Process Quality. BIZ Production & Operations Management. Sung Joo Bae, Assistant Professor. Yonsei University School of Business

Transcription

1 BIZ Production & Operations Management Process Quality Sung Joo Bae, Assistant Professor Yonsei University School of Business Disclaimer: Many slides in this presentation file are from the copyrighted material in 2010 by Pearson Education, Inc. Publishing as Prentice Hall.

2 Verizon s effort in quality control Video: Can you hear me now? Campaign Tear-down analysis & thorough testing on H/W Network test (H/W test on the side) 1M miles/year testing trip 3M voice calls and 16M data tests Customer involvement (SKT, KT)

3 Costs of Quality A failure to satisfy a customer is considered a defect Defects cause cost Prevention costs Preventing defects before they happen Redesign the processes and products/services Continuous improvement by employees Suppliers involvement Appraisal costs Costs incurred to identify and assess performance problems

4 Costs of Quality Internal failure costs Costs resulting from defects that are discovered during the production of service/product Rework some aspect of product/service should be performed again Scrap the item is unfit for further processing External failure costs Costs that arise when a defect is discovered after the customer received the service or product

5 Total Quality Management A philosophy that stresses three principles for high levels of process performance and quality Customer satisfaction Figure 5.1 TQM Wheel

6 Total Quality Management Customer satisfaction (internal & external) Conformance to specifications (e.g. consistent quality, on-time delivery, delivery speed) Value (= Benefit Cost) Fitness for use (product features or service convenience) Support Psychological impressions (atmosphere, image, or aesthetics)

7 Total Quality Management Employee involvement Cultural change quality at the source Teams Problem-solving teams (aka Quality Circles) Small groups of supervisors and employees who identify, analyze, and solve process/quality problems Special-purpose teams Focus on an important issue such as new policy implementation, new technology implementation, etc. Self-managed teams Highest level of worker participation Members learn all the tasks involved in the operation Job rotation Managerial duties vacation scheduling, hiring, etc. Design processes

8 Total Quality Management Continuous improvement Kaizen ( 改 善 ) A philosophy of continually seeking ways to improve processes Not unique to quality applies to process improvement as well. Problem-solving tools (such as SPC statistical process control) should be given Make SPC a normal aspect of daily operations Build work teams and employee involvement Develop operator ownership in the process

9 The Deming Wheel Plan Act Do Study Figure 5.2 Plan-Do-Study-Act Cycle Problem-solving process

10 Six Sigma A comprehensive and flexible system for achieving success by minimizing defects and variability in the process. Driven by understanding customer needs & use of facts, data and statistical methods Process average OK; too much variation Process variability OK; process off target X X X X X X X X X X X X X X X X X X Reduce spread Process on target with low variability Center process X X X X X Figure 5.3 Six-Sigma Approach Focuses on Reducing Spread and Centering the Process

11 Six Sigma Improvement Model Define Measure Analyze Improve Control Figure 5.4 Six Sigma Improvement Model Continuous efforts in achieving process goals Assumes the analyzability of the processes Emphasizes the involvement from the entire organization, especially the senior management

12 Six Sigma Improvement Model

13 Acceptance Sampling Application of statistical techniques Acceptable quality level (AQL) Criteria for defects that will be accepted (e.g. 5 parts per 100,000) Sample testing should pass this level. Otherwise the full-scale inspection should be done. Linked through supply chains

14 Acceptance Sampling Buyer Manufactures furnaces Firm A uses TQM or Six Sigma to achieve internal process performance Motor inspection Firm A Manufacturers furnace fan motors TARGET: Buyer s specs Supplier uses TQM or Six Sigma to achieve internal process performance Yes Accept motors? No Blade inspection Supplier Manufactures fan blades TARGET: Firm A s specs Yes Accept blades? No Figure 5.5 Interface of Acceptance Sampling and Process Performance Approaches in a Supply Chain

15 Statistical Process Control Used to detect process change A increase in the average number of complaints per day at a hotel An increase in the number of scrapped units at a milling machine Variation of outputs Performance measurement Mean Location; Range or S.D. Spread Variables: continuous scale (length of time, diameter of parts) Attributes: discrete scale (conformance to complex specification) Sampling Complete inspection When cost of failure matters Automated inspection Sampling When inspection cost is high and inspection affects the product or service Sample size Interval between successive samples Decision rules on when to take actions

16 Sampling Distributions 1. The sample mean is the sum of the observations divided by the total number of observations x n i 1 n x i where x i = observation of a quality characteristic (such as time) n = total number of observations x = mean

17 Sampling Distributions The range is the difference between the largest observation in a sample and the smallest. The standard deviation is the square root of the variance of a distribution. An estimate of the process standard deviation based on a sample is given by x i n 1 x 2 or x 2 i n 1 n x i 2 where σ = standard deviation of a sample

18 Sample and Process Distributions Mean Distribution of sample means Process distribution Figure 5.6 Relationship Between the Distribution of Sample Means and the Process Distribution 25 Time

19 Causes of Variation in Process Distribution Common causes Random, unavoidable sources of variation Characterized by Location - mean Spread s.d. or range Shape symmetrical or skewed Assignable causes Can be identified and eliminated Change in the mean, spread, or shape Used after a process is in statistical control

20 Assignable Causes Average Process Dist. Change due to the assignable cause (a) Location Time Figure 5.7 Effects of Assignable Causes on the Process Distribution for the Lab Analysis Process

21 Assignable Causes Average (b) Spread Time Figure 5.7 Effects of Assignable Causes on the Process Distribution for the Lab Analysis Process

22 Assignable Causes Average (c) Shape Time Figure 5.7 Effects of Assignable Causes on the Process Distribution for the Lab Analysis Process

23 Control Charts Time-ordered diagram of process performance Mean Upper control limit Lower control limit Steps for a control chart 1. Random sample 2. Plot statistics 3. Eliminate the cause, incorporate improvements 4. Repeat the procedure

24 Control Charts UCL Nominal Assignable causes likely LCL Samples Figure 5.8 How Control Limits Relate to the Sampling Distribution: Observations from Three Samples

25 Variations Control Charts UCL Nominal LCL Sample number (a) Normal No action Figure 5.9 Control Chart Examples

26 Variations Control Charts UCL Nominal LCL Sample number (b) Run Take action Run usually involves five or more observations show a trend (upward/downward) Figure 5.9 Control Chart Examples

27 Variations Control Charts UCL Nominal LCL (c) Sudden change Monitor Sample number Figure 5.9 Control Chart Examples

28 Variations Control Charts UCL Nominal LCL Sample number (d) Exceeds control limits Take action Figure 5.9 Control Chart Examples

29 Control Charts Two types of error are possible with control charts A type I error occurs when a process is thought to be out of control when in fact it is in control A type II error occurs when a process is thought to be in control when it is actually out of statistical control These errors can be controlled by the choice of control limits

30 Control Charts Error control by the choice of the control limits A type I error occurs when a process is thought to be out of control when in fact it is in control A type II error occurs when a process is thought to be in control when it is actually out of statistical control Wider limits Larger type II error cost for not detecting a shift Smaller type I error < search cost for assignable causes Narrower limits Larger type I error search cost for assignable causes Smaller type II error < cost for not detecting a shift Average Time

31 SPC Methods Used for monitoring current process performance & for detecting any changes in the process Control charts for variables R-Chart, or range chart For monitoring the process variability where UCL R = D 4 R and LCL R = D 3 R R = average of several past R (range) values and the central line of the control chart D 3, D 4 = constants that provide three standard deviation (three-sigma) limits for the given sample size

32 Control Chart Factors TABLE 5.1 FACTORS FOR CALCULATING THREE-SIGMA LIMITS FOR THE x-chart AND R-CHART Size of Sample (n) Factor for UCL and LCL for x-chart (A 2 ) Factor for LCL for R-Chart (D 3 ) Factor for UCL for R-Chart (D 4 )

33 SPC Methods Control charts for variables x-chart UCL x = x + A 2 R and LCL x = x A 2 R where x = central line of the chart, which can be either the average of past sample means or a target value set for the process A 2 = constant to provide three-sigma limits for the sample mean

34 Steps for x- and R-Charts 1. Collect data 2. Compute the range 3. Use Table 5.1 to determine R-chart control limits 4. Plot the sample ranges. If all are in control, proceed to step 5. Otherwise, find the assignable causes, correct them, and return to step Calculate x for each sample

35 Steps for x- and R-Charts 6. Use Table 5.1 to determine x-chart control limits 7. Plot the sample means. If all are in control, the process is in statistical control. Continue to take samples and monitor the process. If any are out of control, find the assignable causes, correct them, and return to step 1. If no assignable causes are found, assume outof-control points represent common causes of variation and continue to monitor the process.

36 Using x- and R-Charts EXAMPLE 5.1 The management of West Allis Industries is concerned about the production of a special metal screw used by several of the company s largest customers. The diameter of the screw is critical to the customers. Data from five samples appear in the accompanying table. The sample size is 4. Is the process in statistical control? SOLUTION Step 1: For simplicity, we use only 5 samples. In practice, more than 20 samples would be desirable. The data are shown in the following table.

37 Using x- and R-Charts Data for the x- and R-Charts: Observation of Screw Diameter (in.) Sample Number Observation R x Average Step 2: Compute the range for each sample by subtracting the lowest value from the highest value. For example, in sample 1 the range is = in. Similarly, the ranges for samples 2, 3, 4, and 5 are , , , and in., respectively. As shown in the table, R =

38 Using x- and R-Charts Step 3: To construct the R-chart, select the appropriate constants from Table 5.1 for a sample size of 4. The control limits are UCL R = D 4 R = 2.282(0.0021) = in. LCL R = D 3 R = 0(0.0021) = 0 in.

39 Using x- and R-Charts Step 4: Plot the ranges on the R-chart, as shown in Figure None of the sample ranges falls outside the control limits so the process variability is in statistical control. If any of the sample ranges fall outside of the limits, or an unusual pattern appears, we would search for the causes of the excessive variability, correct them, and repeat step 1. Figure 5.10 Range Chart for the Metal Screw, Showing That the Process Variability Is in Control

40 Using x- and R-Charts Step 5: Compute the mean for each sample. For example, the mean for sample 1 is = in. Similarly, the means of samples 2, 3, 4, and 5 are , , , and in., respectively. As shown in the table, x =

41 Using x- and R-Charts Data for the x- and R-Charts: Observation of Screw Diameter (in.) Sample Number Observation R x Average

42 Using x- and R-Charts Step 6: Now construct the x-chart for the process average. The average screw diameter is in., and the average range is in., so use x = , R = , and A 2 from Table 5.1 for a sample size of 4 to construct the control limits: LCL x = x A 2 R = UCL x = x + A 2 R (0.0021) = in (0.0021) = in.

43 Using x- and R-Charts Step 7: Plot the sample means on the control chart, as shown in Figure The mean of sample 5 falls above the UCL, indicating that the process average is out of statistical control and that assignable causes must be explored, perhaps using a cause-and-effect diagram. Figure 5.11 The x-chart from the OM Explorer x and R-Chart Solver for the Metal Screw, Showing That Sample 5 is out of Control

44 Application 5.1 Webster Chemical Company produces mastics and caulking for the construction industry. The product is blended in large mixers and then pumped into tubes and capped. Webster is concerned whether the filling process for tubes of caulking is in statistical control. The process should be centered on 8 ounces per tube. Several samples of eight tubes are taken and each tube is weighed in ounces. Tube Number Sample Avg Range Avgs Q: Assuming that taking only 6 samples is sufficient, is the process in statistical control?

45 Application 5.1 Webster Chemical Company produces mastics and caulking for the construction industry. The product is blended in large mixers and then pumped into tubes and capped. Webster is concerned whether the filling process for tubes of caulking is in statistical control. The process should be centered on 8 ounces per tube. Several samples of eight tubes are taken and each tube is weighed in ounces. Tube Number Sample Avg Range Avgs Q: Assuming that taking only 6 samples is sufficient, is the process in statistical control?

46 Application 5.1 Assuming that taking only 6 samples is sufficient, is the process in statistical control? Conclusion on process variability given R = 0.38 and n = 8: UCL R = D 4 R = LCL R = D 3 R = 1.864(0.38) = (0.38) = The range chart is out of control since sample 1 falls outside the UCL and sample 6 falls outside the LCL. This means that the x calculation is not necessary.

47 Application 5.1 Consider dropping sample 6 because of an inoperative scale, causing inaccurate measures. Tube Number Sample Avg Range Avgs What is the conclusion on process variability and process average?

48 Application 5.1 Consider dropping sample 6 because of an inoperative scale, causing inaccurate measures. Tube Number Sample Avg Range Avgs What is the conclusion on process variability and process average?

49 Application 5.1 Now R = 0.45, x = 8.034, and n = 8 UCL R = D 4 R = 1.864(0.45) = LCL R = D 3 R = 0.136(0.45) = UCL x = x + A 2 R = (0.45) = LCL x = x A 2 R = (0.45) = The resulting control charts indicate that the process is actually in control.

50 Process Capability Process capability refers to the ability of the process to meet the design specification for the product or service Design specifications are often expressed as a nominal value (target) and a tolerance (allowance)

51 Process Capability Nominal value Process distribution Lower specification Upper specification Minutes (a) Process is capable Figure 5.14 The Relationship Between a Process Distribution and Upper and Lower Specifications

52 Process Capability Nominal value Process distribution Lower specification Upper specification Minutes (b) Process is not capable Figure 5.14 The Relationship Between a Process Distribution and Upper and Lower Specifications

53 Process Capability Nominal value Six sigma Four sigma Two sigma Lower specification Upper specification Mean Figure 5.15 Effects of Reducing Variability on Process Capability

54 Process Capability The process capability index measures how well a process is centered and whether the variability is acceptable x Lower specification C pk = Minimum of, 3σ Upper specification x 3σ where σ = standard deviation of the process distribution

55 Process Capability The process capability ratio tests whether process variability is the cause of problems C p = Upper specification Lower specification 6σ

56 Determining Process Capability Step 1. Collect data on the process output, and calculate the mean and the standard deviation of the process output distribution. Step 2. Use the data from the process distribution to compute process control charts, such as an x- and an R-chart.

57 Determining Process Capability Step 3. Take a series of at least 20 consecutive random samples from the process and plot the results on the control charts. If the sample statistics are within the control limits of the charts, the process is in statistical control. If the process is not in statistical control, look for assignable causes and eliminate them. Recalculate the mean and standard deviation of the process distribution and the control limits for the charts. Continue until the process is in statistical control.

58 Determining Process Capability Step 4. Calculate the process capability index. If the results are acceptable, the process is capable and document any changes made to the process; continue to monitor the output by using the control charts. If the results are unacceptable, calculate the process capability ratio. If the results are acceptable, the process variability is fine and management should focus on centering the process. If not, management should focus on reducing the variability in the process until it passes the test. As changes are made, recalculate the mean and standard deviation of the process distribution and the control limits for the charts and return to step 3(control chart).

59 Assessing Process Capability EXAMPLE 5.5 The intensive care unit lab process has an average turnaround time of 26.2 minutes and a standard deviation of 1.35 minutes The nominal value for this service is 25 minutes with an upper specification limit of 30 minutes and a lower specification limit of 20 minutes The administrator of the lab wants to have four-sigma performance for her lab Is the lab process capable of this level of performance?

60 SOLUTION Assessing Process Capability The administrator began by taking a quick check to see if the process is capable by applying the process capability index: Lower specification calculation = = (1.35) Upper specification calculation = = (1.35) C pk = Minimum of [1.53, 0.94] = 0.94 Since the target value for four-sigma performance is 1.33, the process capability index told her that the process was not capable. However, she did not know whether the problem was the variability of the process, the centering of the process, or both. The options available to improve the process depended on what is wrong.

61 Assessing Process Capability She next checked the process variability with the process capability ratio: C p = = (1.35) The process variability did not meet the four-sigma target of Consequently, she initiated a study to see where variability was introduced into the process. Two activities, report preparation and specimen slide preparation, were identified as having inconsistent procedures. These procedures were modified to provide consistent performance. New data were collected and the average turnaround was now 26.1 minutes with a standard deviation of 1.20 minutes.

62 Assessing Process Capability She now had the process variability at the four-sigma level of performance, as indicated by the process capability ratio: C p = = (1.20) However, the process capability index indicated additional problems to resolve: ( ) ( ) C pk = Minimum of, = (1.20) 3(1.20)

63 Application 5.4 Webster Chemical s nominal weight for filling tubes of caulk is 8.00 ounces ± 0.60 ounces. The target process capability ratio is 1.33, signifying that management wants 4-sigma performance. The current distribution of the filling process is centered on ounces with a standard deviation of ounces. Compute the process capability index and process capability ratio to assess whether the filling process is capable and set properly.

64 Application 5.4 a. Process capability index: x Lower specification C pk = Minimum of, 3σ Upper specification x 3σ = Minimum of = 1.135, = (0.192) 3(0.192) Recall that a capability index value of 1.0 implies that the firm is producing three-sigma quality (0.26% defects) and that the process is consistently producing outputs within specifications even though some defects are generated. The value of is far below the target of Therefore, we can conclude that the process is not capable. Furthermore, we do not know if the problem is centering or variability.

65 Application 5.4 b. Process capability ratio: C p = Upper specification Lower specification 6σ = = (0.192) Recall that if the C pk is greater than the critical value (1.33 for four-sigma quality) we can conclude that the process is capable. Since the C pk is less than the critical value, either the process average is close to one of the tolerance limits and is generating defective output, or the process variability is too large. The value of C p is less than the target for four-sigma quality. Therefore we conclude that the process variability must be addressed first, and then the process should be retested.

66 Quality Engineering Quality engineering is an approach originated by Genichi Taguchi that involves combining engineering and statistical methods to reduce costs and improve quality by optimizing product design and manufacturing processes. The quality loss function is based on the concept that a service or product that barely conforms to the specifications is more like a defective service or product than a perfect one.

67 Loss (dollars) Quality Engineering Lower Nominal Upper specification value specification Figure 5.16 Taguchi s Quality Loss Function

68 International Standards ISO 9000:2000 addresses quality management by specifying what the firm does to fulfill the customer s quality requirements and applicable regulatory requirements while enhancing customer satisfaction and achieving continual improvement of its performance Companies must be certified by an external examiner Assures customers that the organization is performing as they say they are

69 International Standards ISO 14000:2004 documents a firm s environmental program by specifying what the firm does to minimize harmful effects on the environment caused by its activities The standards require companies to keep track of their raw materials use and their generation, treatment, and disposal of hazardous wastes Companies are inspected by outside, private auditors on a regular basis

70 International Standards External benefits are primarily increased sales opportunities ISO certification is preferred or required by many corporate buyers Internal benefits include improved profitability, improved marketing, reduced costs, and improved documentation and improvement of processes

71 Industry Life-cycle as an S-Curve Performance Maturity Takeoff Discontinuity Ferment Source: Foster (1986) t

72 The S-Curve Maps Major Transitions Performance Maturity Takeoff Discontinuity Ferment Source: Foster (1986) t

73 Success trap Fit Congruence in strategy, critical tasks, people, org design, culture Success Size & Age Organizations get larger, more structured, & older Success in stable environment or Failure When environments shifts Inertia

74 Build an Ambidextrous Senior Team Ambidextrous senior teams must manage both more mature, operationally focused businesses and higher growth, emerging businesses High performing senior teams show: High conflict, high respect decision making capabilities High levels of trust and truth telling The ability to manage divergent incentive systems and career paths Coupled with processes that support the divergent management of quite different business units E.g. Resource allocation processes that allow for different time horizons, milestones, rates of return Incremental Innovation Unit Senior Management Teams Discontinuous Innovation Unit

75 End of Process Quality Session