVDS3ZoY2QDYQ4hi4eD/KU7OzRmOOUGYiOjRwk//fTTSWOGlFA8XRn1 0YcdY0S601LJ4VuYCp6gadOmXXLJJTNmzPjuu++EqcR2aXUbshLt9Mqm/3Zd/vH8w7IS2JeD6lua RStZ2gsCgoAgIAh0DgL4X5RFhLgQ+AqD8rAn6ITlbHlThUPgbVFFZok1ISqFuBD+stItDQgToT2L /vz973/ndZd4FL4i3kWthvu9733PWAwZB83EiRNpcOqpp9JGNVAF5RQdwWajjmMsYQg+sqOcTXyL GkhT4TIqbgZpCCf9h7AVVQxXwSVuoJgvmx5LVgbMv73g4inhcaEBzWLGTjoKAoKAICAIJBoB3D0Y JP7973/joFFjqUxglbpMQVgMJzAYdRAnkaII7HOEPCC+ogHuJJV7bCQ2kxZk5ObgyuEj3/KX7kaW ssEtlIGEjXhegmBUyo9hd4GOqO6MqLrzlw3NUQ9XlCI9bComV7YYEEi9++67Y+jWVV2efPJJ71K2 MN8rrrgilDK5Z57cumVX80qtmmHgJkylq06ijCsICAKCgEkEsFtALxSxwH2DDYMdCsXSHSJC7jEx IlATfD00Iy4EioBLiF6wFnY4gt+HkBFYAoyBJGfcRmpoglQIhaFcG/uQIaiPKroPEYHf8PGLL77A lIJ8BnrllVfoS8QkdhqUIRoGORARlXBEdAsDwYeohMvaL5h5aINYGvARsQTBnHjiiYcffjjmFow3 4gkyefa9m/VkssI8Q/EVYSoxXCvSRRAQBASBTkaA5zp0gWBYwlawW8BXMGPADI488khiWThCMjCW DFX5DUJA7AgaUjdWJR5DINiBpvAX7wyhJLiNyC5moyViKShHew7SRRW9hWognG8pi0JAJPJhHjS7 7LLLkMAOq+qyQx1bKu7jVEI4tpZzzjlHlZ5jGTs8QbxIQ3EosItu6H/dddfhJ6LmCoYZfFhCVmK4 inqsG8jAItAfJEwlhgtFuggCgoAg0FUIEFPCmj6qLApWE5UnzEYMLMYVSAPMwIgLgQooNoATB9sJ BMI7h9T7W+/pGPXfgjIJbz+OyqA28qiVEGNEQzc4CmYeQnHRTTmMYE6GnK5CsvuO28MtK+rEeNtX hKl034tVNBcEBIFeiADshJAReAn8gGBbHCvEr0BQ8MjABgh3PeCAA6AsBJEQjctxaqvQBmMJqxvC ckgdwrMDS8B/hJfnww8/JJSEDScRkbmsJQSkGDzwCuH6wR+EQUVVdsHZpNxA9Prkk09IHcLEgnDG oiWrG9KMSFuICEQKIfiJaIYniKHJdsZsg90FNxOmII6gKs4jJkIYr1hWYriMe75lRYGi7CvCVGK4 RKSLICAICAJdiADEAqqBgUSVVmP/zjvv/PGPf3zxxRdTQIWP8AAYA6RE5Q9T0kLt4DPCp4PH56KL LlKJzSqfWTXgo5G6DMtRXfjLt+ywQY9UUTj+En1iyFQ7OIwU51CWHtWFVOpf/vKX55133qWXXvrW W28xKNYgHEkqL5qO3jXluhDS7jh0NyMrOAK9UY6qdi18RXJ/uuM1KjoLAoJAr0UAKoCJgg2zhHdt WQUIJAMXD5GwNIAZYHfhr2Ib7LMDicHUgYkF5sERCA1kBZ6hlmj2Jiuqi+I67LAZZIWB4BzqIJuq oWKQFb5VxWAQzmacKRgM4+IGwn4DMVLChazEfCV3M7Ly+OOPc+0a25IlS2KeuXQUBAQBQUAQSHIE sF5QYYV4FPJucL4obdUagWyEr0I++BZSgo8GJ4valFtH5fuQJYQjhr4c5wg+HdZBZAc3kLGQIc1U F9xAqhl/6Wj4a7DrGJnPKjmZ2F5ld6ENXiek0ctIruY43XEhwVRwPyGWbxlCrX0oWwwIdJty+zHM TboIAoKAICAIdHcEIAeEj2Dn4NlP9i+hIbfccgsMhpARsnXIAMJ6ASEg1oQMHbxCcBFydjBskLEM xSFq5P333+c49g9oDaIwiuCRgXBQfwWXDfiwljIRMBhaVLALJnyIyGefffbHP/4RBxABKKwuRMox BwlV4Vs0gXmgAClCSL799ttxAKEhQkiZRjiuKxY1pCodWc24gQitxQjEV5AqfFgSsxLDNSlkJQbQ pIsgIAgIAoJA1yDAI5/8GpZBZnhcPFAQeAOEBlcLCylzEMZA3Cs7lELBpAFZgYuwqJAqaPv888+T BQ1NgWTQEvbDQRgMO/AJSAaBujTgIAvPkaWMVwjWAgFin4MQF5Z3xpuDKeWjjz6iiD5dFixYwCJB dCeGhvaqO7QGvxUBv1hlzjrrLAgKVhZq6V5zzTVCVmK4dLqZGyiGGUoXQUAQEAQEgZ6EgOFMwVCB oQXqgKnDWCHZmwpgSsFeAnswlk2OiIMRGUMXQ6ZRypZvVdwJJMl7IGPfWN4ZMoSbSa0cZMiJOLo0 CIWAWFbk2hAEBAFBQBDoNggQC/LAAw8Q5Yrrh4AVnDKEj7CPr4ccZigCVc75FopATC6l3qjMxrcc xDEE//jggw9wEmGGwQ1E36lTpzJzvDwUdlNuINZnxjoC+WCHdZVxAxHA+9hjjykbDPYS5NCS2BdM NVdeeSUjkgGEJriBaEmyNGYbaAqj4G8ivAZqRcAKFAebEPs0FstKDFebkJUYQJMugoAgIAgIAl2D AGSFQBAiQoiKJaiWYrLsELYCISB+BRsGpVCgC5AV3ECwE5gH1heOqyWEXnvtNUJb8CJBI3DQsBoi 03j33XeJLME8Q3AJRzjOQWgKFIfjuIGgOKy3zMH77rvvL3/5i5o5O/h0GB3ydNppp9GSgisEr0Br ICV4pgipgUihJDlKkBjUgFHddNNNQlZiuHTEDRQDaNJFEOieCKyYpcp3atvkhaXBJ6E3CvltRx/v Vbq6JxyidbdEgMuTkvYQDuwoLBWEkQPbCaSE5GEOwkhYiAe/D/vs4L6h+D3LNRP0ylccJH+Yv2xU bKOjggCzB5EoHOQvMtVBLChGnRW6q16qnL/3RhuIiBoRUw0hMkoOhhYsK8Sp0EAJp01U5Ta65elJ mNJCVhIGrQgWBJIPgUkLNqnM/01TF42ctSKIgqUbV81c3rZkWkl45VfMmr4s+aYnGvUCBLh6MVfg UiEbGW8L3IUCtewErrqsisXRgK9oQBdcP5g3VO4xacY4fRRgWEfUQs38NQ4SfWIkJ9NXdceuE4gx XZCmFoJWwtUii3iI4ElQGSWcNsY6z73gRFk8xW7vBiIzzVhF01psqHuIAzJ+mUmuYZKrFz/+IqED gRWzJm+c0c5DShdOfnjUkrnjbZhSJsyjEURmycRlk0devVTfnVbqdRzuQnvtK+9msJq5JZ7DNu3D eEFbEEg0AnhYSN4h0BUigs8FfxBHoCNUh4MWMDqEAKOIKn9CNX3CUzBvkONDMwwtkAa+xcIBQcFD hPWFLqo6C98ikwQivoKm4NYhA4iBoB0kAbGDTLgIuc3KOonrB0MLLdGHNGasO7iWCOlFDT5SRw6m wnLNKKA4CmL5ioUMEw1Rz5TvXWOtO+5zVhKktlWSQ8mBDEW8pMjIDzM7SzQMIyRODS1RL0Ent5eK XT7TsKxgW1kwCX7R/h8QUd96Dvsf57Dq6/m+XVR7816KqEw7GRAw6EVQZWAqmF66Sk8CVhT1kS1O BMQNFJEwJKoBBqHwJw9745w5cwgdT5QGkeQmv4aRZiDfh0GgdI1tbInNVrpm6bwJ+nvihHlL15Qa HfyO03rqRJrbSqZp1hhjK5k226a6mwhzkfMhCCQCAUJYwmQmE9SCbSMR45qRSdiK+axpMwJ7bZue TFZwcHSEE4bYo03XnvunnnoKBYL+pUQjdOGee+4RDbsWgZ45+orF88aNgn2UjNUNLGrzoiF+x0vG 2hYt06gM3iDfWJfxc/Wuy8ep72UTBAQBQcB6BHoyWYloGOAOm6B4F/Mniow4Gof6S+abeVEJapn8 GiZo4j1S7NKrR3p4+5yxm3Rq0mEa8TWO+B0vmTafkFwtEWPR1Bn0g7xcrUXowl2UwAm22ZGCcnsk ojIpQUAQ6AwEun2ALXdJOEdQqNRXZMCzZkRgA9LYVL57+O7xn4RQQ5A1R05/ePnErPz6179OqIZh EIhTwzCS40dVJAgCgoAgYCEClY8869pdUXz3dAtliigLEej5ZCU8WF1IViI+y8mUI1idwogUOwrD xuK8GuJBILyGEScYp+bSXRAQBASB+BFoa2rZdd09tYveRFT2SYcP/tfv7QW58YsVCdYi0JPdQAop LCvXBds4bi2UsUmLGLPCOlixSbaqV/JraNVMRY4gIAj0NgRa1m3efOLliqmwNbzzeemRFzZ+sqq3 4ZD88xXLSlK7gf72t79dddVVyewGCqOhWFaS//dvUkOWTWFTq6jIJgj0GATqXnt71/R73NX+7viU zPR+995YOE1bZlm2JEGg55OVpI1ZifgsX7FixYQJE6j9TMnFoJdLRAlmLrJ43EDhNbREPTNTkDYJ ReCrr7768MMPqRVx0kknHXLIIQkdS4QLAp2GQP2bH1be/w9lTfEbFGcQR/Iumlxw8ZRO00cGioBA nHVaurw70wulQ5ivjC5xdjcz/VBDqON///vfw/x9/PHHE61hRPkxa2gGfzMASpuuQoBanyzV9oc/ /GHBggXU92RHFSCXTRDoSQiszTvO719Pml2PmYvErHQxnQ2fGHzFFVd0sX6h06qV5smgYZdD1PMU YCkTYpXuvfdezCqPPPIIZIVVbSm99c477/S8ycqMBAFBIPkR6PluoPDnIB4niMmzG0/qssSsmARZ mlmFAKTk3XfframpYf0UzCqXXnrptddey8cLL7zwrbfeeuihh0j4ZzlZq4YTOYJAlyOwLv94Px3G 1Lzf5VqJAn4IRE9WfNZC8/lgIbjGGmsRZUZkGxKzkji6ZkHMSqdcThGvImkAHcFUBk3BfNK3b19+ NTASVntgvXv2YSp//OMfR4wYcfDBBxNBJZG2csH0JASErHSLs5mMbiAWgdVXd7Vm40UQO3bgxnFs 3daMEYeUMInB48ePJ2Yl6IrkcQwYddfk1zDqKUkHLwQ2bdrEKZ47dy4Lw1555ZWffPLJXXfdNXPm zGeeeYZVxwmtnTJlyksvvcTBRx999IADDqDEYqjCP4KrICAICAIJQsAqstJedFstZsbr8qxZehVu fRURv4/akvT6ZjSeTFtjvZEVtinayq4JmrAhtrS0tKSk5PLLL0/0QOHlS8xKMHysu5zarzWvC6xr T3gSjQ5Z//Of/7xo0aLRo0dTKxnWMnHixAMPPHDp0qUsfE9gNTRl+fLl2FocDgfr1t54442/+tWv TjnlFMVfZRMEBAFBoNMQiIWsdCwvwjqtuqalC6cvmrqJqONNUxdN1+nK0lVj52uLm9nmBHwsXTjH tlyPUJ69xtN43GyvJdQwKFgzfRYC9CyDEuw/3ILPP//8J554wprBopeSm5uLXvPnzw/zl9fZ6AVb 1qNzNEzw5WSz6cv0bVqwarFGnGXzIABLJiLq1FNPnTVrVnNzM/4dyMp//vOfyy67jOP8OshVxoJy 4oknLl68+Oyzz+bvRRddBL9hh2VsCbwVKAUBQUAQ6DQEYiErkxZovETfls9UZMVYPn7iVJu+yvwk z3LyYwM/Bi5JP0lbqd76DQdK+KytJ5980vpRTUtkYaDw6qE/QQNDhgwxLdLihp2jYaIvpwRdXRZj 3bniIBwbNmyApuDWwd1DCO2LL774pz/96fXXX4eOHHvssW+88cY555yD94faKpCY1tZWCMpvfvOb yZMnw3KIun355ZdhM52rtYwmCAgCvReBWMhKIFody8cvW2TTmcdSnbJoLCbgY6gl6XvnSUj+Yvad r6FcTgn9LVCLFv4B4cC5w0Dr16+nngqrULFP5CxeIcwtau1PFvskV/mggw4iunbQoEFY2i644AIM lhyE5Ugac0JPkwgXBAQBbwQsIitey8fP19eJn7Rqjvey8d4fQy1J3ztPTPiYlbFjx3Y5LJ2vYYlc Tok86/vuu29ZWRnpPwxCPhc1VFQwCh8JVYGgkKKszCqHHnroa6+9tmbNGqK79t9/f6wsWFxwXN58 882nn3464SxKiGyCgCAgCCQcgYSUt1s+08u2rzmLfD5aOiTVIBKEEZIt0TSUhrynRtScsMcwbSzR MAyAcWpoiXrK2dhpl5MlZzz5hXz55ZfUUGlsbMQBhMkE7w+85Ntvv0VzrCwQF/KT77777jvuuONf //rXc889t2fPHg5Svnbt2rW0mTRpEpTl7bff5tvkn6xoKAiER0Aq2HaLK8Qay0rEh27iGhAbmCCg kWyJ2qE0NB8REmqCgRrquS8qyUpLwgr1wXteYQCMU0OrALTkLIgQbwRY4gejyNd