MODEL FOR THROUGHPUT PERFORMANCE MEASURES AND
MONITORING SYSTEM IN SEMICONDUCTOR INDUSTRY
NG POH LAY
A Project Report Submitted in partial fulfillment of the requirement for the award of the degree of Master of Engineering (Industrial Engineering)
Faculty of Mechanical Engineering
Universiti Teknologi Malaysia
MAY 2009
iii
ACKNOWLEDGEMENT
Thanks for the academic guidance and technical support especially from Dr.
Muhamad Zameri Mat Saman and all others faculty members who involves in off campus Master of Engineering in Industrial Engineering, Universiti Teknologi
Malaysia, Skudai, Johor, Malaysia.
Appreciate on the opportunities and supports given from the management of
STATS ChipPAC Malaysia Sdn. Bhd. semiconductor plant for this throughput academic study and project of piloting the propose methodology and application of model for throughput performance measures and monitoring system in assembly plant cross departments.
I’m thankful to have my lovely family members and friends who always be there with me throughout this project study.
iv
ABSTRACT
Manufacturing Science has been a proven discipline to further enhance traditional manufacturing practices where manufacturing performance can be quantify, measure and improve. Factory Physics is one of the approaches in manufacturing sciences. Through the application on Factory Physics principles, manufacturing performance gap can be systematically identify for throughput maximization. Managing manufacturing throughput is always a challenge as most of the manufacturing shop floor information available are either qualitative or quantitative but in piece mail which practically make us almost impossible to directly use the information for systematic scientific analysis. This is mainly driven by the dynamic of operation behaviors primary influence by manufacturing variability.
Variability in manufacturing is unavoidable and sometime just simply unpredictable.
It consists of the varies in people skills, process requirement, equipment performance and products. Buffering the shop floor with excessive work in progress (WIP) is the most common undocumented tribal knowledge to overcome throughput uncertainty.
As a result some people even describe that running manufacturing is an art. Four fundamental models for throughput performance measures and monitoring system for a semiconductor assembly plant being developed and introduced. Overall manufacturing system performance expected to be further enhanced and a common ground for continual improvement to be established through this systematic modeling approach with the application of the basic concept of Factory Physics
Principles. Based on initial mid range financial analysis over seven months, there is a potential opportunity for USD48000/week incremental Gross Margin through additional 20% of quality throughput couple with 30% Cycle time improvement.
v
ABSTRAK
Sains pembuatan terbukti berkesan untuk menilai dan meningkatkan proses pembuatan traditional. ‘Factory Physics’ merupakan salah sebuah unsur sains pembuatan. Dengan mengamalkan prinsip ‘Factory Physics’, jurang perlaksanaan pembuatan dapat dikesan dengan sistematik untuk meningkatkan pengeluaran.
Pengurasan pengeluaran adalah cabaran yang amat besar bagi sesebuah kilang kerana informasi yang biasanya diperolehi adalah terhad dan tidak mencukupi untuk analisis yang sistematik. Ini disebabkan oleh perwatakan operator dan pembuatan yang berubah-ubah. Variasi pada proses pembuatan ini susah dielakan mahupun diramalkan. Variasi ini merangkumi kemahiran pekerja, keperluan proses, prestasi mesin dan produk. Penimbal kepada kerja dalam proses (WIP) yang tinggi adalah cara yang umum digunakan untuk mengurangkan varaisi dalam pengeluaran.
Disebabkan inilah, pengurusan pembuatan adalah suatu seni. Intregrasi untuk menilai dan memantau prestasi pengeluaran diperkenalkan dan dibina. Penilaian dan penganalisaan keadaan sekeliling, penimbangan kapasiti sediada dan pengendalian kerja dalam proses (WIP) dalam lingkunan masa akan dibincangkan. Keseluruhuan prestasi sistem pembuatan dijankakan akan meningkat melalui kaedah sismatik yang dicadangkan dengan menamalkam prinsip ‘Factory Physics’. Berdasarkan penilaian kewangan dalam tujuh bulan terkini, terdapat peluang peningkatan sebanyak
USD48000/minggu pendapatan kasar daripada peningkatan kualiti pengeluaran 20% dan pembetulan masa pemprosesan sebanyak 30%.
TABLE OF CONTENTS
CHAPTER TITLE
DECLARATION
ACKNOWLEDGEMENT
ABSTRACT
ABSTRAK
TABLE OF CONTENTS
LIST OF TABLES
LIST OF FIGURES
1 INTRODUCTION
1.1
Background of the Project
1.2
Objective and Scope of Project
1.2.1
Capacity Planning
1.2.2
Work In Progress Management
1.2.3
Cycle Time Management
1.3
Problem Identification
1.4
Significance
1.5
Thesis Structure
1.6
Resume
PAGE vi
4
5
2
4
6
1
9
9
10 ii iii iv v vi ix x
2 LITERATURE REVIEW
2.1
Overview
2.2
Factory Physics Management Principles
2.3
Lot Sizing in Sequential Operation with Min
Cycle Time
2.4
JIT-Kanban System
2.5
Performance Measures for Production System
2.6
Resume
3 MODEL DEVELOPMENT
3.1
Overview
3.2
Throughput Financial Modeling
3.3
Throughput Concepts & Approach
3.4
Throughput Execution Strategy
3.4.1
Factory capacity balancing
3.4.2
Capacity Interdependencies
3.4.3
Work in Progress (WIP) and
Cycle Time Management
3.4.4
Develop Sustainable Models
3.5
Resume
4 PROJECT IMPLEMENTATION AND RESULTS
4.1
Overview
4.2
Project Implementation Phases 1
4.2.1
Machine Throughput Model
4.2.2
Factory Capacity Model
4.2.3
Input – Output - WIP Model
4.2.4
Cycle Time – Lot Size Model
4.3
Project Implementation Phase 2
4.3.1
Throughput Monitoring
4.3.2
Cycle time Monitoring
4.3.2.1
WIP and Cycle Time Report
43
47
47
30
38
40
42
43
28
30
21
21
22
23
18
19
20
26
27
14
15
17 vii
11
11
13
viii
4.3.2.2
Daily Cycle Time Report 48
4.3.2.3
Queue and Process Time by Lot 48
4.4 Resume
Report
4.3.2.4
Conversion Model for Execution 51
Analysis
51
5 DISCUSSION
5.1
Overview
5.2
Review of Achievement
5.3
Critical Appraisal
5.3.1
Project Strength
5.3.2
Project Limitation
5.3.3
Comments
5.4
Future Research
5.5
Resume
52
53
56
56
57
58
59
60
6 CONCLUSION
7 REFERENCES
61
62
8 BIBIOGRAPHY 63
LIST OF TABLES
TABLE NO.
1.1
2.1
3.1
4.1
4.2
4.3
4.4
4.5
TITLE
Summary of Preliminary Findings
Summary of Literature Review
Financial Modeling
Capacity Limiter Chart
Throughput Financial Analysis
Throughput Model eUPH Computation
OEE Computation
4.6 Manning Ratio Computation
4.7 TPT Computation
4.8 Input of Capacity Model
4.9 Machine Inventory Capacity Model
4.10 Machine Conversion Model
4.11
4.12
4.13
4.14
4.15
4.16
4.17
4.18
4.19
4.20
5.1
Input-Output-WIP Model
Detail Input-Output-WIP Model
Cycle Time Lot Size Model
Data Monitoring
Coefficient of Variance
WIP and Cycle Time Report
Daily Cycle Time Report
Queue and Process Time by Lot Report
Lot On-Hold Analysis
Conversion Model for Execution
Summary of Throughput Achievement
PAGE
41
41
43
46
46
37
37
39
39
40
6
17
19
29
29
32
33
35
47
48
49
50
51
55 ix
x
3.4
3.5
3.6
3.7
3.8
4.1
4.2
4.3
5.1
5.2
FIGURE NO.
1.1
TITLE
LIST OF FIGURES
Ideal Production System
1.2
2.1
Manufacturing Sciences and Components
Theoretical Cycle Time Curves
3.1 Basic Queue Model
3.2 Ideal Production System in Detail
3.3 Design Capacity
PAGE
3
8
12
20
21
22
Machine Capacity Breakdown
Sea of Inventory
Pull and Push System
Die Issuance Flow
Flow Integration
23
24
25
26
27
Machine Throughput Modeling
Total Manufacturing Hour Breakdown
Queue Time and Process Time Plot
Cycle Time Performance Trend
31
33
50
55
Relationship among Key Performance Measures 69
CHAPTER 1
INTRODUCTION
1.1.
Background of the Project
In today’s ever competitive marketplace, where the level of semiconductor manufacturing capability among competitors are rapidly raising. It is expected to deliver a higher quality output, low cost product at right time and quantity to achieve competitive advantage.
In line with the above expectation, apply sciences into manufacturing and integrating right level of performance measures into execution strategy is essential.
Traditional manufacturing system approaches which solving single phase problems without integrating and clear understanding on basic manufacturing behaviors with interdependencies no longer adequate today. This is mainly due to the increased of manufacturing variability and the complexities of product technology requirements which are always in a flux. Variability namely from difference in people skills, multiple product flows in same value stream, high volatile on demand ordering rate, complexity in frequent equipment conversion to maximize install capacity, dependency on equipment processing time against product types, product quality compliances and more. Anyway, variability can never be eliminated from system but
2 it can always minimize to certain controllable level for a more predictable performance and delivery.
Factory Physics principles will be applied in this study to ensure manufacturing variability are well understood before setting up a more practical and sustainable throughput performance measures, analysis and monitoring system. Area of improvement can then be identify, measure, improve and sustain systematically at all stages of assembly process for next level of operational excellence.
1.2.
Objective and Scope of Project
The objectives of this project are to develop and define a systematic and scientific way in managing and sustaining performance measures and monitoring system which encompass Capacity Planning, Work In Progress (WIP) and Cycle
Time management in a semiconductor assembly operation which ultimately further enhance assembly line throughput performance.
Four fundamental performance measures models will be develop at different levels of aggregation planning stage to integrate information and data. These models will be use as basis to perform productivity analysis and proactive performance monitoring. They are: i.
Machine Throughput Model ii.
Factory Capacity Model iii.
Loading-Output-WIP Model iv.
Cycle Time–Lot Size Model
3
Throughout this exercise, overall throughput performance expected to increase not less than 20% couple with 30% cycle time improvement within 7 months. This will then yield total incremental Gross Margin of USD 48000/week based on current financial assumptions.
The scope of this project is to further enhance throughput improvement on assemble the unit of integrated circuit chip in a semiconductor industry from silicon wafer to a complete unit.
Plant throughput is a bottom-line delivery as results from integrated productivity performance measures for a product specific capacity and flow cross process operations. Figure 1.2 illustrated the ideal production system where the capacity measures comprehend the build-in concept of Theory of Constraint (E M.
Goldratt and J.Cox, 1984). Individual operation capacity measures are express as in
UPH
1
, UPH
2
and UPH
3
as stated in Figure 1.1 An overall system flow measure is referring to actual output or throughput. It is the minimum measures of UPH x
DA WB BE
To increase throughput need to improve UPH
2 rate based on Theory of
Constraint
Input
UPH
1
UPH
2
UPH
3
Output or Throughput Rate = UPH
2
Figure 1.1: Ideal Production System
4
There are three key fundamental parameters for throughput enhancement.
They are: i.
Capacity Planning ii.
Work In Progress (WIP) Management iii.
Cycle Time Management
1.2.1
Capacity Planning
Capacity planning scope encompasses the definition and itemization of standard measures of individual equipment performance parameters such as unit per hour and overall equipment efficiency at product and individual process level. Line capacity can then be define with the integration of both. Proper application on Line
Balancing technique (MCS Media, 2003) and the application on Theory of
Constraint (E M. Goldratt and J.Cox, 1984) are key. Regular capacity monitoring and reconciliation for the needs of additional capacity by investing in equipment capital for mid-long range planning are essential to ensure ramp capability are always in place to support business continuity.
1.2.2 Work In Progress (WIP) Management
Work In Progress (WIP) Management scope encompasses the definition of individual process work in progress triggering limits includes constant work in progress (CONWIP) and work in progress (WIP) staging arrangement which align to predefined push and pull strategies. The dependencies on die loading process will
5 also be discussed as it impact overall Assembly plant inventory level at given time.
The daily WIP level monitoring process by operation and inventory turns developed to serve as an actual real time performance feedback mechanism to always keep WIP in manageable level whenever variability exists in any point of the flow.
1.2.3
Cycle Time Management
Cycle Time Management scope encompasses the applications of Factory
Physics Principles namely Little’s Law (Factory Physics, 2006). Lot level cycle time tracking automation, aging lot movement aligning to equipment conversion strategy,
Lot on-hold disposition business process and individual lot level queue time analysis for continual improvement to be established.
The propose approaches of redefine, define, develop, monitor and sustaining the integrated performance measure are basis for reinforcing continual improvement as the process cycle continues. Shop floor raw performance data or information will be channel up systematically as it is basis for value added analysis, process improvement and streamlining. Silo or single phase performance measures without bottom-line value stream integration can be easily avoided.
Quality and Safety compliance measures assume as built in process which is given from the study. Also, this performance measurement system excluded external customer and supplier business process and delivery cycle time.
Conducting the study in Assembly operation consider good representative for the plant performance as Assembly Operation usually covers more than 80% of plant capital and resources investment for a standard semiconductor Assembly-Test plant
6 in Asia. Test operation which conducting the functional and structural test on assembled chip is out of scope from this study.
1.3.
Problem Identifications
Initial findings were done through observation and interview. Summary of qualitative challenges listed in Table 1.1 below. Improvement focus on capacity, work in progress and cycle time are essential to ramp the factory to next level of excellent with the same available resources.
Table 1.1: Summary of Preliminary Findings
Caused of Failure Effects Process Failure Mode
Capacity Process
Frequent change in incoming product mix
(package; body size; device)
Differrent cut off for different product
Unpredictable experdite lot list
Work In Progress Management Process
Late in die delivery
Maximize Loading
To meet various or non standard customer requriement
Would like to support urgent customer demand
Fire fight to meet commit output timely as capacity limitation is not being consider upfront in input process
Capacity gaps between plan vs actual due to different awailable day in specific calendar work week
Impact conversion schedule; jerk the flow in pipeline; impact throughout and cycle time
Not able to meet required FIFO target
None linear die supply
No auto FIFO lot triggering system. Lack of visibility in lot movement via automated
Waste capacity system due to multiple operation are represent in single operation code in system
High cycle time; too resouce intensive to track every lot movement manually and accurately
High work in progress in pipeline but all are not align to custmer product line item requriement Conversion based on WIP
Cycle Time Process
Not clear on material disposition plan;
Ageing lot in onhold location whether it is equipment related or incoming customer related issues
Lack of capacity, target and actual performance sharing periodically as well No clear visibility on the chain effects to line execution and performacne; Not able to drive to true capability as resource capability alignment to the right level of people.
High queue time as work in progress stage more than capacity; additional conversion caused low output; high consumptiom on techical resources
Impact cycle time
Not able to decide and plan right strategy on equipment conversion (Output or WIP) ahead of time. Reactive execution prolong cycle time.
No direct cycle time and on schedule delivery measure which align to actual mfg process flow
Lack of target and measurement sharing with right level of people
Not able to monitor and predict Cycle Time &
On schedule Delivery effectively
7
Preliminary problem statement identified based on preliminary findings stated in Table 1.3. Preliminary problem containment includes: i.
Capacity a.
Develop capacity model which optimized between input or loading and shiftly capacity to minimize potential variability downstream. b.
Update actual performance capability at all time to manage potential product schedule or delivery skew. Right level of equipment dedication strategy may apply to minimize the impact upstream. c.
Synchronized production cutoff timing cross process operations and customers. ii.
Work In Progress Management a.
Optimize equipment conversion through input scheduling and batching strategy from upstream. b.
Automated lot aging reporting. This can be done by creating unique operation code for all process operations in current production system. c.
Leverage on expeditor to expedite lot movement based on FIFO to avoid lot aging. d.
Define maximum work in progress triggering limit for each operation correlate to throughput and cycle time target. iii.
Cycle Time Management a.
Automated lot cycle time reporting. This can be done by creating unique operation code for all process operations in current production system. b.
Set business process to clear lot on hold triggering limit. Define specific timeline for disposition with clear ownership.
8 c.
Detail out lot movement cross process operations through automated production lot movement report. Valuable information such as lot processing time and queue time are good to identify problem for continual improvement and analysis.
There are many disciplines under manufacturing sciences which is applicable to further enhance throughput performance. Factory Physics (Factory Physics, 2006) is one of the approaches chosen to drive manufacturing throughput in this project.
The application of Factory Physics Principles (Factory Physics, 2006) describes the fundamental operation behavior of operations such as work in progress, capacity and cycle time relationships. This is important principle which should eventually help to better understand and quantify performance measures systematically.
Figure 1.2 summarized few key component and approaches for operation control under manufacturing sciences. Factory physics is one of it and we will discuss more on this section throughout this project.
9
Factory
Physics
Variability &
Buffer
Analysis
Throughput
WIP/Inventory
Mgt & Control
Cycle Time
Operation
Research
Manufacturing
Sciences
Probability
& Statistics
Optimization
Modeling
Scheduling &
Sequencing
Modeling
Queuing /
Processing
Modeling
Simulation
Descriptive
Statistics
Statistical
Process
Control
Six Sigma
(DMAIC)
Statistical
Distributions
Lean Mfg
(TPS)
Industrial &
Mfg
Engineering
Value Stream
Mapping
Process
Flow/Mapping
Waste
Elimination
Just-In-Time
Kanban
System
Motion &
Time Study
Man &
Machine
Capacity
Work/Process
Simplification
Pull & Push
System
Capacity Mgt
Theory of
Constraints
Linear/Nonline ar
Programming
Nonparametri c
Statistics
Kaizen
Line
Balancing
5 S
Space &
Layout
Planning
Resources &
Capital Mgt
Lot/Batch Size
Figure 1.2: Manufacturing Sciences and Components
1.4.
Significance
Throughput performance is most significant productivity measures compare to others in the entire assembly plant as it has the highest potential in financial improvement measures as it link directly to bottom line manufacturing unit cost estimation which consist in all 5 cost models in predefined cost structure as stated below i.
Depreciation ii.
Fix Overhead iii.
Variable Overhead iv.
Direct Labor (DL) / Indirect Labor (IDL) v.
Direct Materials
10
1.5.
Thesis Structure
This thesis consists of total six chapters. Chapter 1 begins with the introduction and background of the project. Objectives and scope are based on limited time and resources throughout this project study. Significances of this project are based on the current environmental scan and preliminary findings based on existing system limitations. This project was then structured to align to the predefined objectives.
Chapter 2 focuses on literature review summarized from total four related literatures and manufacturing sciences related theoretical principles. This was the basis for this project write up.
Chapter 3 elaborates on model development principles, concept and methodology. Critical approaches in applying sciences into manufacturing for a more systematic performance measures and monitoring system will be discussed in details.
These models cover critical value chain from financial modeling to shop floor execution modeling.
Chapter 4 discuss on project implementation and results. There are 2 phases of implementation. Phase 1 discusses on introducing of the models developed for performance measures tracking and Phase 2 discusses on the monitoring systems for sustaining purposes.
Chapter 5 reviews on the overall achievement based on results from pilot run.
Some critical appraisal and future researches are also suggested.
11
Chapter 6 summarized and concluded all the learning and results throughout the projects.
1.6.
Resume
In summary, the project focus on developing models for systematic performance measures and monitoring system for throughput enhancement were discusses through the application of manufacturing sciences. Approaches and methodology are aligned based on predefined objectives and scopes with relevant literature review as well as preliminary findings. Thesis structure at the end of this chapter also provided a good overview for this entire project write-up. Next chapter will be discuss on details literatures for more in dept understanding.
12
CHAPTER 2
LITERATURE REVIEW
2.1. Overview
Based on the problem identification from Chapter 1, main source for literatures are from journals published which related to manufacturing sciences, system and controls which play a roles in influencing manufacturing throughput.
This includes capacity planning, material flow, work in progress controls, performance measures and Factory Physics principles.
2.2. Factory Physics Management Principles
Factory Physics principles provide practical Scientific relationships such as the “VUT equation”. Where Cycle time in queue is equal to a variability factor of
(V) multiply by utilization factor (U) multiple by effective process time (T). It also provide a balance “portfolio” among WIP, Capacity and Cycle Time for operation management. This expects to lead to a practical buffer management strategy in
13 getting the best possible operation performance for continuous improvement.
Factory Physics framework starts with a fundamental definition of a value stream: a value stream is a structure of flows and stocks providing transformation to meet demand. The most profitable situation is when demand and transformation are completely synchronized. Which in reality it is not easy and it takes efforts to make it happen.
Factory Physics Basic Governing Principles:- i.
When operation variability increases, consistent output expected to decrease and cycle time expected to increase. ii.
Theoretical cycle time curves illustrate that average cycle time equal to infinity when machine utilization over availability equal to one. This can be illustrated in Figure 2.1
CT
U/A U=A
Figure 2.1: Theoretical Cycle Time Curves
Whereas,
CT = Cycle Time
U/A = Utilization /Availability
Availability = Capacity available at given time iii.
Theory of Constraints/Bottleneck Management a.
Indentify constant in system b.
Decide how to exploit the constraint c.
Subordinate everything else to the above decision
14 d.
Elevate the constraint e.
If the previous steps clear the system constraint, go back to 1st step in identify constraint in system but do not allow inertia to cause same system constraint iv.
Kingman’s equation a.
Cycle time in queue (CT) equals to variability (V) multiply by machine utilization (U) and multiple by processing time (T). This can be express based on formula stated as CT = V*U*T v.
Conservation of Mass a.
At a steady state system, input quantity will be equal to output quantity. vi.
Little’s Law a.
Work in progress (WIP) equals to throughput multiply by cycle time. This can be express based on formula stated as WIP =
Throughput * Cycle Time b.
Critical level of work in progress equals to process bottleneck rate multiply by pure machine processing time exclude queue time.
This can be express based on formula stated as Critical WIP =
BNR (Bottleneck rate) * RPT (Pure Processing Time exclude
Queue time)
15
2.3. Lot Sizing in Sequential Operation with Min Cycle Time
There are 2 methods of lot sizing: traditional economy of quantity (EOQ) lot sizing or a method of lot sizing that seeks to minimize cycle time. The setting is for a sequential" as opposed to a simultaneous" operation. Simultaneous operations are those that work on several parts as one time (e.g., heat treat). A sequential operation is one in which a processor that can work on parts sequentially but must spend some time in a setup or changeover before moving to a different part. The classic example is a punch press.
The lot size has a large impact on the system cycle time. The larger the batch the greater the capacity required since relatively more time is spent in production rather than in setups. As results, with more capacity, utilization will be lower. This was also brought up in one of the principles of Factory Physics where it stated when utilization approaches 100%, cycle time increases dramatically. Also, once utilization goes below 80% or so, queue time is very low so that increasing lot sizes to increase capacity will have less impact at low utilization. This is also illustrated in
Figure 2.2
Two approaches on lot sizing being discussed as listed below. This will be elaborate in subsequent chapter. i.
EOQ Lot Sizing ii.
Min Cycle Time Lot sizing
16
2.4
JIT-Kanban System
Conventional JIT and Kanban system effectiveness measures are much rely on relative comparison which are less effective. JIT-Kanban philosophy is now able to elaborate through complete ANOVA statistical experiments results.
JIT emphasizes “zero concept” which means achievement of the goals of zero defects, zero queues, zero inventories, zero breakdown and so on. It ensures the supply of right parts in right quantity in the right place and at the right time.
Push and Pull system are two types of production systems, which operate equally in opposite sense and have their own merits and demerits. i.
Single Card Kanban System ii.
Two-Card Kanban System
Categories of blocking mechanisms also being discuss which includes:- i.
Single Card - Instantaneous a.
Blocking due to part-type b.
Blocking due to queue size c.
Dual blocking mechanism ii.
Two Card - Non Instantaneous a.
Blocking due to part-type b.
Blocking due to queue size c.
Dual blocking mechanism d.
Blocking mechanism Operative on Material Handling.
17
2.5
Performance Measures for Production System
Performance of production system was summarized as a function of all 6 main manufacturing deliverables as listed. However, the weight on significant of each of the following performance measures will depends on the situational needs of a company from time to time while align to its business strategy. i.
Cost - Unit product cost, unit labor cost, unit material cost, total manufacturing overhead cost, inventory turnover (raw material, WIP, finished goods), machine capacity, direct and indirect labor productivity. ii.
Performance - No. of standard and advance features, product resale price, no. of engineering changes, mean time between failures. iii.
Delivery - Delivery time, Master Production Schedule performance, inventory accuracy, average lateness, percent of on time delivery. iv.
Flexibility - No. of products in the line, No. Of available options,
Minimum size order, Average production lot size, Average volume fluctuations that occur over a time period divided by the capacity limit,
No. Of parts produced by a machine, Possibility of producing parts on different machines. v.
Quality - Internal failure (scrap, rework) and external failure (field or systems failure) cost, warranty cost, quality of incoming materials. vi.
Innovativeness - No. of new products introduced per year, lead time to new design, level of R&D investment, consistency of investment over time
18
2.6. Resume
Based on the review on fours journals listed, it shows that applying given principles and analogy will have high potential in improving overall Throughput in semiconductor manufacturing environment. Refer to summary Table 2.1 below.
Table 2.1: Summary of Literature Review
CHAPTER 3
MODEL DEVELOPMENT
3.1.
Overview
In this chapter, we will cover important and practical methodology and approaches for Throughput performance measures.
The purpose of this chapter mainly to ensure factory business models are well understood to be able to analyzed potential gain at any stage. From technical perspective, throughput measures concepts and approaches and execution strategy are discussed including the development of sustainable models as to serves as baseline for proactive die loading planning. This model expect to integrate and loop all critical throughput performance measures with the understanding of loading schedule by devices and variations, conversion and machine down time allowances and process flow and characteristic differences to meet given cycle time.
20
3.2. Throughput Financial Modeling
Throughput Financial Modeling and Analysis were performed to understand actual business model. A simplified revenue vs cost calculations logic and assumptions applied as listed below: i.
Revenue equal to output based on target constraint module output for baseline cost calculation. Factory output equal to constraint output based on Theory of Constraint. ii.
Value Add (VA) = Average Selling Price (ASP) – Material Cost iii.
Gross Margin $ Saving = 90% VA($/Unit) * Incremental Output iv.
Cost $/Kunit Saving = 90% VA($/Unit) / Incremental Output (Ku/Wk)
Example of calculations is illustrated in calculation Table 3.1 below.
Table 3.1: Financial Modeling
VA Calculation for Constaint Operation
Factory Output (Ku/Wk)
ASP
Factory Rev ($K/Wk)
Material Cost ($K/Wk)
Units Material Cost
VA [value add]
22025
$ 0.099
$ 2,190
$ 1,095
$ 0.05
$ 0.0497
Note: refer to finance
Note: refer to finance on material usage % over Revenue
GM $ Calculation for Constraint Operation
Constraint Output (Ku/Wk)
GM$
22025
= 90% VA[value add] * Constraint Output (Ku/Wk)
= 90% * 0.0497 * 22025
= 985.47
Cost $ / Unit Calculation for Constraint Operation
Constraint Output (Ku/Wk)
Cost $/Unit
22025
= 90% VA[value add] / Constraint Output (Ku/Wk) *1000
= 90% * 0.0497 / 22025 *1000
= 0.0020
0.00203
21
3.3.
Throughput Concept and Approach
There are 2 types of Throughput required to be computed. They are Plan
Throughput and Actual Throughput.
Plan Throughput: Velocity (UPH
V
) is use to level set on plan/install UPH
V
capability in current system. It is important as it also use to identify gaps to achieve balance line through UPeH and OEE improvement with designed constraint. It can be representing as formula UPH
V
= UPeH * OEE
Velocity (UPH
V
) use as basis of throughput measure as it can be represent based on Basic Queuing Theory. Figure 3.1 illustrated the relationship.
Flow based on Arrival rate ( λ )
λ
Queue
µ
Velocity based on Install capacity (µ)
1 Velocity (µ) > Flow ( λ )
2 WIP (L) = Throughput ( λ ) * CT (W)
To increase Throughput ( λ ) need to increase WIP (L) to right level with given CT (W).
L = λ /(µ - λ )
3 Since Velocity (µ) > Flow ( λ ); Need to focus on µ ensure it is not limiter for λ .
Figure 3.1: Basic Queue Model
Actual Throughput – Integrate Flow (UPH
F
) into Velocity (UPH
V
). By integrating Flow into Velocity performance measure to materialize the ultimate cost saving. Any potential variability in production line to close gap between UPH
F
&
UPH
V.
It can be represent as formula UPH
F
= Actual Output
Figure 3.2 illustrations on Ideal Production System (Apply Theory of Constraint)
22
Plan
Throughput
Input/Loading
DA
UPH
V1
WB
UPH
V2
UPH
V3
BE
To increase Velocity need to improve UPH
V2
Velocity/Install Capacity :
UPH
V
= Min UPH
Vi
= Min UPeH i
* OEE i
Actual
Throughput
DA WB
Input/Loading
UPH
F1
UPH
F2
UPH
F3
BE
To increase Flow rate need to improve UPH
F2 to meet UPH
V2
Flow/Throughput : UPH
F
= Min UPH
Fi
3.4.1.
Figure 3.2: Ideal Production System in Detail
3.4.
Throughput Execution Strategy
Factory capacity balancing
Doing right level equipment dedication at package level is required. This will then align to factory constraint by shaping factory capacity planning. Machines are dedicated by package level. Figure 3.3 is the pictorial diagram below illustrates the designed capacity.
23
3.4.2.
Figure 3.3: Design Capacity
Capacity Interdependencies
Machine capacity are computed based on Unit Per Hour (UPH) and Overall
Equipment Efficiency (OEE) within a predetermine manufacturing cut off interval manufacturing time. Weekly basis is commonly applied but it can also being computed in daily or shiftly basis. Machine Capacity = UPH * OEE % * 168 hours
Machine UPH consist of pure processing cycle time and standard rework or retest perform by machine automatically should also be comprehend as part of standard UPH.
Machine OEE consist of forecasted downtime either schedule or unscheduled as well as setup time or conversion required to operate the machines. In some cases,
24 special periodic machine monitoring required then it also comprehend as part of
OEE standard.
Manufacturing hours is straight forward. It is basically numbers of hour’s machine run relative to specific cutoff time measures. 168 hrs/week is commonly used to represent 7 days/Wk * 24 hrs a day. Figure 3.4 illustrates the detail machine breakdown.
3.4.3.
Figure 3.4: Machine Capacity Breakdown
Work in Progress (WIP) and Cycle Time Management
In order to obtain constant throughput with desire cycle time. All possible variability on the system needs to be reduced. Optimization between WIP and cycle time are essential. High WIP in system will create high cycle time. To keep cycle time low for faster material turn, we need to ensure WIP are always at manageable level which is not too high.
25
Below are few practical example of WIP monitoring mechanism. Proactive planning on product line loading involves reconciliation on beginning on hand
(BOH) WIP and expected end on hand (EOH) WIP after minus out target input and additional assembly starts. These is by assuming all required resources such as equipment, headcount, raw materials are all in-place and aligned to build request per given cycle time.
Gradual reduction in WIP/inventory level helps to surface and identify areas for continual improvements as illustrates in Figure 3.5 below.
Figure 3.5: Sea of Inventory
Factory cycle time is compute based on the sum of each process processing time. Normally, theoretical estimate were done as start to understand the system potential. Theoretical lot cycle time includes pure processing time and queue time.
Cycle Time = Queue Time + Pure Processing Time
Sometime only pure machine processing time for a lot being computed as theoretical cycle time (without queue time) to understand minimum cycle time can stretch to without having lot to queue. This is also good to serve as basis for machine mechanical and process improvement.
26
Timing diagram can be use as alternatives to sequence lot movements from operation/process to operation/process if required. In most cases, actual lot movement can also be monitor through and production lot movement and tracking system such as MES plus, Workstream, Promise and etc…Pull and push system need to apply to ensure there is no excessive staging and queuing for lot throughout the process. Figure 3.6 illustrated the pull and push system.
Machine Group 1 push full wafer run in cassette to Machine Group 2 base on shiftily target output and capacity.
Wafer run will split into Sublot after Machine Group 2 process and push to Machine Group 3 station in WIP 1 station.
Machine Group 4 will pull Sublot from WIP 2 station to process lot in Sublot level. WIP 2 station serves as Magazines
Lot will travel in Sublot level which push from Machine Group 5.
Figure 3.6: Pull and Push System
Die release based on pull and push can also be further elaborate in flow diagram in Figure3.7 below in terms of material batching and releasing based on constraint tool capacity.
27
Figure 3.7: Die Issuance Flow
Clear communication, calibration and education to all key stakeholders are very important to setup common pace to enable smooth execution.
3.4.4.
Develop Sustainable Models
Several models will be developed at different aggregate planning stage to integrate critical information and using is as basis for productivity computation.
Looping all critical throughput performance measures with the understanding of
Loading schedule by devices and variations, conversion and machine down time allowances, process flow and characteristic differences to meet given cycle time.
28
The concept information integration model to close loop process communication are illustrate in Figure 3.8 at macro level:-
Figure 3.8: Flow Integration
Capacity are core information to facilitate the balancing among the Assembly starts quantity, Assembly output quantity as well as optimizing WIP on hand to maximize throughput while execute within given cycle time.
There are few pre-implementation modeling requirements and checks that need to watch, modeling at lowest process requirement level to understand conversion requirement, data input and reconciliation at shiftly level to minimize linearity impact as static modeling approach being use cross modeling methodology.
3.4.5 Resume
In summary modeling concept, methodology and approaches being discussed in this chapter to enable good throughput improvement plan.
29
CHAPTER 4
PROJECT IMPLEMENTATION AND RESULTS
4.1.
Overview
Preliminary data analysis was performed based on first 3 months pilot data.
This is mainly to understand the trend on demand projection and possible forecast fluctuation. The complexity of install capacity changes due to demand and product mix changes were also studied in order to setup basis for Throughput improvement to match the holistic Throughput target. Financial Analysis (Target Saving) aligns to
Plan Throughput also being analyzed.
In summary, first 3 months data were collected as table show below where
Plan Throughput does have potential to improve from 19.3Mu/Wk to 22.8Mu/Wk based on 1 st
limiter throughput capability. Chronological changes from month to moth were also being monitor and tracked, these includes equipment inventory changes, UPH improvement, product mix changes and others where those are key parameters has direct impact to Throughput performance. Data summarized in Table
4.1 below.
30
Table 4.1: Capacity Limiter Chart
FORECAST
1st LIMITER
Equip Capacity (Ku/Wk)
Equip Inventory
Est Capacity (Ku/Wk) Changes
1) Product Mix
2) Additional/Reduction Equip
3) UPH Improvement
4) Others
Total Changes
2nd LIMITER
Equip Capacity (Ku/Wk)
Equip Inventory
Est Capacity (Ku/Wk) Changes
1) Product Mix
2) Additional/Reduction Equip
3) UPH Improvement
4) Others
Total Changes
3rd LIMITER
Equip Capacity (Ku/Wk)
Equip Inventory
Est Capacity (Ku/Wk) Changes
1) Product Mix
2) Additional/Reduction Equip
3) UPH Improvement
4) Others
Total Changes
Month 1
Machine W
19376
527
-
2323
-
-
2323
Machine D
20552
64
-
3690
-
-
3690
Machine D
20552
64
-
3690
-
-
3690
Month 2
Machine W
22025
563
-1970
1511
1240
900
1681
Machine D
22336
58
-432
0
4800
-1320
3048
Machine L
23581
8
-1204
0
0
0
-1204
Month 3
Machine D
22814
57
858
-380
0
0
478
Machine W
23046
573
517
504
0
0
1021
Machine L
23959
8
378
0
0
0
378
Assuming Plan Throughput able to sustained at ~22.8Mu/Wk, target incremental Gross Margin can go up to $48K/Wk and Cost $/unit of 0.0013 saving expected. Table 4.1.2 below is the summary detail of financial analysis.
Table 4.2: Throughput Financial Analysis
VA Forecast
Target Output (Ku/Wk)
ASP
Target Rev ($K/Wk)
Material Cost ($K/Wk)
Units Material Cost
VA [value add]
Month 1
19376
$ 0.099
$ 1,916
$ 916
$ 0.05
$ 0.052
Month 2 Month 3
22025
$ 0.099
$
22814
0.099
$ 2,190
$ 1,095
$ 0.05
$ 0.050
$
$
$
2,268
1,134
$ 0.05
0.050
Target GM $
Machine D
Machine W
Machine L
Target Cost $/ Unit
Machine D
Machine W
Machine L
Target Output Ku/Wk
Month 1 Month 2 Month 3
20552
19376
20552
22336
22025
23581
22814
23046
23959
Target GM $K/WK
Month 1 Month 2 Month 3
$ 955
$ 900
$ 955
$
$
$
999
985
1,055
$
$
$
1,021
1,031
1,072
Incremental GM $K/Wk
Month 1 Month 2 Month 3
$ 4
$ (62)
$ (54)
$ (122)
$
$
(87)
(47)
$
$
$
48
24
46
Target Output Ku/Wk
Month 1 Month 2 Month 3
20552
19376
20552
22336
22025
23581
22814
23046
23959
Target Cost $/ Unit
Month 1 Month 2 Month 3
$ 0.0023
$ 0.0024
$ 0.0023
$
$
$
0.0020
0.0020
0.0019
$
$
$
0.0020
0.0019
0.0019
Cost $/ Unit Saving
Month 1 Month 2
$ 0.0011
$ 0.0009
$ 0.0009
$
$
$
0.0011
0.0012
0.0012
Month 3
$
$
$
0.0013
0.0013
0.0013
31
WIP reduction activities were pilot together with cycle time improvement to ensure all activities can be effectively synch-up with cycle time improvement.
Preliminary results show that cycle time within target with proper WIP management.
In summary, this project has high potential Throughput as well as efficiency improvement through cycle time measures. Besides tangible improvement list, those intangible implications such as process sustainability and change management were apply in parallel to ensure project success.
4.2.
Project Implementation Phases 1
Mathematic Models being develop as basis to consolidate, calculate and standardize performance measures. This was done after clear understanding on current process addition and limitation. Purpose is to sustain good process and enhance on current limitation. Total four static excel models were created. These includes:- i.
Machine Throughput Model ii.
Factory Capacity Model iii.
Loading-Output-WIP Model iv.
Cycle Time –Lot Size Model
4.2.1.
Machine Throughput Model
This is to model at lowest aggregate planning level parameters. The purposes of this model are to consolidate, calculate and standardize shop floor machine level performance parameters. There are few key parameters includes Effective Unit per
Hour (eUPH) which is basically represent average numbers of units a machine can
32 process in an hour, Overall Equipment Efficiency (OEE) which is equipment efficiency measures after considering PM, setup, conversion, downtime and etc...Machine Availability (MA) is where the time when machine is not down due to scheduled or unscheduled. Variability Gap (VG) is basically a form of buffer expression to protect factory against equipment, labor & WIP, Manning Ratio (MR) is calculated values to represent numbers of machines an operator can operate at any one time and lastly Theoretical Throughput Time (TPT) were computed to understand the theoretical time required for a lot to complete the entire processes
Figure 4.1 is pictorial diagram illustrate conceptual machine throughput modeling. It is express in following mathematic equations associate to all parameters and metrics listed.
Machine Throughput = eUPH * OEE * Mfg Hrs
1. Define second per unit (pure machine time)
2. Define lot to lot setup time
3. eUPH = 3600 /(Second/unit + Lot to lot Setup)
1. Define Availability
• Aggregate Down Time for both schedule and
Unscheduled to Mfg hrs
• Availability = Mfg hrs – Aggregated Down Time
2. Define Variability Gap
3. OEE = Availability / (1 + Variability Gap)
1. Mfg hours = 168 hours a week (7X24)
Note : The above is assuming Labor/headcount resources is always there.
Figure 4.1: Machine Throughput Modeling
Sample outlook on static mathematic machine throughput model in Table 4.3 where it integrates all machine parameters and compute the standardize performance measures. Input cells are shaded in yellow and others are auto calculated based on mathematic formula setup during modeling.
33
Table 4.3: Throughput Model eUPH (Effective Unit Per Hour) is defined as average number of units that can be produced in 60 minutes when machine is up to production which includes any task associated with a lot. All the lot to lot setup factors in addition to the pure processing time of the machine are accounted to create an effective or average unit per hour.
eUPH = 3600 /(Second/unit + Lot to lot Setup)
34
Table 4.4 is the snap shot of model interfaces for eUPH computation.
Table 4.4: eUPH Computation
OEE % (Overall Equipment Efficiency %) is computed based on the % of time a machine is capable of producing product.
OEE = Availability / (1+Variability Gap)
Figure 4.2 below is the conceptual model illustrate on the OEE% measures aggregation in 1 week. (168 hours/wk)
• Total Mfg Hours = 168 hours/Wk (7 X 24)
• Availability = 168 hrs/wk – Down Time
• Down Time = Aggregated Schedule +
• Variability Gap = (Availability – OEE)/OEE
Figure 4.2: Total Manufacturing Hour Breakdown
Availability is the time the machine is not down due to scheduled or unscheduled PMs
35
Availability = 168 hrs/wk – Down Time
Down Time includes the following:- i.
Scheduled Down Time (Hrs/Wk) = PM/Calibration + Conversion +
Shiftly change consumable or setup ii.
Unscheduled Down Time (Hrs/Wk) = Total Repair + Total Assist iii.
Total Repair = 168 * MTTR (hrs) / MTBF (hrs) + a.
MTBF: Mean Time Between Failure b.
MTTR: Mean Time to Repair iv.
Total Assist = 168 * MTTA (min)/60min / MTBA (hrs) a.
MTBA: Mean Time Between Assist b.
MTTA: Mean Time to Assist
Variability Gap It is an expression of buffer capacity required at tool level to ensure factory are protected against machine, labor and WIP variability. This will help to ensure the non constraint tools can consistently feed the constraint tool. The variability gap will be smaller at constraint tools.
Variability Gap = (Availability – OEE)/OEE
Variability Gap is needed as in order to run equipment we need equipment to be available, WIP to be present, Labor (MA or Tech) to run the WIP at require time.
Managing operation variability is difficult because not all of the equipment is available at all time, not all of the MA and Tech are available at all time and the right amount of WIP is not always there. Therefore, these resulted factory throughput being unpredictable and Factory throughput time (Cycle Time) being unpredictable.
Variability propagate whenever there is variation in equipment availability leads to uneven WIP levels, variation in labor (MA/TECH) availability, add to
36 uneven WIP levels. The resulting variation in WIP flows and levels lead to variation in factory throughput.
We can control variability in various ways. Engineering work to reduce variation in equipment availability by improving equipment reliability, reducing failures and assists, decreasing setup, PM etc…Operation to address variability by buffering WIP and Capacity. Buffers are necessary to help minimize and dampen the effects of upstream variability. This designed to help compensate for variations in
WIP flows through buffer or dedication, ensure labor is present to run WIP, model to have additional excess equipment capacity by adding Variability Gap.
Things to consider in defining Variability Gap includes normalized it against
OEE% (independent of OEE). Understand actual tool performance and sometime automation is necessary for real time performance update. Number of similar equipment in the pool also plays some roles. Note that the more similar equipment in the pool, the lesser the variability gap required. Besides, at which operation the type of equipment is position in the line and variability of upstream process and equipment are equally important.
Table 4.5 below is the snap shot of model interfaces for OEE%
Table 4.5: OEE Computation computation.
37
Manning Ratio (MR) is a study on labor efficiency which applies to
Manufacturing Assistance (MA) who is assign to operate multiple machines at same time. Work methods and elements being evaluated to optimize the utilization of equipment with given availability (eg. if MR 1:16 this means 1 MA capable to manage 16 machines at given time without impact equipment throughput and within labor utilization standard.
There are 3 key elements in computing manning ratio. External Labor
Elements Time represents MA activities time which needs to be done before machine can continue/start to operate. In other words while MA performing External
Elements, machine expect to be waiting/idling. These includes schedule and unscheduled down time need MA assistance.
Internal Labor Elements Time represent MA activities time while can be perform concurrently which no stoppages on machines expected. No net loss in equipment productivity. Machine time actually express in terms of machine cycle time
There are 2 levels of MR calculation: i.
Level 1 MR : Assuming no interferences exist a.
Level 1 MR (Manning Ratio) = (External + Internal Labor Time) /
(External Labor + Machine Time) ii.
Level 2 MR : Added interferences factor (Commonly 15% added to offset the interferences factor) a.
Level 2 MR (Manning Ratio) = Round down (Level1 MR * 85%)
Table 4.6 is the snap shot of model interfaces for MR computation.
38
Table 4.6: Manning Ratio Computation
Theoretical Lot Throughput (TPT) define as the amount of time that it takes a unit to be process through the entire process line. Lot throughput time is use as basis to forecast On Schedule Delivery (OSD) and Cycle Time calculation. Total sum of processing time and queue time at each process operations were use to compute the factory cycle time.
It this modeling, pure machine processing time, number of equipment sharing per lot, equipment fill-up time as well as estimated queue time were comprehended.
Theoretical Lot Throughput time (min) = Equipment Fillup Time + Lot Queue Time
+ Units per machine / eUPH. Table 4.7 is the snap shot of model interfaces for TPT computation.
Table 4.7: TPT Computation
39
4.2.2.
Factory Capacity Model
The purposes of this model is to consolidate individual machine throughput by product when then integrate it with mix of demand forecast and requirement at factory level. This model will then be use to maximize daily production plan against throughput capacity.
Optimization on the model will be done by adjusting loading to ensure all plan die loading within machine capability with gradual conversion schedule at given time without execution jerk cause high variability.
General business process and procedures involves consolidation of demand and die schedule from customer and sales. All die loading/scheduling plan must align to available equipment throughput capability avoid over loading within given time cross operations for each product line item. Generic guidelines for loading schedule: i.
Ramp rate align to capable conversion rate ii.
Device variety and quantity of small lots iii.
Recovery need if there is any miss in previous schedule iv.
WIP in pipe line should not more than cycle time target. Upfront loading control if needed. v.
Latest equipment throughput capability (eUPH and OEE) must be validated by key stakeholders. vi.
Optimized plan will then be review, calibrate and buyoff in daily
Operation meeting prior execution.
Table 4.8 below is the snap shot of the portion of model interfaces for demand schedule from customer and sales. Load leveling can be done here to ensure workload balancing from operation shift to shift.
40
Table 4.8: Input of Capacity Model
Device Pkg density Body size
CS4353-CNZR/B1
TLG1100GC
ISL6236IRZA-TR5281
ISL6262ACRZ-T
21032BSQFNE3CSS
HD
HD
ISL6208CRZ-T HD
3215TEH-5730-CEA-Y8-POL STD
HD
HD
STD
ISL65426HRZ-T
ATA5812-M80MLP
ATA5823-M80MLP
CS4206A-CNZR/C1
8835TEJ-7719-ABA-Y8-POL HD
I22/FGPNURD0680U0AE8 STD
ISL6237IRZ-T STD
STD
STD
STD
STD
3x3
5x5
5x5
5x10
7x7
7x7
6x6
7x7
4x4
3x3
3x2
4x4
8x8
5x5
40
0
0
67
0
0
0
Mon Mon Tues Tues Wed Wed Thurs Thurs Fri
21 42 78 78 72 85 106 106 106
100
0 0
70 50
0 20
0
30
0
52
0
52
0
52
0
52
0
121
42
100
0
121
63
100
0
200
63
80
48
220
53
80
48
220
42
0
48
220
40
48
0
220
40
0
0
220
40
0
0
250
40
0
0
67
0
0
0
0
0
45
15
45
0
0
0
0
0
0
0
0
32
0
12
45
0
0
0
0
10
0
45
24
24
30
45
0
0
0
0
0
52
0
50
0
45
53
50
45
0
0
0
0
50
0
45
0
56
0
20
50
0
45
58
0
20
0
Fri
106
52
6
250
40
0
0
50
0
45
0
58
0
0
Sat Sat Sun Sun
106 106 106 106
52
6
52
6
52
6
52
6
350 350
40 40
0
0
0
0
350
40
0
0
350
0
0
0
0
58
0
0
50
0
45
58
50
45
0
0
0
0
0
58
0
0
50
0
45
Table 4.9 below are the sample 2 processes machine inventory, jig and fixture checks model interface. This is important to ensure specific requirement for product line item are considered.
Operation
Process 1
Package
Package A
Machine allocation
Delta
Package B
Machine allocation
Delta
Total req't
Machine allocation
Delta
Table 4.9: Machine Inventory Capacity Model
Mon D/S Mon N/S Tue D/S Tue N/S Wed D/S Wed N/S Thu D/S Thu N/S
0.57
0.47
0.33
0.37
0.17
0.40
0.42
0.21
0.80
0.23
0.80
0.33
0.80
0.47
0.80
0.43
0.80
0.63
0.80
0.40
0.80
0.38
0.80
0.59
0.45
1.20
0.75
0.48
1.20
0.72
0.62
1.20
0.58
0.80
1.20
0.40
0.63
1.20
0.57
0.50
1.20
0.70
0.70
1.20
0.50
0.73
1.20
0.47
1.02
2.00
0.98
0.96
2.00
1.04
0.95
2.00
1.05
1.17
2.00
0.83
0.79
2.00
1.21
0.90
2.00
1.10
1.12
2.00
0.88
0.94
2.00
1.06
Fri D/S
0.27
0.80
0.53
0.86
1.20
0.34
1.13
2.00
0.87
Fri N/S
0.28
0.80
0.52
0.87
1.20
0.33
1.15
2.00
0.85
Sat D/S
-
0.80
0.80
0.83
1.20
0.37
0.83
2.00
1.17
Sat N/S Sun D/S Sun N/S
-
0.80
0.80
0.80
0.80
0.80
0.80
0.84
1.20
0.36
-
1.20
1.20
-
1.20
1.20
0.84
2.00
1.16
-
2.00
2.00
-
2.00
2.00
Process 2
Package A
Machine allocation
Delta
Package B STD
Machine allocation
Delta
Package B HD
Machine allocation
Delta
Package C
Machine allocation
Delta
Total req't
Machine allocation
Delta
-
4.00
4.00
9.21
14.00
4.79
12.83
27.00
14.17
5.86
8.00
2.14
15.07
53.00
37.93
-
4.00
4.00
7.33
14.00
6.67
14.13
27.00
12.87
4.90
8.00
3.10
12.23
53.00
40.77
-
4.00
4.00
6.61
14.00
7.39
16.81
27.00
10.19
4.94
8.00
3.06
11.55
53.00
41.45
-
4.00
4.00
6.68
14.00
7.32
20.39
27.00
6.61
4.94
8.00
3.06
11.62
53.00
41.38
-
4.00
4.00
5.75
14.00
8.25
16.61
27.00
10.39
5.32
8.00
2.68
11.07
53.00
41.93
-
4.00
4.00
9.13
14.00
4.87
11.81
27.00
15.19
5.10
8.00
2.90
14.23
53.00
38.77
-
4.00
4.00
9.31
14.00
4.69
17.08
27.00
9.92
2.39
8.00
5.61
11.71
53.00
41.29
-
4.00
4.00
2.18
14.00
11.82
17.81
27.00
9.19
-
8.00
8.00
2.18
53.00
50.82
-
4.00
4.00
2.71
14.00
11.29
20.29
27.00
6.71
0.96
8.00
7.04
3.67
53.00
49.33
-
4.00
4.00
2.75
14.00
11.25
17.29
27.00
9.71
1.27
8.00
6.73
4.02
53.00
48.98
-
4.00
4.00
-
14.00
14.00
16.10
27.00
10.90
-
8.00
8.00
-
53.00
53.00
-
4.00
4.00
-
14.00
14.00
16.36
27.00
10.64
-
8.00
8.00
-
53.00
53.00
-
4.00
4.00
-
14.00
14.00
-
27.00
27.00
-
8.00
8.00
-
53.00
53.00
-
4.00
4.00
-
14.00
14.00
-
27.00
27.00
-
8.00
8.00
-
53.00
53.00
Table 4.10 below is the sample of number of machine conversion expected aligned to demand loading. In case the number of conversion excess capability leveling the demand schedule required.
41
Customer
DIALOG
DIALOG
FREE
FREE
FREE
SMIC-TJ
CNXT-W
CNXT-W
CNXT-W
CNXT-W
CNXT-W
CNXT-W
Device
D1755AD-NB3J-FA2
D1759AC-NA3J-FC2
99CD90-V2373-1D
99CD90-V2685-2J
99CD90-V2920-1F
CD90-V9925-1MTR
11270-11Z-002-A
20493-21-050-A
20493-31-049-A
20493-51-051-A
20493-52-052-A
20493-53-053-A
Table 4.10: Machine Conversion Model
STD
STD
STD
STD
STD
STD
STD
STD
STD
STD
STD
STD
Pkg density eUPH OEE
80 75%
Throughput
Capacity Mon D/S Mon N/S Tue D/S Tue N/S Wed D/S Wed N/S Thu D/S Thu N/S Fri D/S Fri N/S Sat D/S Sat N/S Sun D/S Sun N/S
(Ku/shift)
BCCSTDMaxum 4 4 4 4 4 -
138
197
75%
75%
56 75%
170 75%
91 75%
1,218
865
75%
75%
842
363
780
794
BCCSTDMaxum 10
BCCSTD8028 -
BCCSTD8028 2 -
BCCSTD8028
BCCSTD8028
11.0
QFNpSTDMaxum
QFNpSTD8028
-
-
7
4
10
12
7
4
-
75%
75%
75%
75%
QFNpSTD8028 12 12
QFNpSTD8028 3 -
QFNpSTD8028 -
QFNpSTD8028 -
-
-
3
-
-
-
-
-
10
24
12
12
3
4
10
-
36
12
-
-
-
5
4
4
7
-
-
-
-
-
10
7
48
12
4
10
10
-
48
-
-
-
-
-
7
7
4
-
-
-
-
-
-
10
48
7
7
4
10
7
-
48
-
-
-
-
-
4
-
-
-
-
-
10
-
-
48
7
4
-
-
10
48
-
-
-
-
-
7
4
-
-
-
-
-
-
-
10
48
7
4
-
-
-
-
10
7
-
48
-
4
-
-
-
-
-
10
7
-
48
-
4
-
4.2.3.
Input – Output - WIP Model
This model is used to govern the commit input plan with output taking into considering expected WIP on hand. This is a simple model use to avoid excessive
WIP pile-up or bubble up in pipe line. It also gives good estimates on projected output by end of the cutoff. In case actual end on hand (EOH) WIP is higher than
WIP target to meet cycle time target, upfront input control needed.
Beginning On Hand WIP (BOH) is cumulative from previous cutoff. End On
Hand WIP (EOH) is calculated value base on formula below:-
BOH + Loading – Target output (FOL / EOL) = EOH
Table 4.11 below is snap shot of Model summarized table where is clearly show Input-Output-WIP status for 2 consecutive weeks. Note: Font in red means missed target, Output shaded in same color are from same input batch.
42
Table 4.11: Input-Output-WIP Model
Work Week : XX
Mon Tue
Assy Operation A
BOH
Loading /In
4.58
2.77
Target FOL Output 2.79
EOH (WIP) 4.55
4.55
2.68
2.75
4.48
Wed
4.57
2.75
2.67
4.65
Thu
4.45
2.87
2.73
4.59
Fri
4.59
2.96
2.57
4.97
Sat
4.97
1.66
1.02
5.61
Sun
Work Week : XX + 1
Mon Tue Wed
5.61
3.30
2.80
6.11
6.16
2.54
2.69
6.02
6.02
2.51
2.60
5.93
6.02
2.54
2.58
5.98
Thu
5.78
2.65
2.65
5.78
Fri
5.78
2.58
2.76
5.60
Sat
5.60
2.62
2.81
5.41
Sun
5.41
2.56
2.81
5.16
Assy Operation B
BOH
Input EOL
Output
EOH (WIP)
5.88
2.79
2.49
6.18
6.18
2.75
2.61
6.32
6.32
2.67
2.84
6.15
6.35
2.73
2.89
6.19
Total Loading 19.0
Target Output 17.3
6.19
2.57
2.47
6.29
6.29
1.02
1.93
5.38
5.38
2.80
2.80
5.38
5.22
2.69
2.07
5.83
5.83
2.60
2.17
6.26
6.26
2.58
2.86
5.98
6.18
2.65
2.93
5.90
Total Loading 18.0
Target Output 18.9
5.90
2.76
2.93
5.73
5.73
2.81
2.94
5.60
5.60
2.81
2.89
5.52
Total Output from Assembly Operation A 17.3
Target EOL Output 18.0
Total Output from Assembly Operation A 18.9
Target EOL Output 18.8
Table 4.12 below is the snap shot of 2 nd
level of model raw input extract from actual production system. This is where the breakdown composed from raw data input for Input-Out-WIP by devices by operation stated in operation code.
Table 4.12: Detail Input-Output-WIP Model
Red – High WIP
Black -normal
Yellow -low WIP
Blue - expedite, Priority 1
WIP numbers flag in RED font means WIP exceed cycle time allowable target. BLACK font means WIP at normal range, YELLOW font means WIP at low level and BLUE means there is expedite devices WIP need to clear as in 1 st
priority.
43
The logic for WIP trigger is basically based on the predefined cycle time allowed. Formula below are the example if overall cycle time set at 4 days and pure cross operation processing time at 3 days. (cycle time without queue time) i.
Min WIP = Average daily input for last 3 days input * 3 days cycle time ii.
Maximum WIP = Average daily input for last 4 days input * 4 days cycle time.
4.2.4.
Cycle Time – Lot Size Model
Lot size impact cycle time. The bigger the lot sizes the longer the cycle time.
To define optimum lot size which able to meet cycle time target, Cycle Time-Lot
Size Model was developed based on Min Cycle Time Lot Size Technique from
Factory Physic were applied. General procedures involve includes:-
Optimum lot size was defined by analyzing incoming batch size, types of material handler media such as units per tray, units per magazine and others indirect material; etc against duration of processing in an operation. Lot size were adjust till desire cycle time achieved after associate it with machine capacity.
Table 4.13 below is the snap shot of Cycle Time- Lot Size Model use to determine optimum lot size based on static mathematic calculations.
44
Table 4.13: Cycle Time Lot Size Model
4.3.
Implementation Phase 2
4.3.1.
Throughput Monitoring
Monitoring system and business process setup is important to ensure good practices are able to monitor and sustain all the time. Actual throughput performance parameters and data were extracted daily from shop floor production system for performance analysis. The information were consolidated and summarized monthly for operation review. This has become important steps to understand performance baseline before projecting the demand outlook from sales and customers.
45
There are 2 approaches in design monitoring. Monitor through automated IT system data extract which automatically distribute via email to key stakeholders and monitor through manual data extract from local data based which can customized for various improvement analyses.
Monitoring through automated IT system data extract which auto distributed via email to key stakeholders does have advantages and disadvantages. Advantages in this auto monitory are it serves very good and timely information update for 1 st level trigger. 1 st
level triggers such as “Meet” or “Miss”? The disadvantages here is it not able to provide 2 nd
level of information such as Why ahead or behind or suggest alternatives.
Monitoring through manual data extract from localized database has the advantages in analyzing the data based on user predefine requirement. All information for root cause analysis are possible through this approaches. The disadvantages on this is information overwhelm. It normally takes engineers time to extract the data manually and customized to various needs.
Cycle time performance report, actual throughput report are done through automated IT system data extract daily. All other analysis related reports such as Lot onhold report, queue and processing time lags report, machine UPH, conversion,
PM, downtime report are obtained through manual data log extract from various database.
In this project discussion, objective is to drive actual throughput improvement increased by 20% and cycle time reduction 30%. Based on the objective statement target key performance measures monitoring covers both Throughput and Cycle time.
46
Actual Throughput performances on the 7th month on key constraint machines were analyzed. Engineering improvement done in parallel to improve Plan
Throughput from 19Mu to 23Mu capacity per weeks includes:- i.
Wirebond loop height reduction for UPH improvement. ii.
Wirebond capillary life span extension to reduce machine setup time as part of OEE improvement. iii.
Wirebond machine to machine variation reduction improve UPH iv.
Mold Cleaning time reduction through process simplification for OEE improvement. v.
Die attach mechanical indexing time reduction for UPH improvement vi.
Oven cure profile optimization to reduce cure time for UPH improvement vii.
Laser marking sensor and sequence alignment for UPH improvement viii.
Chemical deflash batch processing improve conversion time for OEE improvement ix.
Punch singulation program synchronization to improve UPH x.
Enable offline / pre strip marking to shorten cycle time xi.
3 rd
Optical inspection sampling reduction xii.
Anneal bake elimination as process elimination
Besides introducing systematic approach is driving Throughput and Cycle
Time. All the improvement listed above are run concurrently and aligned to ensure plan throughput and cycle time target can be meet or exceed.
Table 4.14 below is snap shot example of Actual Machine Throughput monitoring using JUMP statistical analysis for key constraints. Example data below shows that performance of UPH
F
(Actual Throughput) at median is in par with UPH
V
(Plan Throughput) after bake in all the expected improvement with less than ~1% gap.
47
Table 4.14: Data Monitoring
Some additional statistical analysis was also conducted to further understand the performance behavior. Coefficient of variance is the dimensionless measures where obtained by dividing Std Deviation over Mean to give a representative on dispersion. Target vs mean good to be monitor periodically as this data will trigger us whether the change in plan throughput is required if they consistently away from mean.
Table 4.15 is the snap shot on analysis for machine performance above based on certain package body size.
Table 4.15: Coefficient of Variance
48
4.3.2.
Cycle Time Monitoring
4.3.2.1.
WIP and Cycle Time Report
Every lot status (operation code, operation name, customer, package, body size, lead frame type, lot no, customer device, create code, quantity, priority, factoryto-date, waiting/running status, OSD, start time, pending time, oper-to-date and machine number) are auto generated by extracting shop floor system. Key parameter definition in this report includes: - i.
Fac-to-date: Lot start launching time until time when report been generated ii.
Oper-to-date: Total period of time where lot in the operation until the time when report generated iii.
OSD: Scheduled delivery date for the particular lot
Table 4.16 below is the example of report snap short in Window based interface and automatically generated.
Table 4.16: WIP and Cycle Time Report
49
4.3.2.2.
Daily Cycle Time Report
Daily cycle time report auto generated and send through mail in daily basis. It include cycle time by day and WTD (week-to-date) based on week cut off. Cycle time is reported based on Average, 95% tile and 100% tile for every customer with their respective package.
Table 4.17 below is the examples of report snap short in Window based interface and automatically generated.
Table 4.17: Daily Cycle Time Report
4.3.2.3.
Queue and Process Time by Lot Report
Daily Queue and Process time by lot report is auto generated in Web Static
Report. The purpose of this report is to use for internal improvement analysis. It consists of operation queue time, operation process time, start time and end time for every operation by lot.
50
Definition/Formulation of queue and process time calculation includes:- i.
Oper Queue Time: period of time when calculate from end time of previous operation until start time of this particular operation ii.
Oper Process time: period of time when calculate from start time of this particular operation until end time of this particular operation
Table 4.18 below is the examples of report snap short in Window based interface and automatically generated.
Table 4.18: Queue and Process Time by Lot Report
51
The data were also translated into histogram manually for further trend monitoring and analysis. See Figure 4.3 below:-
Cycle Time Performance Monitoring – Queue Time and Process Time Plot
1.26 1.23
1.29
210 211 220 223 224 228 230 232 234 240 245 248 249 250 254 255 258 259 260 262 265 266 268 270 280 281 290
21 0 2 11 2 20 22 3 2 24 22 8 23 0 2 32 2 3 4 2 40 2 45 24 8 24 9 2 5 0 25 4 2 55 2 5 8 2 59 2 60 2 62 26 5 2 66 26 8 27 0 2 8 0 2 8 1 2 90
0.71
0.77
Figure 4.3: Queue Time and Process Time Plot
Specific On-hold lot analysis was also done by team to ensure there is no excessive lot staging time. The goal should be less than 0.5 days for these lots. Table
4.19 below is the example of analysis review weekly.
Table 4.19: Lot On-Hold Analysis
52
4.3.2.4.
Conversion Model for Execution Analysis
Conversion strategy is always important to maximize throughput with minimum cycle time. Attached is the example of conversion tracking done manually everyday beginning of shift before to decide on actual conversions requirement aligned to target throughput based on available WIP in particular stage.
If WIP is too high, machine capacity constraint and expected too long in cycle time to clear the WIP, decision will be made to reduce or stop input based on push and pull process methodology discussed earlier. Table 4.20 below for example.
Table 4.20: Conversion Model for Execution
Target per
Conversion
Shift
Qty bonded 3rd
AL:7X7 ISL6266AHRZ-T 114.2
C7:3X3 ISL6208CRZ-T
ISL6609ACRZ-T 18.1
4.4 Resume
In summary this chapter covered overall implementation plans from planning in developing and introduced models to actual monitoring for each performance measures tracking.
53
CHAPTER 5
DISCUSSION
5.1.
Overview
A systematic approaches mainly utilizing Factory Physic Principles and
Theory of Constraint as basis was introduced to define factory throughput performance measures & analysis and monitoring system.
Performance measures & analysis was done by introducing and developing 3 levels of aggregated planning and Modeling from unit input to output. i.
Level 1 Performance Measures – Machine Throughput ii.
Level 2 Performance Measures – Factory Throughput which this is the stage to Integrate Machine Throughput & Customer Demand iii.
Level 3 Performance Measures - Balance between Input, Output and WIP for Optimal Cycle Time Performance
54
Monitoring System and Business process mainly concentrate on monitoring actual throughput and cycle time performance for sustainability after every cycle of continual improvements. This was offered in 2 levels data extraction i.
Level 1 Monitoring – Monitor through automated IT system data extract which automatically distribute via email to key stakeholders ii.
Level 2 Monitoring – Monitor through manual data extract to understand
2 nd
level detail information normally based on data triggering in Level 1.
This was done by extracting raw data from local data based which can customized for various improvement analyses typically using JUMP statistically data analysis tool.
5.2.
Review of Achievement
As to re-cap on the project objective, this is drive overall throughput performance improvement not less than 20% couple with 30% cycle time improvement within 7 months. This will then yield total incremental Gross Margin of 48K/Week based on current financial assumptions.
On throughput performance measures development perspective, fundamental aggregated planning models were developed and introduced which are able to integrate information, parameters and metrics at different aggregated planning stage.
Performance measures are also built in the models where systematic and scientific data analysis can now be done easily in managing the performance measures and analysis for the semiconductor assembly operation.
55
During execution to actual throughput performance, various Factory Physics
Principles were applied which includes, Theory of Constraint (Bottleneck
Management), Little’s Law (WIP = Throughput * Cycle Time), Minimum Cycle
Time Lot Sizing approaches, Push and Pull System, Conservation of Mass (At steady state In equal to out). Besides, the balancing on Cost vs Performance vs Delivery we review periodically.
On system monitoring a sustainable business process for continual improvement were developed. Total 3 window-base monitoring reports were introduced. These include, WIP and Cycle Report, Daily Cycle Time Report and
Queue and Process time by Lot Report. There are also various report can be manually extract further scientific analysis. The performance monitory systems basically were done by splitting it into 2 different monitoring approaches. i.
Monitor through automated IT system data extract which automatically distribute via email to key stakeholders ii.
Monitor through manual data extract from local data based which can customized for various improvement analyses.
Table 5.1 below summarized the Plan Throughput achieved and sustained at
20% improvement over seven months despite multiple changes on business strategies which includes tools shipped to other sister sites, multiple iterations of product mix changes due to demand forecast fluctuations.
56
Table 5.1: Summary of Throughput Achievement
Cycle time performance also successfully improved >30% through the same period of time. Figure 5.1 below illustrate the trend overtime.
Month 3Month 5Month 5Month 6Month 7Month 7-
Figure 5.1: Cycle Time Performance Trend
57
In summary, systematic approaches on throughput performance measures were successfully defined with proven install capability. However, due to global economic crisis, actual demand way below capacity resulted actual financial $ incremental Gross Margin and Unit Cost yet to be realized. Once the demand picks up and utilizes all the 20% incremental plan throughput capacity, financial gain is just be a given item. Cycle target were also met per plan with >30% improvement over 7 months.
5.3.
Critical Appraisal
5.3.1.
Project Strength
This project is able to standardized, level set a common baseline and pace within and cross organization team in driving Throughput and Cycle Time improvement.
Aggregated models in each planning stage are able to consolidate information and integrate key parameters & metrics and also compute plan throughput effectively. Aggregated models are built in Microsoft-excel is flexible, easy to interpret, manage and best to introduce to all level of stakeholders ranges from shop floor owners, engineers, supervisors to managers as well.
Formulations and algorithms built in mathematic excel models are standardized, transparent to all stakeholders, easy to modify to align with new team strategy for process fine tuning. Aggregated models are straight forward, handy for operation execution and training to excess anywhere at least for the start.
58
Automated reporting standardizes performance measure reporting and declaration cross functional groups. Automated reporting eases performance monitoring. This eliminates possibility of human glue and excessive time spends extracting and formatting shop floor data to monitor performance. Automated reporting also help to increased labor productivity by replacing and offloading daily highly repetitive compiling effort from engineers.
This project is also able to streamline supply chain for better value in maximizing factory throughput.
5.3.2.
Project Limitation
Overtime, excel aggregated models might not suitable in long run in terms of information control and documentation. Centralized database should be introduced for matured factory.
This project still has room for improvement such as to further reduce human glue into system. Aggregated models were developed strictly using static modeling approach which has potential is losing certain level of data granularity and accuracy.
Other modeling approaches such as dynamic modeling approach should also be considered when required.
The “optimize” solution stated might not be really optimized, it still based on human experience and judgment from time to time for decision making.
59
Automated reporting expects to trigger via email. Proper control needed to consolidate and further streamline individual reporting from time to time before it blown out to too many reports which carry certain level of duplicate information.
Last but not least, performance measures changes as business strategy change.
Modification on approach and modeling might required from time to time.
5.3.3.
Comments
Change management is key to shaping factory performance measures and monitoring as it impacting the way how people perform daily route work. Few learning from this project implementation:- i.
Ensure stay intact with key stakeholders at all time to facilitate the change. ii.
Make yourself available, supportive, helpful and attentive. iii.
Ensure approaches, methodology, algorithm, formulation use in setting up performance measures are always transparent to all at all time iv.
Educating key stakeholders. Help them to bridge the before and after gaps v.
Get people involve and participate throughout the change process vi.
Apply good listening skills to ensure inputs are taken from shop floor people. Be patient. vii.
Focus on process gap not people issues
60
5.4.
Future Research
Researcher should consider further explore dynamic throughput modeling
(eg. simulation or other tools) instead of using static excel mathematic modeling for higher degree of forecast accuracy. This includes quantifying interdependencies and relationship among Manning Ratio, OEE and UPH.
Figure 5.2 below show the concept of dependencies but it is almost impossible to quantify it quantitatively through static modeling.
Figure5.2: Relationship among Key Performance Measures
Real time cycle time prediction before lot closes for proactive cycle time lag trigger. This will be good research as it needs massive compilation of data overtime to predict cycle time performance behavior in sequent downstream operation.
Capitalize statistical data analysis in defining best approaches to further improve performance measure through repeatability and reproducibility.
61
5.5.
Resume
Systematic approach in managing throughput performance measures proven yield positive and desire results. This means that the theoretical modeling that introduced is applicable to actual industrial practices.
Anyway, the concept and integrated approach discussed in this project might be new to certain environment, there is always a good practice to do synergy check and balance before any immediate project proliferation to new environment. This is to ensure it meets one’s objective and need.
62
CHAPTER 6
CONCLUSION
In summary this project meets the initial objective and expected results by applying Factory Physics Principles, Theory of Constraint and Pull and Push System.
The approach of using Factory Physics as fundamental to define performance measures analysis and monitoring system proven effective but it might not the only way to applying sciences into manufacturing. Other approaches are encouraged to explore. The weight age on criticalness of one performance measures might change from time to time whenever there is a change in company business acumen and strategies. Therefore, it always recommended to revisit whenever needed. Change management need to apply to ensure smooth implementation for a sustainable, repeatable and reproduce-able continual improvements.
REFERENCES
Goldratt, E.M., and Cox, J. (1994). “The Goal”.
USA: North River Press.
Spearman, Mark L. (2006). “Factory Physics Principles for Managers”
Arizona: Factory Physics, Inc.
Sendil Kumar, C. and Panneerselvam R. (2005). “JIT-Kanban System”
International Journal of Manufacturing System
H’ng Gaik Chin (2004). “Performance Measures Relationship”
University Technology Malaysia. Thesis M.Sc
63
64
BIBIOGRAPGY
Marvin E.Mundel, Ph.D., P.E and David L. Danner, Ph.D., P.E. (1994). “Motion and
Time Study” (Seventh Edition)
USA: Prentice Hall, Inc.
Andrew S. Grove, (1995). “High Output Management” (2 nd
ed.).
USA: Vintage books Edition, Random House, Inc
Frederick Winslow Taylor, (2006). “The Principles of Scientific Management”
USA: Cosimo, Inc.
Jay HeiZer and Barry Render, (2006). “Operation Management” (Eight Edition)
New Jersey: Pearson International Inc.
Sunil Chopra and Peter Meindl. (2007). “Supply Chain Management” (3 rd
Edition)
New Jersey: Pearson International Inc.
Jim Collins, (2001). “Good to Great”
USA: Harper Collins Publisher, Inc.
Robert Heller, (2002). “Manager’s Handbook”
London: Dorling Kindersley Limited.