Skip to main content Skip to course navigation

Confirmation Experiment

Learning Objective

By the end of this lesson, the learner should be able to:

  • Understand how DOE results are statistically analyzed.
  • Identify factors that are influential under the selected analysis.
  • Evaluate model adequacy before using a model for optimization.
  • Understand the purpose of optimization and Response Surface Methodology (RSM).
  • Compare predicted performance with actual process performance.

1. Introduction

After an experiment has been completed, the collected data must be analyzed to determine what the experiment has revealed.

DOE analysis helps identify important factors, evaluate whether the model adequately represents the experimental data, and support the selection of process settings that meet the defined objectives and constraints.

The Black Belt therefore moves from:

Experiment → Statistical Analysis → Model Evaluation → Optimization

2. Significant Factors

A significant factor is a factor supported as influential under the selected model and analysis.

The purpose is not simply to identify factors that changed during the experiment, but to determine which factors provide evidence of influence on the response under the analysis being used.

This helps focus improvement efforts on factors that matter to the process.

3. Model Adequacy

Model adequacy considers whether the statistical model represents the experimental data sufficiently for its intended use.

Before using a model to identify optimum settings, the Black Belt should evaluate whether the model is appropriate for the purpose.

If the model does not adequately represent the experimental data, optimization based on that model may not provide reliable guidance.

Therefore:

Analyze → Check Model Adequacy → Then Optimize

4. Optimization

Optimization means finding process settings that best satisfy the defined response objectives and constraints.

The optimum is therefore not simply the setting that produces the highest or lowest response.

The Black Belt must consider:

  • The desired response.
  • Process requirements.
  • Experimental findings.
  • Relevant constraints.

5. Response Surface Methodology (RSM)

Response Surface Methodology (RSM) is a family of methods used to model and optimize responses, particularly when curvature in the response relationship matters.

RSM can therefore support the Black Belt when the relationship between process factors and the response is more complex than a simple linear relationship.

6. DOE Analysis Tools

The source identifies the following tools and techniques:

  • ANOVA for DOE
  • Regression models
  • Response Surface Methodology (RSM)
  • Optimization
  • Minitab DOE analysis

These tools support the transition from experimental data to process understanding and optimization.

7. Application Example

Pharmaceutical Process

A pharmaceutical process is optimized for two responses:

  • Tablet hardness
  • Dissolution

The DOE results are analyzed to understand which process factors influence these responses and to identify settings that satisfy the defined objectives.

This illustrates that optimization may involve more than one response, requiring the Black Belt to consider the overall objectives rather than optimizing a single characteristic in isolation.

8. Case Study

Plastics Manufacturing

A plastics manufacturer studied:

  • Molding temperature
  • Pressure
  • Cooling time

A regression model was used to identify optimal settings.

The resulting settings reduced warping defects by 25% and improved throughput.

Black Belt Learning

The case demonstrates how DOE analysis can move beyond identifying influential factors toward selecting operating conditions that improve process performance.

9. Predicted vs. Actual Performance

An important step after optimization is to compare the predicted performance from the model with the actual performance achieved when the selected settings are implemented.

This provides practical evidence about whether the model-based optimization is behaving as expected.

This leads directly to the next lesson:

Lesson 21 — Confirmation Experiment

where the selected settings are tested to confirm the predicted improvement.

10. Black Belt Perspective

A Black Belt should follow a disciplined sequence:

  1. Analyze the DOE results.
  2. Identify influential factors.
  3. Evaluate model adequacy.
  4. Define the response objectives.
  5. Consider relevant constraints.
  6. Identify suitable settings through optimization.
  7. Compare predicted and actual performance.
  8. Conduct a confirmation experiment.

The key principle is:

Do not optimize a model until you have established that the model is adequate for its intended use.

11. Lesson Practice

  1. Why should model adequacy be checked before optimization?
  2. How do constraints affect an optimum?
  3. Why should predicted performance be compared with actual performance?
  4. What is meant by a significant factor in DOE analysis?
  5. What is the purpose of ANOVA in DOE analysis?
  6. When is Response Surface Methodology particularly useful?
  7. Why is optimization more than simply finding the highest or lowest response?

12. Key Learning Points

  • DOE analysis identifies important factors and evaluates the experimental model.
  • Model adequacy should be considered before using a model for optimization.
  • Optimization identifies settings that satisfy defined response objectives and constraints.
  • RSM supports modeling and optimization when curvature matters.
  • ANOVA, regression, RSM and Minitab DOE analysis are important supporting tools.
  • Predicted performance should be compared with actual performance.
  • Confirmation testing provides the next stage of validation.

13. Lesson Conclusion

DOE analysis converts experimental results into useful process knowledge.

The overall progression is:

Run Experiment → Analyze Results → Identify Significant Factors → Check Model Adequacy → Optimize → Confirm

A Black Belt should treat optimization as a data-based decision, supported by an adequately evaluated model and followed by confirmation in the actual process.