Controlchart英文版

1、Controlchart英文版What is a Control Chart? A trend chart with control limits Graphical representation of process performance, where data is collected at regular time sequence of production Valuable tool for differentiating between a common cause & a special cause Evaluating whether a process is or

2、is not in a state of statistical control It lets the data talk by itself & basis for data-driven decisions2Control LimitsA typical control chart consists of three lines :-Upper ControlLimit (UCL)Center Line(CL)Lower ControlLimit (LCL)CL: The average (measure of location) process performance when

3、 the process is in-controlUCL & LCL: The range of usual process performance when the process is stable. Lines drawn 3 standard deviations (3 sigma) on each side of the center line.3How to Set-up a Control Chart? Select an area which allows for fruitful application Identify critical parameters to

4、 be monitored based on the feedback from customers, level of quality & reliability, process performance, etc. as reflected in the FMEA of the specific operations Inherently unstable parameters or high excursions Low process capability Critical process inputs Customers required parameters Select

5、measurement tools. Verify the measuring equipment is accurate, stable & capable by performing Measurement Capability Analysis (MCA).4 Select appropriate type of control chart to be used Decide sampling plan (based on variance component analysis from PDC) How samples are to be taken Subgroup samp

6、le size (n) Frequency of data collection Gather data to establish the control chart. A minimum of 30 subgroups is required over a time frame as determined by the sampling plan. Plot the data in time order on a Trend ChartHow to Set-up a Control Chart?5 Check the control chart assumptions & see y

7、our site statistician for advice if these assumptions are not met. Compute the control limits & plot them on the trend chart Outliers identification & exclusion Exclude the Out-of Control (OOC) points or outliers for which there are verified/confirmed special causes from the chart Re-compute

8、 the control limits, excluding the OOC points If there are fewer than 30 points remaining at any time, collect more data. Its very important that the control limits are calculated using at least 30 subgroups.How to Set-up a Control Chart?6 Validate the computed control limits against data collected

9、by re-plotting the control chart with data & new control limits Do the limits detect known problems? Are the limits too sensitive? Would they flag problems you do not know how to react to? Use the control limits established to monitor the critical parameter identified For each & every identi

10、fied critical parameters, every machine should have a separate control chart with separately computed control limitsHow to Set-up a Control Chart?7Control Chart ClassificationsClassifications of control charts are depending onthe type of data Variables data A characteristic measured on a continuous

11、scale resulting in a numerical value E.g. Kerf Width, BLT, Void Size, Bond Pull Strength, Coplanarity, Ball Height, etc. Attributes data A characteristic measured by # of conforming & non-conforming to a specification. Output is classified as pass/fail or accept/reject. E.g. Broken Wire, Lifted

12、Bond, FM, Chipping, Bent Lead, etc. Can be expressed in terms of fraction, percentage, count or DPM8Control Charts For VariablesControl ChartSymbolDescriptionSample SizeX - RXMean of SampleMust Be Equal(Mean-Range) ChartRRange of SampleMust Be EqualX - MR (Individual-XIndividual MeasurementOneMoving

13、 Range) ChartMRRange Between Individual MeasurementTwoX - S (Mean-StandardXMean of SampleMust Be Equal Deviation) ChartSStandard Deviation of SampleMust Be Equal Note: For X - MR Chart, MR Chart is optional.9Control Charts For VariablesNote: S Chart can also be used for any subgroup sample size (n)

14、especially for automated SPC system as standard deviation (s) is a better estimator for within lot variation.10Control Limits for Variable Control ChartsNotes:1. Constants A2, A3, D3, D4, B3 & B4 are given in Appendix A.2. The Moving Range (MR) Method is used to determine the control limits for

15、Mean Chart at Intel A/T sites.11Why MR Method is used to determine Control Limits for Mean Chart? Most Intel A/T production processes have a larger run-to-run variation than within-run variation Traditional control chart formulas developed in the 20s by Walter Shewhart considerably underestimate con

16、trol limits, i.e. too narrow 12Traditional vs. MR Method859095100105110115OrderAvg=99.1LCL=96.5UCL=101.8Mean of Meas.708090100110120130OrderAvg=99.1LCL=76.7UCL=121.5Mean(Meas.)Traditional control chart formulasare used.Moving Range (MR) Method is used.X-bar Control ChartX-bar Control Chart13Caution

17、When Using MR Method If there is autocorrelation, MR(Summary Stat) will underestimate the true process variation & the control limits will be too narrow If autocorrelation is evident, see your site statistician for better control limits computation method Methods for calculating control limits:

18、Std Dev Method Moving Range (MR) Method Percentile Method (Normal Probability Plot) Traditional Shewhart Method14TimeDataConsecutive lotsare very similarLots far apart maybe very differentTime-related condition where consecutive data values are correlated (i.e. dependent)Data values collected nearby

19、 in time are very similarData values collected far apart in time may be very differentTend to drift over time; some drift gradually, others may have occasional sudden changes in direction between periods of relative stabilityAutocorrelation15X - R Chart Concept It plots some specific characteristic

20、or measurement Consists of Two Portions : X Chart Plots the mean of the X values in the sample Shows the changes of the mean of one sample to another R Chart Plots the range of a sample Shows the changes in dispersion or process variability of one sample to another16Computing Control Limits for X -

21、R Chart Obtain at least k = 30 subgroups with n = 3, 4 or 5 on a data sheet Compute the Mean for each subgroup, X= (X1 + X2+ X3 . + Xn) / n Compute the Range for each subgroup, R = Xmax - Xmin Compute the Moving Range for each subgroup, MRi = | Xi - Xi-1 |17Computing Control limits for X - R Chart C

22、ompute the Overall Mean, X = (X1 + X2 + X3 . + Xk) / k Compute the Average of Range, R = (R1 + R2 + R3 . + Rk) / k Compute the Average of Moving Range, MR = (MR2 + MR3 + MR4 . + MRk) / (k - 1)=18 Compute the Control Limits: Constant D3 & D4 are given in Appendix A Draw the control limits on both

23、 the X - R chart respectively X Chart UCL (X) = X + 2.66MR CL (X) = X LCL (X) = X - 2.66MR R Chart UCL (R) = D4R CL (R) = R LCL (R) = D3R=Computing Control limits for X - R Chart19 X ChartUCL (X) = X + 2.66MR= 7.64 + 2.66(0.68) = 9.45CL (X) = X = 7.64LCL (X) = X - 2.66MR= 7.64 - 2.66(0.68) = 5.83 R

24、ChartUCL (R) = D4R= 2.114(0.55) = 1.16CL (R) = R = 0.55LCL (R) = D3R = 0Example of Computing Control Limits for X - R Chartn = 5, D3 = 0, D4 = 2.114201) Calculate the control limits for X - R chart based on data collected from WW35.1 - WW37.1.2) Plot the control limits on the charts & monitor fo

25、r WW37.2 and WW37.3. ExerciseInterpretation of X - R ChartSome special causes of out-of-control for X Chart Changes in machine setting or adjustment MS-to-MS technique inconsistent Changes in material R Chart Machine in need of repair or adjustment New MSes Materials are not uniform22Attributes Cont

26、rol Charts Attribute control charts are useful when it is difficult or impractical to monitor a process numerically (on a continuous scale) A defect is an individual failure to meet a single requirement A defective unit is a unit that contains one or more defects23Control Charts For Attributes24Cont

27、rol Limits for Attribute Control ChartsConventional FormulasControl ChartUCLCLLCLpnpcuCCC*Notes:1. * The formulas are based on a a = 0.10.25 The 3 Standard Deviation Method is highly recommended in HVM environment for p, np, c or u Chart. To be discussed later. Methods based on the Binomial or Poiss

28、on distribution are also used depending on the application. Consult your site statistician for advice on this. As for CCC Chart, it is based on Geometric distribution & works well with low dpm environment.Control Limits for Attribute Control Charts26p Chart Concept It plots proportion of defecti

29、ve units in a sample The proportion of defective units in a sample can be in terms of fraction, percent or dpm It allows us to chart production processes where sample size cannot be equal It can be successfully used for processes with 10,000 dpm and above27Binary DataLot12345*Distribution ofIndividu

30、al LotDistribution ofLot MeansOverall Distributionof Combined LotsGood= 0Bad= 101Lot MeanVariance ComponentsMany products are manufactured in batches or lots. Due to random fluctuations, these lots will vary in quality even though the process is stable.28Variance ComponentsThe total variation can be

31、 separated into 2 variance components:s s2 2Total = s s2 2Lot + s s2 2Withinwheres s2 2Total = The total variation in the processs s2 2Lot= The amount of variation between lots. Its is a measure of how different the lots are from each other.s s2 2Within= The amount of variation within each lot. It i

32、s a measure ofhow different the parts within each lot are from each other.Note: Most conventional statistical analysis methods for Binary data ignore the existence of variance components, andassume that the population defective rate is constant & does not vary from lot to lot. The result of this

33、 is a rather unrealistically low estimate of the true process variation.29Computing Control Limits for p Chart with 3 Std Dev Method Obtain at least k = 30 subgroups or lots. Data collected in # of units inspected & # of units rejected. Compute the defective rate from the ith lot (i =1,2,.,k),pi

34、 = # of units rejected / # of units inspected Compute the average of the ps, Compute the standard deviation of the ps, ()Sppk1pi2i 1k=ppkii 1k=30 Compute the Control Limits: - CL (p)= p - UCL (p) = p + 3Sp - LCL (p) = p - 3Sp When LCL UCL or Point LCL For an automated SPC system with automated appli

35、cation of SPC trend rules, its highly recommended to add 4th rule to detect large shifts in mean, i.e. 2 out of 3 consecutive points at least 2 std dev beyond the centerline, on the same side Add other rules depending upon process knowledge ability to respond criticality of the monitor sensitivity r

36、equirements for the monitor43Trend Rule Recommendations Only use the trend rules that signal process instabilities for which you are capable of responding Justification needed for not using other SPC trend rules Std dev & range charts may choose not to react to Point 1.0 Definitely a problem : |Change Ratio| 1.5Note: This Change Ratio utility is available in SPC View.49Given:UCLcurrent = 11.36LCLcurrent = 7.64F