2Quality ControlPurpose: monitor and maintain quality levels of products/services/processes, and try to improve product/service/process designs in ways that result in better quality products, services, and processesQuality Control Departments:Traditional focus inspectionsNewer focus quality training, improvements, working with suppliers, few inspections
3Quality ControlProcess variation causes a process to be less consistent, resulting in lower qualityW. Edwards Deming 2 causes of variationCommon causes: normal (typical) variation, inherent to the process, difficult/expensive to reduce; example– a machine is not very accurate because it is oldSpecial causes: due to a specific problem, not inherent to process, usually easy/cheap to fix; example– a machine’s quality level is lower one morning, and after you investigate you discover that operator has hangoverTo improve quality, identify and fix special causes of variation first (low hanging fruit)
4Process Variation2 different machines can do the same operation. Which should be used?10,000 units of a gear shaft needed – set up machine and do test run of 100 units.9.009.059.19.158.958.98.85specsdiameter (mm)machine A20% out of spec
5Process Variation Which should be used, machine A or machine B? 9.00 9.059.19.158.958.98.85specsdiameter (mm)machine B50% out of spec
6Quality Control Throughout Productive Systems Raw Materials,Parts, andSuppliesProductionProcessesProducts andServicesAcceptanceTestsControl ChartsQuality ofInputsMonitoring Quality ofPartially Completed ProductsOutputsConversion
7Monitoring Process Quality Run Diagrams useful when starting a process(most likely time for errors)measure every pieceplot measurementslook for outliers and patterns to investigateprocess is in-control if no outliers or patterns (just typical randomness)
8Run Diagram—1st Attempt Outer Diameters of 30 Pieces
9Run Diagram—2nd Attempt Outer Diameters of 30 Pieces
10GP4890 ROSEBURG FOREST PRODUCTS SHIPPING TEMPERATURE 1/21/93 – 2/15/93
11Control Charts X chart and R chart when measuring a variable (e.g., length, weight, volume, viscosity)assumes normal distribution of sample meansX is the measurement of one unitX is the mean measurement of one sampleX is the expected value of the measurementR is the range of one sample (high – low)R is the expected value of the range
12Control Chartsp chartwhen measurement has only two outcomes (e.g., yes/no, defective/nondefective, good/bad)assumes binomial distributionp is fraction or percent of bad parts in a samplep is the expected value of p
13Control Chartsc chartwhen measuring the total number of defects in a samplesample size is 2 or more unitsassumes Poisson distributionc is number of defects in a samplec is the expected value of c
14Control Chartsu chartwhen measuring the number of defects on one unitsample size is one unitassumes Poisson distributionu is number of defects on one unitu is the expected value of uonly difference between u and c charts is sample size (1 or many)
15Control Limits X chart – for sample means Upper control limit (UCL) = X + ARLower control limit (LCL) = X – ARX = expected value of X (average over many samples)R = expected value of R (average over many samples)A = a constant from Table
16Control Limits R chart – for sample ranges UCL = D2R LCL = D1R D1 and D2 are constants from Table
17Control Limits p chart – for fraction/percent defectives in sample p = expected fraction/percent defectives in samplesn = size of one sample (number of units)
18Control Limits c chart – for total number of defects in a sample c = expected number of defects in samplesX and R chart example: Suppose a company wants to startusing X and R control charts. They have collected 25samples of 5 units in each sample to estimate X and R.
24GP4890 ROSEBURG FOREST PRODUCTS % CAUSTIC 1/15/93 – 2/13/93 UPPER CONTROL LIMITLOWER CONTROL LIMIT32.521.510.5
25p Chart Example Process placing labels on jeans Label placement is either defective or non-defective.Typically, about 1.5% of labels are considered defective.Compute control limits for a p chart, with sample size = 200.If a sample has 6 bad labels, is this in-control?
26c Chart ExampleTwice a day, Ford takes a sample of 5 cars after the painting operation to count the number of bad spots on the freshly painted surfaces. The typical number of defects per sample is 3.1 bad spots. Compute 3σ control limits.
27Good/Bad & Investigate or Not AboveUCLBelowLCLX chartR chartp chartc chartu chart
28Guidelines for Determining Which Control Chart to Use Identify the item or product to be evaluated for quality.What characteristic is to be measured?Should every item be checked, or should a random sample be taken?Is the characteristic measured on a continuous scale?Is the item either good or bad, or is the number of defects on one item important?Should the random sample size be one unit, or should it be more than one unit?
29Control Chart Examples 1. A local building contractor builds large custom homes. He wants to use a control chart to monitor the number of problems that customers find in the finished homes. What type of control chart should be used?2. A manufacturer of semiconductors chips plans to use a control chart to monitor the quality of chips they produce. Due to the complexity and density of circuits on each chip, typically 5% to 15% of the chips are faulty, which is about normal for the industry. Faulty chips are returned by customers for a full refund or replacement. What type control chart should be used?3. A professional proofreader checks manuscript pages for typing errors. What type of control chart should be used to monitor the quality of their proofreading?
30Control Chart Examples 4. As cereal boxes are filled in a factory they are weighed for their contents by an automatic scale. The target is to put 10 ounces of cereal in each box. What type of control chart should be used to monitor how well they are achieving their target?5. What type of control chart should Microsoft use to monitor the quality of their software programmers in developing computer code for assigned portions of larger software products?6. What type of chart should a business school use to monitor the quality of incoming MBA students based on GMAT scores?