Statistical Process Control (SPC) – Advanced

Course Description
This second course in SPC fills a void left by many, if not virtually all, textbooks on SPC.  Found among the standard assumptions for SPC are 1) no between subgroup variation, 2) normally distributed data, 3) uncorrelated data in time (no serial or autocorrelation), and 4) homogeneous subgroups. However, these so-called standard assumptions in textbooks seem to be more the exception than the rule for real industrial processes. Thus, this course presents the practical side of SPC implementation, including SPC chart design for number of subgroups, frequency of subgroups, number of out-of-control (OOC) rules, how to handle multilevel batched processes, non-normal data, serially correlated data, fixed effects within subgroups, as well as special SPC techniques such as zero-inflated models for detection limit censored data and normalized charts.

This is the practical side of SPC that is usually not taught in textbooks but is essential to meet the practical demands of industry. By following techniques from this course, SPC charts will better be able to separate common from special cause and be truly effective for making the correct process decisions. These methods are a “must” for anyone who has struggled with making SPC effective in real industrial situations.

Who Should Attend
This course is designed for engineers, quality professionals, researchers, and managers who need to implement SPC in industry with real-world problematic data such as batched processes, non-normal data, and data correlated in time.

Learning Objectives
Through training, participants will:

  • Know how to design effective SPC charts and measure the performance of SPC
  • Be able to make effective SPC charts for both 2 and 3 level nested processes, handle fixed effects, correlated data in time and non-normal data
  • Know how to use zero-inflated distributions and reliability distributions for censored data
  • Know how normalized charts work and their advantages

Course Outline

  • Uses of SPC:  On-Line vs. Off-Line, SPC vs. STC
  • Origins of SPC
  • Review of Basic SPC
    • Common versus Special Causes
    • Data Type
    • Selecting the Right Control Chart
    • Why SPC Works
    • Errors on Control charts
    • CLCR
    • Assumptions of Traditional SPC as Listed y Shewhart
  • Four Problems with SPC in Industry

SPC Chart Design and Performance Measures

  • Types of SPC Performance Measures
    • Practical Metrics
    • Theoretical Metrics of ARL and the OC Curve
  • Design of SPC
    • Subgroup Size
    • Subgroup Frequency
    • Number of OOC Rules
    • Economic SPC Design

SPC for Nested Processes

  • Variance Component Estimation
  • Estimation Using Mean Squares
  • Variation for the Mean in Batched Processes
  • Overall Standard Deviation vs. Variance Component Estimates
  • SPC for 2-Level Batched Processes
  • SPC for 3-Level Matched Processes

SPC for Fixed Effects

  • Random vs. Fixed Effects
  • Two Strategies
    • Separate Charts for Fixed Factor Levels
    • Individual Chart Solution

SPC for Serially Correlated Data

  • Handling a Linear Trend
  • Describing Serial Correlation
    • Autocorrelation
    • Partial Autocorrelation
  • k-Sigma Limits
  • EWMA for Data with Positive Autocorrelation

SPC for Non-Normal Data

  • Effect of Non-Normality
  • Assessing Non-Normality
  • Fitting a Distribution Model
  • Setting Probability Limits
  • Using a Johnson Transformation
  • Group Exercise with Hands-On Data Collection

Special Topics

  • Handling a Special Mixture Case with a Zero-Inflated Distribution
    • Other Censored Data Limits
  • Robust Estimation of Control Limits
  • Normalized Charts

Basic SPC or the equivalent

Course Format
16 hours
Instructor-led class training, with opportunities to practice learned skills using prepared data, live demonstrations, and data collected real time in class
Minitab or JMP Statistical Software

Follow QSG on LinkedIn!
Become a QSG Member today!

Always Keep Improving!