Automated Detection and Root Cause Analysis of Casting Defects

Collaborative research with the Nissan Casting Automotive Plant

As part of a collaborative research initiative with Nissan, this project pioneers using explainable and generative AI for fault detection, diagnosis, and root cause analysis in high-pressure die casting processes. There are several elements to this novel Industry 4.0 project, including new part tracking equipment and processes to capture which parts are being processes when certain data is collected from different machines, novel automated visual inspection using new cameras and AI defect detection, and novel deep learning explainable AI root cause analysis.

Leveraging the new real manufacturing data now available on each part from each location in the plant when the part was processed, the system integrates a Generative AI architecture to detect anomalies within the dataset with high precision. Association of anomalies with inspected defects on a part-by-part basis enables root cause analysis of the detected defects. Explainable AI provides interpretable insights into the model’s decisions, enabling engineers to trace faults back to their root causes in near real-time.

This work not only aims to significantly reduce defect rates on the production line but also contributes to the broader field of trustworthy AI in industrial systems. By embedding transparency into AI-driven quality control, the project strengthens Nissan’s commitment to innovation while showcasing The University of Melbourne’s leadership in applied AI research with direct industry relevance.