1、Statistical Process Control,Statistical Process Control,Variability exists in everything - people, machines and naturePurpose - to be sure that processes are performing in an acceptable mannermonitor over timeforecast if will still be well in futuredetermine factors that make it out of control,Stati
2、stical Process Control,Design quality checks into the process - eliminate need for inspectionbuild better product during designmake product withstand variation - Taguchi Method,History,Prior to IR craftsmen were responsible for entire process - pride of workmanshipDuring 1920s W. Shewart of Bell Lab
3、s build Control Charts - ASQ formedDuring 1950s Deming introduced QC techniques to Japanese1961 Feigenbaum wrote Total Quality Control,Statistical Process Control,Two main subjectsAcceptance Sampling - test random samples of finished goods and accept or reject whole lotProcess Control - sample proce
4、ss output at different points in time - see if process is in control - determine cause of out-of-control behavior,Statistical Process Control,Monitor the production process in one of three places - before, during or afterPhysically examine some of the unitsBasic issues;where do we inspecthow much an
5、d how often do we inspectdo we inspect variables or attributesBalance two costs: inspection and service,Rules of thumb for where to inspect;,1. Inputs - do not pay for goods that do not meet standards2. Finished Goods - protect your reputation3. Before a costly operation - do not waste time and labo
6、r4. Before an irreversible operation - rework only up to a certain point5. Before a covering operation - painting or assembly,Types of Sampling,Defn: Attribute Sampling - check for the presence or absence of a product characteristicDefn: Variable Sampling - product characteristics that can be measur
7、ed on a continuous scale,Statistical Process Control,used to evaluate the output of a processmost processes exhibit some kind of variationrandom variation - natural variation, uncontrollableassignable variation - can identify causedetermine whether non-random sources of variation are responsible for
8、 problems,Control Charts,Prevent production of excessive number of defects - look for causes of variabilityA time-ordered plot of sample statistics Control limits - determine the difference between random and assignable variation - usually 3 Each new sample statistic is plotted and compared to the c
9、ontrol limits,StatisticalProcessControl,Types of Control Charts,1. Variable Control chartsA. Means Chart - monitors central tendency B. Range Chart - monitors dispersion- both can be used to monitor same process- use 20 - 25 samples to build initial limits- take samples of 5 - 9 observations - reaso
10、nable amount of time and costs less,Types of Control Charts (cont),2. Attributes Control ChartsA. P-Chart - count number of defects in sampleB. C-Chart - count number of defects per unit-building control charts adds to the time and cost of production- attribute sampling is easier than variable - nee
11、d larger sample sizes,Process Capability,Specify ranges for certain product dimensions during designTolerancesMax and min acceptable level of outputBearing for rotating shaft - diameter is 1.25 inches 0.005Process has = 0.02 inches - a 3- control chart limits are wider than tolerances,Capability Ind
12、ex,Shows how well parts fit into the range specified by design limitsIf design limits are wider than process limits, the process can shift and still produce good partsCpk =,Capability Index,Example - light bulbs are produced with an average life of 90 hours. The target design life is 100 hours. The
13、UTL = 120 and the LTL = 80 hours. If = 4.8 what is the Cpk?,4,Acceptance Sampling,PurposesDetermine policy for accepting or rejecting a production lot of materials or componentsSample quality is used to make decisionAdvantagesLess handling damageFewer inspectorsApplicability to destructive testingEn
14、tire lot rejection (motivation for improvement),5,Acceptance Sampling,DisadvantagesRisks of accepting “bad” lots and rejecting “good” lotsAdded planning and documentationSample provides less information than 100-percent inspection,Acceptance Plans,1. Single-Sample Plans - only one sample define n an
15、d cOperating Characteristic Curve - graphs the performance of planAcceptable Quality Level - defines good lotsLot Tolerance Percent Defective - defines bad lotsAverage Outgoing Quality - the percent defects in lots leaving inspection,Acceptance Plans (cont),- Producers Risk ( ) - probability of reje
16、cting good lot- Consumers Risk ( ) - probability of accepting a bad lot- can reduce errors by not biasing the sample and taking a larger one- use the Poisson distribution to compute probability of acceptance,9,Operating Characteristic Curve,a = .05 (producers risk),Acceptance Plans,Plans considers t
17、he interaction among the above criteria - it considers the objectives of the supplier and consumerThe producer would like low probability of rejecting a good lot and consumer wants to not accept a bad lotHP and Japanese vendor,Acceptance Plans (cont),2. Double-Sample PlansA. Draw one sampleB. if number of defects is less than a lower limit, accept lot if number of defects is greater than some upper limit reject lotC. if number of defects is between the limits, take another sample,演讲完毕，谢谢观看！,
