Statistical Tools for Process Improvement - Applications Jairo Muñoz, Ph.D., CMfgE Iowa Precision Industries (319) 364-9181 Ext. firstname.lastname@example.org
Contents of Todays Meeting Statistical Tools for Process Improvement The Concept of Variation Process Capability –Actual vs. Potential. –Example: Cut to length machinery. The Role of Industrial Experimentation
A METHOD FOR CONTINUOUS IMPROVEMENT Values Vision Goals Objectives Strategies Improvement tools and awareness Planning Customer & Supplier Mission Responsab. Activities Customers Suppliers Requirements Feedback Plan – Do – Check -- Act Process Flow C/S Measures Requirements Baseline Feedback Analysis Improve ID the process Define boundaries Analysis Measure Solve/Test Implement Fool proof Monitor variation System change Benchmark Train Lessons learned Make sure were doing the right things Make sure were doing things right Eliminate things that were not doing right Maintain and improve what were doing right Process analysis Problem solving Hold the gains
What is variation? Variation is the inability to maintain a constant performance. –T–There is natural (NORMAL) variation. –T–There is induced (SPECIAL) variation. Variation means $.
Seven Deadly Sins - Waste Over-production & producing early Delay Transport(material handling) Processing Inventory Motions Defects
THE COST OF VARIATION 68% 10% 22% Costs associated with producing a poor quality product. Costs of assuring a good quality product
Are We Good Enough? 200,000 wrong drug prescriptions yearly 15,000 newborns dropped in hospitals Unsafe drinking water 1 hr each year No utilities for 8.6 hrs each year 2 short/long landings at all major airports each day 500 incorrect surgical procedures weekly 9 misspelled words on every page of every magazine At 99.9%:
STATISTICAL QUALITY CONTROL Need for C/A? Improved enough? Test process options More detail needed? Select condition needing to improve Analyze current process Define data to be collected Collect data Analyze data Implement needed changes Record process changes Record unusual events Establish regular process monitoring NO Yes NO BRAINSTORM PROCESS FLOW DIAGRAM CAUSE / EFFECT DIAGRAM CHECK SHEETS PARETO, SCATTER, REGRESn CONTROL CHARTS DOE, EVOP
SPC, Cp or DOE? Eliminate Special Causes of Variation Check distribution. Decrease variability. USE FACTORIAL DESIGNS Continue Production Is Process Predictable? Is Process Capable? NO Yes
The statistical approach focuses on problem- solving by providing a rational rather than emotional basis for decision-making. It provides the basis for on going improvement. Dr. W.E. Deming
MEASURES ENGINEERING –Percent documents issued on time; ECRs per project. PRODUCTION PLANNING –Actual/Planned deviation; Time lost waiting for parts. MIS –Average response time; Data entry errors per _____. PURCHASING –Purchases/sales ratio; Total dollars; Percent shortages.
Cant Let Obstacles Slow Us We tried that ten years ago Were too busy to fix these problems We dont do things that way here But those companies arent like us –We have different problems Well change, but lets do it very slowly That wont work here
The output of a process in control (predictable)may be compared against its specification. For measurement data, process capability indexes are usually expressed as a ratio of total variability and the specification range. Percent defective may be predicted when the process capability is known for an in control process. If the process capability is sufficiently high, end item inspection becomes a waste of time.
Process Variation Vs. Specifications C p = UTL – LTL = 1.33 = C k 6 C p = 1.33; C k = UTL – Mean = 0.5 3 C p = 1.33; C k = 0 C p = 1.33; C k < -1
Cp in Cut to Length Applications Used as a sales tool Used as a final test tool Used by customer as a verification tool Potential process capabilities of up to 2.1 in short runs (with tolerances of +/- 0.030) Actual process capabilities of up 1.7 in short runs. Customer runs of up to 1.35 for +/- 0.005 machines
Process Fallout - Centered Process Six-sigma philosophy
References Box and Draper, Evolutionary Operations. Box, Hunter and Hunter, Statistics for Experimenters Montgomery, Design and Analysis of Experiments Snee, Hare and Trout, Experiments in Industry Wheeler, Tables of Screening Designs Wheeler, Understanding Industrial Experimentation Software: –www.statease.comwww.statease.com –MatLab; Design Ease; Design Expert; SAS.