SQE 品质工程师 职责

按一下以編輯母片標題樣式,按一下以編輯母片,第二層,第三層,第四層,第五層,*,品質控制,(,統計製程管制,),余德成,國立高雄海洋科技大學運籌管理系,2007.5.13,POM2007,國立中山大學人力資源管理研究所,Unit 6,大綱,前言,基本的控制模式,TQC,SPC,抽樣方法,品質管制方法,More,前言,TQM,失敗的原因,管理有兩種,連續改善,基本的控制模式,TQM,失敗的原因,連續改善,Concepts,如何,連續改善,?,5-Why,改善工具,Concept-1,Concept-2,如何,連續改善,?,5-Why,?,改善工具,魚骨圖,基本的控制模式,基本概念,TQC,戴明獎審查檢點表,SPC,統計製程管制,(Statistical Process Control; SPC),統計思維,(,Statistical Thinking ),品質特性,(,Quality Characteristics),資料型態,(,Types Of Data),變異型態,(,Types of Variations),統計方法,(,Statistical Methods),抽樣方法,(,Sampling Methods),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,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,Attributes,計數值,Variables,計量值,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,品質特性,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,資料型態,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,變異型態,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,變異的原因,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,!,產品規格與品變異,Grams,(,a) Location,Average,(,平均值,),共同原因,(,a) Location,Grams,Average,特殊原因,統計方法,統計圖表,統計分配,管制圖,檢定,迴歸,讓資料說話,.Know-why,-3,s,-2,s,-1,s,+1,s,+2,s,+3,s,Mean,平均值,68.26%,95.44%,99.74%,=,Standard deviation,=,標準差,The Norma,l,Distribution,常態分配,Mean,平均值,Central Limit Theorem,Standard deviation,樣本標準差,Theoretical Basis of Control Charts,UCL,管制規格上限,Nominal,中心線,LCL,管制規格下限,1 2 3,Samples,Control Charts,管制圖,1 2 3,Samples,UCL,管制規格上限,管制圖,Assignable causes likely,可能的特殊原因,1 2 3,Samples,UCL,管制規格上限,Nominal,中心線,LCL,管制規格下限,管制圖,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,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,製程管制的三種顯示型態,Uniform,Normal,Beta,Distribution of sample means,樣本平均值分配,Standard deviation of the sample means,(mean),Three population distributions,群體分配,群體與樣本間之關係,Target,At a fixed point in time,固定時間,Time,Target,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,機遇原因之觀察,Process Control Charts,製程管制圖,Control,Charts,Variables,Charts,Attributes,Charts,Continuous,連續的,Numerical Data,Categorical or Discrete,離散的,Numerical Data,計量,計數,管制圖型態,Quality Characteristic,variable,attribute,n1?,n=10 or,computer?,x and MR,no,yes,x and s,x and R,no,yes,defective,defect,constant,sample,size?,p-chart with,variable sample,size,no,p or,np,yes,constant,sampling,unit?,c u,yes,no,管制圖的選定,Produce Good,Provide Service,Stop Process,Yes,No,Assign.,Causes?,Take Sample,Inspect Sample,Find Out Why,Create,Control Chart,Start,Statistical Process Control Steps,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 n50 for p-chart),3) Construct the Control Chart,建立管制圖,4) Analyze the data relative to the control limits. Points outside 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,如何使用管制圖,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,間控製程平均數,Example: Measure 5 samples of solder paste Plot,如計算錫膏厚度之平均值,再點圖,X,Chart,平均值管制圖,use historical data taken from the process when it was “known” to be in 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 size) taken at regular intervals,樣本資料是在一定間隔的時間裡取得,process mean,m,estimated as the average of the sample means (the grand mean or nominal value),假設製程平均值,m,與樣本平均值相同,process standard deviation,s,estimated by:,製程標準差,s,估算由,standard deviation of all individual samples,所有個別值樣本之標準差,OR mean of sample range R/d,2, where,或樣本平均值,/ d,2,sample range R = (,Rmax-Rmin,), d,2,= value from look-up table,全距為,R, d,2,可由查表得知,平均值與標準差估計,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,),X-bar vs. R charts,1. Take samples and measure them.,取樣量測,2. For each subgroup, calculate the sample average and range.,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 revise 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,如何建立管制圖,Alarm,No Alarm,In-Control,管制內,Out-of-Control,失控,Type 1 and Type 2 Error,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,管制圖異常之判定,Test Probability Type 1 Error,2/3,7/7,10/11,16/20,1/1,2(0.00135),0.0027,0.0052,(0.5),7,0.0078,0.00586,0.0059,Type 1 Errors for these Tests,Suppose,m,1,m,Type 2 Error =,where,F,(z) denotes the the cumulative probability of a standard normal,variate,at z,Power = 1- Type 2 Error. Power increases as ,n increases, as (,m,1,-m,) increases, and as,s,decreases.,Extension to,m,1, 1,C,pk,量測之意義,Lower specification limit,Upper specification limit,(a) Acceptance sampling,(b) Statistical process control,(c),c,pk,1,確認並降低製程變異,TYPE I ERROR = P(reject good lot),a,or producers 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),b,or consumers 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,生產者與消費者冒險率,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,品質的定義,Operating Characteristic Curve,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,作業特性曲線,AQL,LTPD,b,= 0.10,a,= 0.05,Probability,of acceptance,P,a,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,Operating Characteristic Curve,OC,曲線,Average Outgoing Quality (AOQ),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,平均出廠品質,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,AOQ Curve,平均出廠品質曲線,Process,製程,Variation,變異,Data,資料,Statistical Tools,統計方法,Statistical Thinking,統計思維,Statistical Methods,統計方法,從,統計思維,到,統計方法,統計訓練,訓練課程,認證體系,薪資調整,抽樣方法,雙次抽樣計劃,多重,(,連續,),抽樣計劃,如何選擇抽樣之方法,Double Sampling Plans,雙次抽樣計劃,Take small initial sample,抽取少量之原始樣本,If # defective upper limit, reject,If # defective between limits, take second sample,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,Multiple (Sequential) Sampling Plans,多重,(,連續,),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, 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,Choosing A Sampling Method,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,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,品質管制方法,Sampling vs. Screening,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,抽樣與篩選,Acceptance Sampling,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,允收抽樣,Sampling Plan,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,抽樣計劃,More.,The End,運籌帷幄,決勝全球,觀念與方法,!,