Total Quality Management BUS 3 – 142 Statistics for Variables Week of Mar 14, 2011
Page 2 2 Ishikawa’s Basic Seven (7) Tools of Quality –Process Maps –Check Sheets –Histograms –Scatter Plots –Control Charts –Cause & Effect (“Fishbone”) Diagrams –Pareto Analysis
Page 3 3 Reasons to carefully monitor processes –Ensure compliance to specifications –Continuous Improvement –Checking for dispersion –Look for variation and reducing the variation –Understand Randomness vs. Abnormality
Page 4 4 Separating Random variation from Non-Random variation –Product Quality –Machine performance –Budgets –Forecasts –Body temperature –Traffic patterns Process Control provides data to isolate Real Problems vs. natural variability in a process Not every imperfect measurement or event triggers immediate Corrective Action
Page 5 5 Remember The purpose of Process Control is to quickly detect abnormal data and trends for appropriate Corrective Action and to enable Continuous Improvement It is not about IDEAL performance, it is about Controlled, sustained performance It is not meant to predict exact future performance but helps predict RANGES of future performance
Sampling
Page 7 7 Key Statistical Measures Mean –Average Standard Deviation –A measure of variability around the mean –The basis for using probability in anlysis Upper Control Limit –A calculation around the process mean, based on a Normal Distribution Lower Control Limit –A calculation around the process mean, based on a Normal Distribution
Page 8 8 Monitoring Samples vs. Entire populations –Lower cost –Less time –Less disruptive –A practical alternative when destructive testing is required
Page 9 9 Factors when selecting Sample Groups –Ensure that every piece has the same probability of being chosen to be sampled –Gather data at selected time intervals (e.g. every 1 minutes / hour / shift) –Gather data at selected Quantities produced (e.g. every 25 th unit, 100 th unit) –Understand significant inputs or regular events (e.g. Time of Day, Shift changes, preventative maintenenance)
Constructing Process Control Charts
Page Key Elements when implementing Process Inspection –What Type of Inspection –Population –Random –Which sub-groups –Which critical attributes to be sampled –Size of samples –Who will perform the inspection –Who will monitor and analyze the data
Page Generalized Procedure for Developing Process Charts 1.Identify critical operations where inspection might be needed If the operation is performed improperly, the product will be negatively affected 2.Identify critical product characteristics that will result in either good or poor functioning of the product 3.Determine whether the product characteristic is variable or attribute 4.Select the appropriate Control Chart 5.Establish the Control Limits and use the chart to continually monitor and improve 6.Update the limits when changes have been made to the process * Adapted from Foster, Quality Management, Fourth Edition, Prentice Hall
Page Control Charts: Variable & Attribute Data –Variables –Weight –Thickness –Height –Heat –Tensile strength –Attributes –Pass / Fail –Defects (Parts Per Million) * Adapted from Foster, Quality Management, Fourth Edition, Prentice Hall
Page Histogram Before using a Tool designed for a Normal Distribution, Make sure that the data are Normally Distributed
Page Normal Distribution Additional discussion Page 340
Page Key Definitions
Page Control Chart Example
Page Control Chart Example, Continued UCLX-BarCLLCL
Page Interpreting Control Charts –All points lie within the Control limits –The point grouping does not form a particular form Control and Randomness –Run: When points line up on the same side of the Center Line. Three points together above or below the line is a run. A run of 7 points is considered an abnormality. A run 0f 10 out of 11 or 12 out of 14 is also considered a run –Trend: A continued rise or fall of 7 points; can cross the center line (a specific run) … Also known as “Drift” Key signs of Non-randomness Additional analytics on Page 345
Page Interpreting Control Charts
Page Additional Points on Control Charts –Control Limits are Calculated, Specification Limits are not calculated –The Sample Factors APPROXIMATE 3 Standard Deviations –The smaller the sample size, the greater the uncertainty (see Factors on p347) –Control Limits should be CONSTANT –Recalculate UCL and LCL only after process has CHANGED
Page Control Chart Summary Inputs Process Outputs Establish Variable or Attribute Date Define Key Characteristics to Measure Confirm Normal Distribution Choose Data Gathering Methodology Train Users Collect Data Plot Data Check for Randomness Identify Randomness Identify non-Randomness Stop production if necessary Discover improvement opportunities Recalculate Control Limits
Page Moving Range Charts –Variable data only –Volumes are very low –Single points are recorded –Not samples or subgroups, –Requires a Normal distribution
Process Capability
Page Illustration of Process Capability vs. Product Specifications UCL LCL X USL LSL Observation Key Characteristic (dimension, functionality, delivery, etc..) The Supplier is likely to produce conforming parts all the time
Page Illustration of Process Capability vs. Product Specifications UCL LCL X USL LSL Observation The Supplier is likely to produce a quantity of non-conforming parts Key Characteristic (dimension, functionality, delivery, etc..)
Page Applying Process Capability to Supplier Selection –If a Supplier’s process consistently meets or exceeds Customer Specifications, consider the following: Increasing spend on the items (if not Single Source) Introducing new items to be supplied Partnerships and collaborative design where appropriate –If a Supplier’s process misses Customer Specifications, consider: Changing the Supplier Changing the Specification (when possible) Improving the Supplier (if business case justifies)