-中醫內科學(含中藥學)訓練基準(草案)-目 錄 頁碼 壹、課程基準 1 訓練課程與訓練期間 訓練內容 貳、參考指引 6 中醫婦科學訓練基準
02 spc訓練教材
description
Transcript of 02 spc訓練教材
1
Statistical Process ContrStatistical Process Control ol
統計製程管制統計製程管制
2
Chapter OutlineChapter Outline 概述概述Statistical Thinking and Statistical Methods 統計思維與統計方法 Statistical Process Control (SPC) 統計製程管制
Types of data 資料型態 Constructing control charts 如何架構管制圖 Interpreting control charts 管制圖之說明 Process capability 製程能力
Acceptance sampling 允收水準 Inspection process 檢驗程序 Quality measures 品質的量測 Sampling vs. screening 抽樣與篩選
3
Process 製程
Variation 變異
Data 資料
Statistical Tools 統計方法
Statistical Thinking Statistical Thinking
統計思維統計思維Statistical Methods Statistical Methods
統計方法統計方法
Statistical Thinking and Statistical Thinking and Statistical MethodsStatistical Methods
統計思維與統計方法統計思維與統計方法
4
Statistical ThinkingStatistical Thinking統計思維統計思維
Key Concepts 主要觀念 Process and systems thinking 製程與系統的思維 Variation 變異 Analysis increases knowledge 分析可以增加知識 Taking action 可以採取行動 Improvement 可以用來改善
Role of Data 資料的角色 Quantify variation 量化的變異 ( 變動 ) Measure effects 量測的效應
5
•
“You can’t improve a process that you don’t understand” 你若對製程不懂 , 就無法改善製程
Without a Process ViewWithout a Process View若無製程的觀點若無製程的觀點
•People have problems understanding the problem and their role in its solution (turf). 吾人在其問題的理解與對策執行的角色扮演上會有問題 It is difficult to define the scope of the problem. 難以定義問題範圍 •It is difficult to get to root causes. 難以找到真正的要因 •People get blamed when the process is the problem (80/20 Rule). 吾人在當製程是真正問題時 , 會遭到責備 •Process management is ineffective 製程管理沒有效果 •Improvement is slowed 改善緩慢
6
•
Without Understanding VariatiWithout Understanding Variationon 若不了解其變異若不了解其變異
•Management by the last data point 永遠是用最後的資料作管理 ( 永遠在頭痛醫頭 , 腳痛一腳 , 沒有源頭置根本的觀念 ) •There’s lots of fire fighting 火災不斷
•Using special cause methods to solve common cause problems 用特別的方法處理共同要因的 ( 一般性 ) 問題
•Tampering and micromanaging abound •修改與小事的管理老是存在 •Goals and methods to attain them fail •目標與方法無法達成 •Understanding the process is handicapped •只知道製程是個問題
• Learning is slowed 學習慢 •Process management is ineffective 製程管理沒有效果 •Improvement is slowed 改善慢
7
•
Without DataWithout Data若是手上沒有資料若是手上沒有資料
•Everyone is an expert: 每個人都是專家 •Discussions produce more heat than light 討論不斷
•Historical memory is poor 歷史的記憶模糊 •Difficult to get agreement on: 難以得到協議若
•What the problem is 無法得知問題是什麼 •What success looks like 無法得知其成果將如何 •Progress made 或由哪一製程所產出
•Process management is ineffective 製程管理是無效的 •Improvement is slowed 改善慢
8
“Early on, we failed to focus adequately on core work processes and statistics.” 初期若核心工作製程與統計無法適當集中 , 其結果…
David Kearns and David Nelder, Xerox Corporation
Without Statistical ThinkingWithout Statistical Thinking若無製程統計的思維若無製程統計的思維
•Your management and improvement processes are handicappe 吾人的管理與改善將有障礙 •It’s like 其像
–Football without a passing attack 足球未經核准即攻擊 –Growing a lawn without fertilizer 草地未經施肥 –Doing research without measurements 研究未做量測資料 –Playing golf without your irons 不用自己的球竿打高爾書球
9
SECURE
STORE
KIT
Load Program Load Pick/Place Load Reflow Profile Load Stencil
Screen Solder Paste
Parts SMT
Placement
I / R ReFlow
Clean
PEM Parts (ASIC, ADC, DAC) Placement
& Hand Solder
Clean
Second Level Assy. Touch-up solder joints
Mechanical Installations
Staking/Bonding
Clean
Electrical Functional
Test
Clean Bake Conformal Coat
Post Test InspectionAcceptanc
e Test
Electrical
Controlled
Storage
Inspection Checkpoint
Inspection Checkpoint
Inspection Checkpoint
Inspection Checkpoint
Inspection Checkpoint
Inspection Checkpoint
Inspection Checkpoint
Through-hole and Plastic Parts Preparation Tin Components
Form & Cut Axial Leads
Through-hole Component Placement
& Hand Solder
Clean & Inspection Checkpoint
PWB Preparation: Clean
Ink Stamp Bake
Production Operation
Inspection Operation
Test Operation
Material Control Operation
KEY
Manufacturing Flow Diagram of PWB AssemManufacturing Flow Diagram of PWB Assembly PWB bly PWB 組裝之製造流程圖組裝之製造流程圖
10
SECURE
STORE
KIT
Load Program Load Pick/Place Load Reflow Profile Load Stencil
Screen Solder Paste
Parts SMT
Placement
I / R ReFlow
Clean
PEM Parts (ASIC, ADC, DAC) Placement
& Hand Solder
Clean
Second Level Assy. Touch-up solder joints
Mechanical Installations
Staking/Bonding
Clean
Electrical Functional
Test
Clean Bake Conformal Coat
Post Test InspectionAcceptanc
e Test
Electrical
Controlled
Storage
Inspection Checkpoint
Inspection Checkpoint
Inspection Checkpoint
Inspection Checkpoint
Inspection Checkpoint
Inspection Checkpoint
Inspection Checkpoint
Through-hole and Plastic Parts Preparation Tin Components
Form & Cut Axial Leads
Through-hole Component Placement
& Hand Solder
Clean & Inspection Checkpoint
PWB Preparation: Clean
Ink Stamp Bake
Production Operation
Inspection Operation
Test Operation
Material Control Operation
KEY
Manufacturing Flow Diagram of PWB AsseManufacturing Flow Diagram of PWB Assembly PWB mbly PWB 組裝之製造流程圖組裝之製造流程圖
11
Depends on levels of activity and job responsibility. Depends on levels of activity and job responsibility. 依據活動的層級與工作執掌依據活動的層級與工作執掌
Where we're headed Where we're headed 我們朝何方我們朝何方
Managerial processes Managerial processes to guide us to guide us 用管理的程序來指導我們用管理的程序來指導我們
Where the work gets Where the work gets Done Done 讓所需的工作被執行完成 讓所需的工作被執行完成
Strategic Strategic 策略上的策略上的
Managerial Managerial 管理上的管理上的
Operational Operational 作業性的作業性的
Executives Executives 高階決策層高階決策層
Managers Managers 經理階層經理階層
Workers Workers 現場員工現場員工
Use of Statistical ThinkingUse of Statistical Thinking運用統計思維運用統計思維
12
• Executives use systems approach. • 決策者運用系統導向策略 • Core processes have been flow charted • 主要程序已被流程圖表化 • Strategic direction defined and deployed. • 策略方向的訂定與展開 • Measurement systems in place. • 適當的量測系統 • Employee, customer, and benchmarking studies are used to
drive improvement. • 是以員工 , 客戶與 benchmarking 的研究被用來主導改
善 • Experimentation is encouraged. 鼓勵實驗 •
Statistical Thinking at the Strategic LevStatistical Thinking at the Strategic Levelel
決策者之統計思維決策者之統計思維
13
.
• Managers use meeting management techniques • 經理利用會議管理技巧 • Standardized project management systems are in place. • 適當的標準化專案管理系統 • Both project process and results are reviewed. • 此專案的流程與結果已被審核 • Process variation is considered when setting goals. • 當設定目標時 , 流程的變異已被考慮 • Measurement is viewed as a process. • 量測點被視為一個流程 • The number of suppliers is reduced • 供應者數目減少 • A variety of communication media are used. • 廣泛的傳訊媒體被採用
Statistical Thinking at the Statistical Thinking at the Managerial Level Managerial Level
經理階層統計思維經理階層統計思維
14
• Work processes are flowcharted & documented • 工作程序已被流程圖表化與書面化 • Key measurements are identified. 主要量測點已被確認
– Time plots displayed 時間的圖示被展現 • Process management and improvement utilize: • 製程管理與改善採用
– Knowledge of variation, and 變異觀念的知識及 – Data analysis 資料分析
• Improvement activities focus on the process, not blaming employees. 改善工具著重於製程 ,而非責備員工
•
Statistical Thinking at the Operational LevelStatistical Thinking at the Operational Level
現場員工的統計思維範例現場員工的統計思維範例
15
Statistical Thinking at the Operational LStatistical Thinking at the Operational Levelevel
現場員工的統計思維範例現場員工的統計思維範例A Recent Experience 最近的經驗 Huge quantities of data 大量的資料 Limited understanding of structure 在有限度理解的結構上 Consultants applied artificial neural nets 顧問群運用人工神經網狀系統 Didn’t work但不成功
16
Statistical Thinking at the Operational LeStatistical Thinking at the Operational Levelvel
現場員工的統計思維範例現場員工的統計思維範例A Recent Experience 最近的經驗 Artificial Neural Nets apply nicely in many situations (NIST Examples): 人工神經網狀系統出色地運用於許多領域 :
Optical Character Recognition光學文字辨識系統 Finger Printing 指紋辨識 Face Printing for the FBI 相貌辨識
Example等案例上
17
…….But, .But, 但但•Unless you sample the process taking the right amount of the right kind of data (rational subgroups) you will never approach process understanding. •在抽驗的流 ( 製 ) 程裡若你無法取得正確的數量與資料 (合理的樣組 ),你將無法深入了解此一流 ( 製 ) 程 •Without process understanding, there is no process control. •流 ( 製 ) 程若不了解 ,就無所謂的流 ( 製 )程管制
18
Key Learnings from Key Learnings from Statistical Thinking EffortsStatistical Thinking Efforts
由統計思維的努力中由統計思維的努力中 ,, 吾人學到的吾人學到的要點要點Statisticians don’t understand Statistical Thinking as
well as they think they do. 統計的思維不僅要懂而且也要會做 Those who do understand it have limited access to managerial and strategic levels. 真正了解統計思維的人 , 在管理與決策上之能力較少受限制 There’s much more work to be done. 較多的事能被完成
Spread the word 口令的展開 Focus on process 著重製程
19
Characteristics for which you focus on defects 其特性著重於缺點 Classify products as either ‘good’ or ‘bad’, or count # defects
以產品的好 .壞 ,缺點數量來看e.g., radio works or not
如收音機是否可以播放
Categorical or discrete random variables屬不連續的雖機變數
AttributesAttributes 計數值計數值 VariablesVariables 計量值計量值
Quality CharacteristicsQuality Characteristics 品質特性品質特性
Characteristics that you measure, e.g., weight, length
其特性可被量測而得 , 如重量,長度等 May be in whole or in fractional numbers
可以以整數或分數表達
Continuous random variables
連續的隨機變數
20
Types Of DataTypes Of Data資料型態資料型態
Attribute data 計數資料 Product characteristic evaluated with a discrete choice 產品資料特性以離散的評估方式選定
Good/bad, yes/no 良品 / 不良品 , 好 /壞 Variable data 計量資料
Product characteristic that can be measured 產品特性能被量測而得
Length, size, weight, height, time, velocity 長度 ,大小 ,重量 , 高度 , 時間 ,,速度
21
Types of VariationsTypes of Variations 變異型態變異型態
Common Cause 共同原因Random隨機
Chronic長期的Small影響小
System problems 系統問題
Mgt controllable 管理上的控制Process improvement 製程改善
Process capability 製程能力
Special Cause 特殊原因Situational局部Sporadic偶而發生Large影響大Local problems局部問題
Locally controllable 可局部控制
Process control 製程管制 Process stability 製程的穩定性
22
Statistical technique used to ensure process is making product to standard 統計技術用於確保製程所製出的產品合乎標準 All process are subject to variability 所有製程受變異性所支配Natural or Common causes 自然或共同原因 :
Random variations隨機變異如設備損耗Assignable causes 特殊原因
Correctable problems 可改善的問題Machine wear, unskilled workers, poor material
如生手 ,材料不良…Objective: Identify assignable causes
目標 :確認特殊原因Uses process control charts
利用管制圖表
Statistical Process ControlStatistical Process Control統計製程管制統計製程管制
23
Causes of VariationCauses of Variation變異的原因變異的原因
Inherent to process固有製程 Random隨機 Cannot be controlled 不可控 Cannot be prevented 無法預防 Examples 如 :
Weather氣候 accuracy of measurements
量測精度 capability of machine
設備能力
Exogenous to process外來因子影響製程
Not random非隨機 Controllable 可控 Preventable 可預防 Examples 如
tool wear 工具磨耗 “Monday” effect週一效應 poor maintenance 維護差
Common Causes 共同原因 Assignable Causes 特殊原因
What prevents perfection? Process variation... 何事阻礙完美 ? 製程變異…
24
Product Specification and Process VariationProduct Specification and Process Variation產品規格與製程變異產品規格與製程變異
Product specification 產品規格desired range of product attribute 產品屬性之期望範圍part of product design 產品設計的一部份
length, weight, thickness, color, …長度 ,重量 ,厚度 ,顏色…等nominal specification(公稱規格 )
upper and lower specification limits(規格上下限 )
Process variability 製程變異inherent variation in processes 製程中固有的變異
limits what can actually be achieved 其實際能被達成之界限值defines and limits process capability 定義並限制製程能力Process may not be capable of meeting specification!
製程是有可能無法達到規格的要求 !
25
Grams(a) Location
Average(平均值 )
Common CausesCommon Causes共同原因共同原因
26 (a) Location
Grams
Average
Assignable CausesAssignable Causes特殊原因特殊原因
27
-3 -2 -1 +1 +2 +3Mean 平均值
68.26% 95.44% 99.74%
= Standard deviation = 標準差
The NormaThe Normal l DistributionDistribution常態分配常態分配
28
X
Mean 平均值
Central Limit Theorem
x
x
n
xx
nμX =μX =
Standard deviation
樣本標準差
μX =μX =
Theoretical Basis of Control Theoretical Basis of Control ChartsCharts
29
UCL 管制規格上限
Nominal 中心線
LCL 管制規格下限
1 2 3 Samples
Control ChartsControl Charts 管制圖管制圖
30
1 2 3 Samples
Control ChartsControl Charts 管制圖管制圖UCL 管制規格上限
Nominal 中心線
LCL 管制規格下限
31
Assignable causes likely 可能的特殊原因
1 2 3 Samples
Control ChartsControl Charts 管制圖管制圖UCL 管制規格上限
Nominal 中心線
LCL 管制規格下限
32
Process Control: Process Control: Three Types of Process OutputsThree Types of Process Outputs
製程管制的三種顯示型態製程管制的三種顯示型態
Frequency
Lower control limit
Size Weight, length, speed, etc.
Upper control limit
(b) In statistical control, but not capable of producing within control limits. A process in control (only natural causes of variation are present) but not capable of producing within the specified control limits; 共同原因變異 and
(c) Out of control. A process out of control having assignable causes of variation. 特殊原因變異
(a) In statistical control and capable of producing within control limits. A process with only natural causes of variation and capable of producing within the specified control limits. 正常型
33
The Relationship Between The Relationship Between Population and Sampling DistributionsPopulation and Sampling Distributions群體與樣本間之關係群體與樣本間之關係
Uniform
Normal
BetaDistribution of sample means 樣本平均值分配
x means sample of Mean
n
xx
Standard deviation of
the sample means
(mean)
x2 withinfall x all of 95.5%
x3 withinfall x all of 99.7%
x3 x2 x x x1 x2 x3
Three population distributions群體分配
34
Visualizing Chance CausesVisualizing Chance Causes機遇原因之觀察機遇原因之觀察
Target
At a fixed point in time 固定時間
Time Targe
t
Over time 連續時間
Think of a manufacturing process producing distinct parts with measurable characteristics. These measurements vary because of materials, machines, operators, etc. These sources make up chance causes of variation. 製造各零件之量測特性會因 4M等機遇原因而發生變異
35
Process Control ChartsProcess Control Charts製程管制圖製程管制圖
Plot of Sample Data Over Time
0
20
40
60
80
1 5 9 13 17 21
Time
Sam
ple
Valu
e
SampleValueUCL
Average
LCL
36
Control Charts
Variables Charts
Attributes Charts
Continuous
連續的
Numerical Data
Categorical or Discrete
離散的
Numerical Data
Control Chart TypesControl Chart Types管制圖型態管制圖型態
計量 計數
37
Control Chart SelectionControl Chart Selection管制圖的選定管制圖的選定Quality CharacteristicQuality Characteristic
variablevariable attributeattribute
n>1?n>1?
n>=10 or n>=10 or computer?computer?
x and MRx and MRnono
yesyes
x and sx and s
x and Rx and Rnono
yesyes
defectivedefective defectdefect
constant constant sample sample size?size?
p-chart with p-chart with variable sample variable sample sizesize
nono
p or p or npnp
yesyes constant constant sampling sampling unit?unit?
c uc u
yesyes nono
38
Produce Good Provide Service
Stop Process
Yes
No
Assign. Causes? Take Sample
Inspect Sample
Find Out Why Create
Control Chart
Start
Statistical Process Control SteStatistical Process Control Stepsps
統計製程管制控制步驟統計製程管制控制步驟
39
Statistical Thinking is a philosophy of lStatistical Thinking is a philosophy of learning and Action based on the followearning and Action based on the following fundamental principles:ing fundamental principles:統計思維哲學之學習與行動基於以下原則統計思維哲學之學習與行動基於以下原則
•All work occurs in a system of interconnected processes, •Variation exists in all processes, and •Understanding and reducing variation are keys to success. •所有工作的產生源於系統互相連結之製程 ,而變異存在於所有製程 , 了解並降低製程的變異是成功的關鍵
40
Using Control ChartsUsing Control Charts如何使用管制圖如何使用管制圖
1) Select the process to be charted 選擇需要被圖表化之製程
2) Get 20 - 25 groups of samples 選擇樣組及樣本大小 (usually 5-20 per group for X and R-chart or n≥50 for p-chart)
3) Construct the Control Chart建立管制圖 4) Analyze the data relative to the control limits. Points outs
ide of the limits should be explained 分析關聯於管制界線之資料 , 點超出界限需能被解釋
5) Once they are explained, eliminate them from the data and recalculate the control chart 一旦澄清 ,消除異常點及原因 ,並重算管制圖資料
6) Use the chart for new data, but DO NOT recalculate the control limits 利用此新資料 ,但無須重算管制界限
41
Type of variables control chart 計量管制圖 Interval or ratio scaled numerical data
間距或比率量測數字資料
Shows sample means over time
算出樣本平均值
Monitors process average
監控製程平均數Example: Measure 5 samples of solder paste & compute means of samples; Plot
如計算錫膏厚度之平均值 ,再點圖
XX Chart Chart 平均值管制圖平均值管制圖
42
Basic Probabilities Concerning the DistriBasic Probabilities Concerning the Distribution of Sample Meansbution of Sample Means
有關樣本平均數之機率分佈 有關樣本平均數之機率分佈
{ }{ } 0006.0=4<4+>
0027.0=3<3+>
xx
xx
σμxorσμx
σμxorσμx
-Prob
-Prob
nσσ x /=Std. dev. of the sample means 樣本平均數標準差 :
43
Estimation of Mean and Std. Dev. Estimation of Mean and Std. Dev. of the Underlying Processof the Underlying Process
在製程控制之下之平均值與標準差估計 在製程控制之下之平均值與標準差估計 use historical data taken from the process when it was “known” to be i
n control 當製程穩定時 , 利用過去所產生之歷史資料 usually data is in the form of samples (preferably with fixed sample siz
e) taken at regular intervals 樣本資料是在一定間隔的時間裡取得 process mean estimated as the average of the sample means (the gr
and mean or nominal value)假設製程平均值與與與與與與與與 process standard deviation estimated by: 製程標準差與與與
standard deviation of all individual samples 所有個別值樣本之標準差
OR mean of sample range R/d2, where 或樣本平均值 / d2
sample range R = (Rmax-Rmin), d2 = value from look-up table, 全距為 R, d2 可由查表得知 ,
n
R R
i
n
1i ==∑
44
X-bar vs. R chartsX-bar vs. R charts平均值平均值 VSVS 全距管制圖 全距管制圖
R charts monitor variability: Is the variability of the process stable over time? Do the items come from one distribution? R 管制圖監控變異性 , 是否整個製程處於安定狀態 ? 有項目超出此一分配嗎 ? X-bar charts monitor centering (once the R chart is in control): Is the mean stable over time? X-Bar 管制圖監控中心 ( 一旦 R 管制圖處於管制狀態 ):平均值於爭個製程是否穩定 ?
>> Bring the R-chart under control, then look at the x-bar chart(先看 R 圖 ,再看 Xbar 圖 )
45
How to Construct a Control ChaHow to Construct a Control Chartrt
如何建立管制圖 如何建立管制圖 1. Take samples and measure them. 取樣量測 2. For each subgroup, calculate the sample average and ra
nge. 每個群組 , 計算平均值與全距 3. Set trial center line and control limits. 製作解析用管制
圖之中心線與管制界限 4. Plot the R chart. Remove out-of-control points and revis
e control limits.畫 R 圖 ,移除異常點 ,再修正管制界限
5. Plot x-bar chart. Remove out-of-control points and revise control limits.畫 R 圖 ,移除異常點 ,再修正管制界限
6. Implement - sample and plot points at standard intervals. Monitor the chart. 管制用管制圖 ,於標準間隔時間取樣 ,監控此管制圖
46
Type 1 and Type 2 ErrorType 1 and Type 2 Error第一種與第二種錯誤 第一種與第二種錯誤
Type 1 ( alpha]
Error
No Error
No Error
Type 2 (Beta)
Error
Alarm No Alarm
In-Control 管制內
Out-of-Control 失控
47
Common Tests to Determine if the Common Tests to Determine if the Process is Out of ControlProcess is Out of Control
管制圖異常之判定 管制圖異常之判定 One point outside of either control limit 一點超出管制界線 2 out of 3 points beyond UCL - 2 sigma 3 點有 2 點在 2 個標準差或以外 7 successive points on same side of the central line 連續 7 點在中心線之同一側 of 11 successive points, at least 10 on the same side
of the central line 連續 11 點有 10 點在中心線之同一側 of 20 successive points, at least 16 on the same side
of the central line 連續 20 點有 16 點在中心線之同一側
48
Type 1 Errors for these Tests Type 1 Errors for these Tests 第一種錯誤 第一種錯誤
Test Probability Type 1 Error
2/3
7/7
10/11
16/20
1/1 2(0.00135) 0.0027
0.0052
(0.5)7 0.0078
11)5.0(11
11)5.0(10)5.0(
10
11
0.00586
iii
20)5.0()5.0(
20
16
20
0.0059
3)0228.0()9772.0(2)0228.0(2
3
49
Type 2 ErrorType 2 Error第二種錯誤 第二種錯誤 Suppose 1 >
Type 2 Error =
1x3xProb
/3
/3
1
1
n
xx
where (z) denotes the the cumulative probability of a standard normal variate at z • Power = 1- Type 2 Error. Power increases as … n increases, as () increases, and as decreases. • Extension to is straightforward
50
X Chart Control LimitsX Chart Control Limits
Sample Range at Time i
# Samples
Sample Mean at Time i
From Table
RAxx
LCL
RAxx
UCL
2=
2+=
n
R R
i
n
1i ==∑
n
x x
i
n
i 1==∑
51
Factors for Computing Factors for Computing Control Chart LimitsControl Chart Limits
管制圖之係數表管制圖之係數表Sample Size, n
Mean Factor, A2
Upper Range, D4
Lower Range, D3
2 1.880 3.268 0
3 1.023 2.574 0
4 0.729 2.282 0
5 0.577 2.115 0
6 0.483 2.004 0
7 0.419 1.924 0.076
8 0.373 1.864 0.136
9 0.337 1.816 0.184
10 0.308 1.777 0.223 0.184
Table
52
Type of variables control chart 計量管制圖 Interval or ratio scaled numerical data 間距或比率量測數字資料
Shows sample ranges over time Difference between smallest & largest values in inspection sample 樣本中最大值與最小值之差
Monitors variability in process間控製程變異性
Example: Calculate Range of samples of solder paste; Plot 計算全距並點圖
RR Chart Chart 全距管制圖全距管制圖
53
Sample Range at Time i 某時間間隔之全距
Samples size
樣本大小
From Table查表
R Chart Control LimitsR Chart Control LimitsRR 管制圖管制界限公式管制圖管制界限公式
n
R R
RD LCL
RD UCL
i
n
1i
3R
4R
==
=
=
∑
Setting up a X-BAR R ChartSetting up a X-BAR R Chart建立建立 X-bar R X-bar R 管制圖管制圖
Take about 20-25 sample groups (n) of the process result. Each sample should contain 4 or 5 observations.
For each sample calculate the average and the range.
Average all the sample averages = X-BAR.
Average all the sample ranges = R-BAR.
Calculate the upper & lower control limit for X-BAR
Calculate the upper & lower control limit for R-BAR
RAxx
LCL
RAxx
UCL
2
2
+=
+=
n
x x
i
n
i 1==∑
RD LCL
RD UCL
3R
4R
=
=
n
R R
i
n
1i==∑
Using an s-Chart Using an s-Chart Instead of an R-ChartInstead of an R-Chart
利用標準差圖取代利用標準差圖取代 RR 管制圖管制圖S-Charts are used when: Tight control of process variation is essential.Sample size equals 10 or more.
a computer can be used to simplify & speed up calculations.
Formulas:
i
i
n
1i
n
)x(xΣs
-==
sBLCL
sBUCL
s
s
3
4
=
=
Control Limits for s-Chart Control Limits for X-bar Chart
sAxLCL
sAxUCL
x
x
3
3
=
+=
56
Example: The first 20 days samples are as follows:SampleNumber X-Bar R
1 45.87 43 44.55 39.23 38.34 47.68 43.11 9.3332 57.1 37.23 51.39 54.83 50.02 50.16 50.12 19.863 29.93 61.78 39.65 32.54 68.48 66.42 49.8 38.554 57.01 59 54 54.48 52 44.57 53.51 14.435 68.38 35.84 55.12 35.64 50.75 55.92 50.28 32.746 69.76 64.21 62.43 67.34 54.43 57.34 62.59 15.337 57.01 27.27 44.8 46.41 67.61 32.5 45.93 40.348 46.65 50.41 54.02 44.3 63.19 45.59 50.69 18.899 33.97 45.59 35.07 55.95 51.41 58.94 46.82 24.9610 54.56 64.9 54.53 57.34 53.34 59.43 57.35 11.5611 45.27 47.11 61.7 46.84 50.79 53.42 50.85 16.4412 41.76 34.34 51.04 45.5 44.98 65.77 47.23 31.4313 31.9 44.12 41.27 59.17 49.03 65.48 48.5 33.5814 45.32 33.91 39.43 37.34 41.43 38.34 39.3 11.4215 60.91 46.36 54.27 26.29 43.53 47.51 46.48 34.6216 64.49 67.11 58.33 62.23 59.23 58.34 61.62 8.77917 48.53 48.56 41.29 44.92 47.8 28.07 43.2 20.4918 54.44 58.49 58.54 63.34 65.43 69.23 61.58 14.819 50.65 44.97 55.32 60.92 35.37 43.18 48.4 25.5420 51.89 50.26 40.28 54.33 61.67 47.83 51.04 21.39
X double-bar R-bar50.42 22.2
57
UCL
LCL
X-bar ChartX-bar Chart
Is the process in control? Are the specifications being met? How can we tell if the variability is in control?
30
35
40
45
50
55
60
65
0 5 10 15 20 25
58
R-ChartR-Chart The R chart measures the change in the spread over time. Plot R, the range for each sample. Lower Control Limit = Upper Control Limit =
RD3RD4
UCL
LCL0
5
10
15
20
25
30
35
40
45
50
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
59
Ex: Control “Commuting times” Ex: Control “Commuting times” Step 1Step 1 Commuting Times (min.) - A.M.Commuting Times (min.) - A.M.
WEEKWEEK1 2 3 4 5 6 7 8 9 10
55 90 100 70 55 75 120 65 70 100
75 95 75 110 65 85 110 65 85 80
65 60 75 65 95 65 65 90 60 65
80 60 65 60 70 65 85 90 65 60
80 55 65 60 70 65 70 60 75 80
71 72 76 73 71 71 90 74 71 77
25 40 35 50 40 20 55 30 25 40
Min
ute
sM
inu
tes
Xbar =Xbar =R =R =
Step Step 22
Step Step 33
X = 74.6 X = 74.6
R = 36 R = 36
n = 5n = 5
UCLUCLLL = X + A = X + A22*R *R
= 74.6 + (.58)*(36) = 74.6 + (.58)*(36)
= 95.48 = 95.48
LCLLCLLL= X - A= X - A22*R *R
= 74.6 - 20.88 = 74.6 - 20.88
= 53.72= 53.72
UCLUCLRR = D = D44*R *R
= (2.11)*(36.0) = (2.11)*(36.0)
= 75.96 = 75.96
LCLLCLRR = D = D33*R *R
= 0= 0
60
Control “Commuting times” Control “Commuting times” (cont.)(cont.)step 4step 4 Commuting times - A.M.Commuting times - A.M.
UCL = 95.48UCL = 95.48
Xbarbar = 74.6Xbarbar = 74.6
LCL = 53.72LCL = 53.72
Xbar ChartXbar Chart
11 101022 33 44 55 66 77 88 995050
100100
7575
R ChartR Chart UCL = 75.96UCL = 75.96
Rbar = 36.0Rbar = 36.0
LCL = 0LCL = 0
11 101022 33 44 55 66 77 88 99
7575
55
3535
61
FigureFigure
62
Type of attributes control chart 計數管制圖
Nominally scaled categorical data 以絕對資料分類e.g., good-bad 如好 ,壞
Shows % of nonconforming items顯示不合格項目 %
Example: Count # defective chairs & divide by total chairs inspected; Plot 計算椅子的不良數除以椅子總檢驗數 , 點圖
Chair is either defective or not defective椅子只有好 與壞兩種
p Chartp Chart不良率管制圖不良率管制圖
Setting up a p ChartSetting up a p Chart建立建立 pp 管制圖管制圖•Take about 20-25 samples of the process result. Each
•sample should be large enough to contain AT LEAST 1 •bad observation. Often for P-Charts samples sizes are •in excess of 100. •For each sample calculate the percentage of bad units.
•Average all the sample percentages together, this is P-BAR.
• Calculate the upper & lower control limit for the P-BAR chart • using the following formulas:
inpp
p)1(*
*3
64
p Chart Control Limitsp Chart Control Limits不良率管制圖管制界限不良率管制圖管制界限
# Defective Items in Sample iSize of sample i
•If individual samples are within 25% of the average sample size then control limits can be calculated using the average sample size:
•z = 2 for 95.5% limits; •z = 3 for 99.7% limits
i
k
1i
i
k
1ii
k
i
p
p
n
xp and
k
nn
n
ppzpLCL
n
ppzpUCL
=
=1=
∑
∑=
∑=
)1(=
)1(+=
•If sample sizes vary by more than 25% of the average sample size then control limits should be computed for each sample.
ipp n
pppLCLUCL
)1(**3±=,
65
Example: p-ChartExample: p-ChartM&M Mars wants to institute a statistical process control on a new candy bar. In order to do so, every shift they sample 50 bars and determine the number of defective ones. They obtain the following data:
Shift # Rejected Shift # Rejected1 7 11 12 11 12 183 5 13 94 18 14 85 6 15 146 8 16 117 22 17 88 9 18 69 2 19 310 1 20 3
66
20 groups of 50 = 1000 samples Total defective = 170 p-bar = 0.17 UCL = 0.17 + 3 x 0.053 = 0.329 LCL = 0.17 - 3 x 0.053 = 0.010 Plotting the % defective shows:
.17.83 50 0.053
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
1 3 5 7 9 11
13
15
17
19
67
Identifying Special CausesIdentifying Special Causes確認特殊要因確認特殊要因
It appears that shifts 4, 7 and 12 were out of control.
Upon further inspection it appears that too much water was added to the process in shifts 4 and 7 and that in shift 12 a new operator started.
Since each of the out of control points have assignable causes, we eliminate them from the data.
The new control chart is then:
68
Now it appears that shift 15 is out-of-control. Further checking shows that the temperature was set
too high during this shift. Therefore, we want to eliminate this point so that in
subsequent tests we can identify when this occurs. If we eliminate this point the new control chart is:
0
0.05
0.1
0.15
0.2
0.25
0.3
1 2 3 5 6 8 9 10 11 13 14 15 16 17 18 19 20
Identifying Special Identifying Special CausesCauses
69
Final p ChartFinal p Chart
UCL = 0.122 + 3 x 0.046 = 0.260 LCL = 0.122 - 3 x 0.046 = -0.016 = 0.0
(negative control limits should be set to 0) Now they should use this chart for all subsequent
sampling until the process changes
0
0.05
0.1
0.15
0.2
0.25
0.3
1 2 3 5 6 8 9 10 11 13 14 16 17 18 19 20
70
Determining if Your Process isDetermining if Your Process is “Out of Control” “Out of Control”
決定你的製程是否在穩定狀態決定你的製程是否在穩定狀態 Establish regions A, B, and C as one, two, and three One or more points fall outside the control limits. 2 out of 3 consecutive points fall in the same region A 4 out of 5 consecutive points fall in the same region A or B 6 consecutive points increasing or decreasing 9 consecutive points on the same side of the average. 14 consecutive points alternating up and down 15 consecutive points within region C.
A
B
C
AB
C
71
Using an np ChartUsing an np Chart建立不良數管制圖建立不良數管制圖
Np charts for number of nonconforming units. 以不合格品之數統計
Converted from basic p-chart 由 p 管制圖演變而來Multiply p by sample size (n). 不良率乘以樣本大小
Formulas:
)1(*3
)1(*3
ppnpnLCL
ppnpnUCL
p
p
Setting up a c chartSetting up a c chart建立缺點數管制圖建立缺點數管制圖
•Take about 20-25 samples from the process. •Each sample contains 1 unit.
•For each unit count the number of occurrences for the •observation of interest.
•Calculate the average number of occurrences per unit. •This is C-BAR.
• Calculate the upper & lower control limit for the C-BAR chart • using the following formulas:
c3cLCL,UCL pp ±=
73
Using an u ChartUsing an u Chart建立單位缺點 犐赯建立單位缺點 犐赯 ?/span>?/span>
• A u chart is used when the unit size inspected for defects • is not constant. In these cases the unit is often referred to • as an area of opportunity (ni). • The average occurrence per area of opportunity • (i.e. the center line) is calculated as: • The same 25% variation rule discussed for p-charts • applies here as well. Control limits are calculated as:
i
iu
nuuLCLu
nuuUCL
/3±=
/3±=
•
•
74
FigureFigure
75 425 Grams
Mean 平均值
Process Distribution 製程分配
Distribution of sample means 樣本平均值分配
Sample Means and theSample Means and theProcess DistributionProcess Distribution
樣本平均值與製程分配樣本平均值與製程分配
76
Process CapabilityProcess Capability 製程能力 製程能力
µ , Nominal value
800 1000 1200 Hours
Upper specification
Lower specification
Process distribution
(a) Process is capable
77
Process CapabilityProcess Capability 製程能製程能力 力
Lower specification
Mean
Upper specification
Two sigma
µ, Nominal value
78
Process CapabilityProcess Capability 製程能製程能力 力
Lower specification
Mean
Upper specification
Four sigma
Two sigma
µ , Nominal value
79
Process CapabilityProcess Capability 製程能製程能力 力
Lower specification
Mean
Upper specification
Six sigma
Four sigma
Two sigma
µ , Nominal value
80
Process CapabilityProcess Capability 製程能力製程能力
Capable
Very capable
Not capable
LSL USLSpecProcess variation
81
Process Capability CProcess Capability Cpkpk
製程能力指數製程能力指數Assumes that the process is:
•under control •normally distributed
•假設製程為穩定且為常態分配
LSL)-(USLoleranceTT ,T
µ(XCa
σ6LSL)-(USL
σ6δ2
σ 6
Tolerance
capabilityProcess Tolerance
Cp
==2/
)=
====
-
Cpk =min(Cpu , Cpl)
Cpu=(USL-µ)/3
Cpl =(µ-LSL)/3
Precision精密度
Capability 準確度
82
Meanings of CMeanings of Cpkpk Measures Measures
CCpkpk 量測之意義量測之意義Cpk = negative number
Cpk = zero
Cpk = between 0 and 1
Cpk = 1
Cpk > 1
83
Statistical Process Control – Statistical Process Control – Identify and Reduce Process VariabilityIdentify and Reduce Process Variability統計製程管制統計製程管制 -- 確認並降低製程變異確認並降低製程變異
Lower specification
limit
Upper specification
limit
(a) Acceptance sampling
(b) Statistical process control
(c) cpk >1
84
Quality Control ApproachesQuality Control Approaches品質管制方法品質管制方法
Statistical process control (SPC) 統計製程管制
Monitors production process to prevent poor quality
監控產品製程以預防不良品質 Acceptance sampling 允收抽樣
Inspects random sample of product or materials to determine if a lot is acceptable隨機抽樣檢驗產品或物料以決定此批是否允收
85
Sampling vs. ScreeningSampling vs. Screening抽樣與篩選抽樣與篩選
Sampling 抽樣 When you inspect a subset of the population 群體批中檢查小批
Screening When you inspect the whole population 群體批中檢查全數
The costs consideration 成本的考量 , 經濟的原則
86
Acceptance SamplingAcceptance Sampling允收抽樣允收抽樣
Accept/reject entire lot based on sample results
整個允收 /拒收是樣品結果為基礎
Not consistent with TQM of Zero Defects
與 TQM 的零缺點不同
Measures quality in percent defective
以缺點百分率測量品質
87
Sampling PlanSampling Plan抽樣計劃抽樣計劃
Guidelines for accepting lot 允收批之指導作業
Single sampling plan單一抽樣計劃N = lot size批量n = sample size (random) 樣本大小c = acceptance number 允收數d = number of defective items in sample 樣本不良項目之數目
If d <= c, accept lot; else reject
若 d <= c, 允收此批 , 其他則批退
88
Producer’s & Consumer’s RiskProducer’s & Consumer’s Risk生產者與消費者冒險率生產者與消費者冒險率
TYPE I ERROR = P(reject good lot) or producer’s risk, too nervous
5% is common 第一種錯誤 = 將好批判成壞批的機率 ,緊張忙亂的錯誤
TYPE II ERROR = P(accept bad lot)
or consumer’s risk, absent- minded
10% is typical value 第二種錯誤 = 將壞批判成好批的機率 , 心不在焉的錯誤
89
Quality DefinitionsQuality Definitions品質的定義品質的定義
Acceptance quality level (AQL)
允收水準Acceptable fraction defective in a lot
允許一批中不良的比例
Lot tolerance percent defective (LTPD)
拒收水準 ,批容許不良率
Maximum fraction defective accepted in a lot
允許一批中最大不良的比例
90
Operating CharacteristicOperating Characteristic (OC) Curve (OC) Curve
作業特性曲線作業特性曲線Shows probability of lot acceptance
顯示批允收的機率Based on 是基於 :
sampling plan 抽樣計劃 quality level of lot批品質的等級
Indicates discriminating power of plan
顯示不同計劃的差異性
91
Operating Characteristic Curve Operating Characteristic Curve OCOC曲線曲線
AQL LTPD
= 0.10
= 0.05
Pro
babi
lity
of a
cce
pta
nce,
Pa
{
0.60
0.40
0.20
0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20
0.80
{
Proportion defective 不良比例
1.00
OC curve for n and c 樣本大小與 c 允收數
允
收
機
率
92
Average Outgoing Quality (AOQ)Average Outgoing Quality (AOQ)平均出廠品質平均出廠品質
Expected number of defective items passed to customer
期望通過客戶之不良項目數Average outgoing quality limit (AOQL) is maximum point on AOQ curve
平均出廠品質界限是 AOQ曲線的最大值
93
AOQ CurveAOQ Curve平均出廠品質曲線平均出廠品質曲線
0.015
0.010
0.005
0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10
AOQL
Average Outgoing Quality
(Incoming) Percent Defective
AQL LTPD
94
Double Sampling PlansDouble Sampling Plans雙次抽樣計劃雙次抽樣計劃
Take small initial sample 抽取少量之原始樣本If # defective < lower limit, accept
If # defective > upper limit, reject
If # defective between limits, take second sample
若不良數 < 下界限 , 允收
若不良數 > 上界限 ,拒收
若不良數界於界限內 ,第二次抽樣
Accept or reject based on 2 samples 允收與拒收是站在此二抽樣樣本上
Less costly than single-sampling plans
比單次抽樣成本低
95
Multiple (Sequential) Sampling PlansMultiple (Sequential) Sampling Plans 多重多重 ((連連續續 )) 抽樣計劃抽樣計劃
Uses smaller sample sizes使用較小的樣本大小
Take initial sample 取出原始樣本
If # defective < lower limit, accept 若不良數 < 下界限 , 允收
If # defective > upper limit, reject
若不良數 > 上界限 , 拒收If # defective between limits, resample
若不良數界於界限內 ,重新抽樣Continue sampling until accept or reject lot based on all sample data
連續抽樣必需站在所有的樣本資料以決定允收或拒收
96
Choosing A Sampling MethodChoosing A Sampling Method如何選擇抽樣之方法如何選擇抽樣之方法
An economic decision 經濟的考量
Single sampling plans單次抽樣計劃
high sampling costs 高抽樣成本
Double/Multiple sampling plans
雙次 /連續抽樣計劃
low sampling costs低抽樣成本