Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Explainable AI for Correct Root Cause Analysis of Product Quality in Injection Moulding

Muhammad Muaz, Sameed Sajid, Tobias Schulze, Chang Liu, Nils Klasen, Benny Drescher

TL;DR

The study tackles explainability in injection-moulding quality prediction by comparing permutation-based SHAP with ICE on RF and MLP models trained from a central composite design dataset. It demonstrates the existence of interactions among six operator-controlled machine settings using the $H$-statistic and shows that SHAP attributions yield correct root-cause identification, while ICE can misattribute due to neglecting interactions. Both RF and MLP achieve very low prediction errors ($\text{MAE}$ in the \(3.7\times10^{-3}\) to \(5.1\times10^{-3}\) range and $\text{MAPE}$ around 0.031–0.043%), enabling reliable, real-time explanations. The findings support deploying SHAP-based cause analysis in automated quality control for injection moulding, improving reliability and reducing scrap, with potential extension to multiple quality characteristics.

Abstract

If a product deviates from its desired properties in the injection moulding process, its root cause analysis can be aided by models that relate the input machine settings with the output quality characteristics. The machine learning models tested in the quality prediction are mostly black boxes; therefore, no direct explanation of their prognosis is given, which restricts their applicability in the quality control. The previously attempted explainability methods are either restricted to tree-based algorithms only or do not emphasize on the fact that some explainability methods can lead to wrong root cause identification of a product's deviation from its desired properties. This study first shows that the interactions among the multiple input machine settings do exist in real experimental data collected as per a central composite design. Then, the model-agnostic explainable AI methods are compared for the first time to show that different explainability methods indeed lead to different feature impact analysis in injection moulding. Moreover, it is shown that the better feature attribution translates to the correct cause identification and actionable insights for the injection moulding process. Being model agnostic, explanations on both random forest and multilayer perceptron are performed for the cause analysis, as both models have the mean absolute percentage error of less than 0.05% on the experimental dataset.

Explainable AI for Correct Root Cause Analysis of Product Quality in Injection Moulding

TL;DR

The study tackles explainability in injection-moulding quality prediction by comparing permutation-based SHAP with ICE on RF and MLP models trained from a central composite design dataset. It demonstrates the existence of interactions among six operator-controlled machine settings using the -statistic and shows that SHAP attributions yield correct root-cause identification, while ICE can misattribute due to neglecting interactions. Both RF and MLP achieve very low prediction errors ( in the to range and around 0.031–0.043%), enabling reliable, real-time explanations. The findings support deploying SHAP-based cause analysis in automated quality control for injection moulding, improving reliability and reducing scrap, with potential extension to multiple quality characteristics.

Abstract

If a product deviates from its desired properties in the injection moulding process, its root cause analysis can be aided by models that relate the input machine settings with the output quality characteristics. The machine learning models tested in the quality prediction are mostly black boxes; therefore, no direct explanation of their prognosis is given, which restricts their applicability in the quality control. The previously attempted explainability methods are either restricted to tree-based algorithms only or do not emphasize on the fact that some explainability methods can lead to wrong root cause identification of a product's deviation from its desired properties. This study first shows that the interactions among the multiple input machine settings do exist in real experimental data collected as per a central composite design. Then, the model-agnostic explainable AI methods are compared for the first time to show that different explainability methods indeed lead to different feature impact analysis in injection moulding. Moreover, it is shown that the better feature attribution translates to the correct cause identification and actionable insights for the injection moulding process. Being model agnostic, explanations on both random forest and multilayer perceptron are performed for the cause analysis, as both models have the mean absolute percentage error of less than 0.05% on the experimental dataset.
Paper Structure (12 sections, 10 equations, 9 figures, 6 tables)

This paper contains 12 sections, 10 equations, 9 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Production setup
  3. Methodology
  4. Chosen Models
  5. Capturing Machine Setting Interactions
  6. Shapley Additive Explanation
  7. Cause Analysis
  8. Results and Discussion
  9. Feature Interactions
  10. Feature Attribution for Cause Analysis
  11. Conclusion
  12. Acknowledgment

Figures (9)

  • Figure 1: Explainable AI broad categories used in manufacturing dam_explainable_2018.
  • Figure 2: Schematic of production setup for training and deployment.
  • Figure 3: PDP plots of machine settings for (a) RF, and (b) MLP.
  • Figure 4: Bar plot of one-versus-all interactions of machine settings measured by H-statistic based on RF.
  • Figure 5: Step plots of packing time (left) and packing pressure (right) vs RF-based weight predictions for different values of the melt temperature.
  • ...and 4 more figures