Introduction to Design of Experiments

This 8-hour course answers the questions:

• What is a Designed Experiment?
• Why and when do I need a Designed Experiment?
• What are the most important elements of a Designed Experiment?
• What is the REAL way to do a Designed experiment?
• And, Now, How do I set up, and analyze a Designed Experiment..in Detail?

We progress from the language of a Designed Experiment to the overall concept that most often guides efficient
experimentation, namely sequential experimentation. The important things to consider in a designed experiment are paramount to success:

• Discussion with the team (customer) for details regarding needs and expectations, as well as budget.
• Detailing the process to be improved, and understanding them in detail as well as the team.
• Talking to all elements of the process - people actually executing the current process, engineers/supervisors/managers.
• Deciding the overall approach that balances budget with expectations (Fractional factorial or other screening experiment); then proceeding after those results are obtained.
• Executing the first experiment, analyzing the results (using MINITAB or other Statistics package).
• Reviewing results for surprises and for “believability” with the process experts.
• Setting up and doing a follow-on experiment that will identify the main (important) drivers in the process and (very importantly) the “interactions” between main drivers that also affect the process.
• Reporting your results in language understandable to Management.

When you leave this course you can start advising Process Improvement teams in the use of experimental design and help them solve their improvement challenges … AND save your company

Course Outline

1. Background: Historical perspective; Why use DOE?
2. Completely Randomized Designs
   • Theory
   • Randomization, Replication and Local Control of Error
   • Experimental Error vs. Sampling Error
   • Analysis of Variance
3. Multiple Comparisons
4. Examples
5. Factorial Designs
   • Main Effects vs. Interactions
   • Fixed factors vs. Random factors
   • Examples
6. Fractional Factorial Designs
   • Resolution, Aliasing & Confounding
   • Graphical techniques for analyzing significance
7. Examples
8. Response Surface Designs
   • Overview
9. Team Exercises

---

 

Introduction to Weibull Analysis

This 8-hour course answers the questions:

• What is Weibull Analysis and why is it useful?
• What type of data do I need for Weibull Analysis?
• I’ve had a couple failures of a Safety critical part, can Weibull analysis help me?
• How can I calculate the number of future events with the few failures I have?
• What should I expect from Managers/Customers when I present my analysis?

We progress from the language of Reliability & Statistics to quickly move into the what, why and usefulness of the Weibull Failure Distribution. Short explanation via analogy: “The Weibull Distribution is to Reliability as the Normal Distribution is to Statistics.“Weibull analysis generally refers to the process of fitting a Weibull distribution to a set of data, usually time or cycles to failure data, for the same reasons that we fit normal distributions to manufacturing data. The fitted Weibull distribution smooths out small sample variation and produces a concise description of the data with just two statistics.

Conclusions may be drawn:

• about the population the data were drawn from
• Will 99.5% of the population survive until the next overhaul?
• about differences between populations
• Is Population A = Population B?
• Is Supplier A’s time-to-failure distribution = Supplier B’s time-to-failure distribution?
• How long do I have to test a new design?
• Enables engineer to make predictions outside the bounds of the data based on the consistent pattern of variation within the data
• Helps to locate outliers

In summary, the fitted Weibull distribution is a tool that helps turn raw data into useful information.

Course Outline:

1. Introduction/Background
2. Statistical Distributions
3. Weibull Plot Overview
4. How are Failures Plotted on a Weibull plot?
5. Weibull Plot considerations
6. More Weibull Plot considerations
7. What are Maximum Likelihood Estimates?
8. Interpreting Weibull plot Results
9. “Weibayes” Analysis (Weibulls with NO data)
10. Weibull Substantiation & Life testing
11. Identifying more than one failure mode in the data.
12. Working with failures only in a large (unfailed) population
13. Confidence intervals & the Weibull
14. Weibulls with Interval data
15. Weibull Risk Analysis

---