Statistical Methods for Setting Specifications

Course Description
Sometimes specifications are given by the customer or marketplace. But when these are not specified externally, a common task for engineers or designers is to set either product or process specifications. One typical case is to set process windows on inputs so as to achieve a given performance on some outputs. Unfortunately, there is very little guidance about best methods for setting these specifications. This course presents statistical approaches to certain specification problems such as setting specs for standardization, determining statistically valid process windows for inputs, setting specs to achieve a given Cpk, using correlation to set specifications, setting rework limits, or determining statistical guardbands for testing. Notice that the specification problems here are distinct and different from geometrical tolerancing which would usually be covered in a drafting or machinist curriculum.

This information is not widely seen together in one place and will give the participant the ability to improve controls on processes with statistically valid limits. The creation of specifications defines processes. It is critical that these process boundaries be set accurately based on data and sound analysis. This course provides proven techniques to set statistically valid process specification limits or to reexamine current process specifications for validity. This is vital knowledge that every engineer, manager, R&D member, or researcher should know.

Who Should Attend
This course is designed for managers, engineers, or researchers who are responsible for setting internal specifications or process windows.

Learning Objectives
Through training, participants will gain the ability to set specifications on either inputs or outputs using statistically valid techniques. Statistically valid specs often receive better compliance when it is known that violating these specifications will have adverse effects on selected outputs.

Course Outline
Introduction to Setting Specifications

  • Purpose of Specification Limits
  • Types of Specification Problems and Associated Statistical Approaches
    • Tolerance Intervals for Specification of Standardization
    • Cpk Performance
    • Correlation Methods
    • RSM Experiments for Process Windows
    • Bivariate Normal Models
    • Optimal Producer Limits
    • Test Guardband Limits for minimal Escapes and Maximum Yield

Setting Specifications for Standardization

  • Typical Methods for Setting Specs for Standardization
  • Three Types of Statistical Intervals
  • Eight Data Types for Setting Specifications for Standardization
    • Tolerance Intervals
    • Assessing the Overall Process Sigma
    • Distribution Fitting
    • Johnson Transformations
    • Zero-Inflated Distributions
    • Censored Data Methods
    • Examples and Exercises
  • Setting Specifications to Achieve a Given Cpk Value
    • Flowchart

Setting Specifications for Process Windows

  • Experimental Designs to determine process windows for 1 to 4 input factors.
  • Analysis of Designs and Compromise Necessity
  • Converting a Contour Plot for Averages Into a Contour Plot for an Individual Data Point
  • Exercises

Setting Specification by Correlation

  • Inverse Prediction
  • Excel Solution
  • Exercises

Special Topics

  • Bivariate Normal Models
  • Optimal Producer Limits for Rework Decisions
  • Statistically Valid Guardband Limits for Testing

Prerequisites
Basic SPC or the equivalent

Course Format
16 hours
Instructor-led class training, with opportunities to practice learned skills using prepared data
Minitab or JMP Statistical Software
Course Instructor: Paul Mullenix – Consultant, Statistical/Quality Methods and Six Sigma for Manufacturing, Research, and Service Industries

Follow QSG on LinkedIn!
Become a QSG Member today!

Always Keep Improving!