Using AI and optical microscopy to improve manufacturing processes for aluminium vehicle parts

Over the past 3 years, a team of experts from the NRC's Automotive and Surface Transportation Research Centre has been working on an innovative project to develop a new method that uses AI to evaluate the characteristics of aluminium components critical to vehicle manufacturing. The project is being led by the NRC's METALTec industrial R&D group under the Advanced Manufacturing program, with funding from Quebec's centre for aluminum R&D, CQRDA (in French only), and through the contribution of METALTec industrial participants.

The goal of this project is to address growing challenges in the vehicle manufacturing industry that, over the last decades, have started a digital transformation leading to the digitization of manufacturing, known as Industry 4.0. Businesses are looking to improve the efficiency and decrease the costs of their manufacturing processes through the use of autonomous technologies such as robotics, AI and machine-learning.

A digital framework to improve manufacturing processes

To overcome these challenges, the NRC team has successfully developed a digital framework that uses AI to automate and improve methods using optical microscopy for analyzing materials such as aluminum and alloys.

This analysis, called material characterization, involves using methods to measure and analyze the structure and properties of aluminum, such as strength and stiffness. These methods are critical to successfully manufacturing any aluminum component and can include techniques such as optical microscopy to analyze images. However, many of the methods being used were developed long before the dawn of digital transformation in vehicle manufacturing. As a result, industry had to address a number of challenges related to data management and process automation before characterization could be fully integrated into a digital manufacturing environment.

In 2019, the NRC team at the aluminium technology centre started to explore different possibilities to find the best way to analyze the precise images obtained from optical microscopy, the characterization technique used by most vehicle manufacturers. The team selected this technique because it is commonly used to study and understand the composition of a material using a precise image.

As part of this research, team members built a variety of aluminum parts they could analyze with optical microscopy. Doing so allowed them to create a database for real-world use. For example, the data was used to gain insight on the mechanical properties and behaviours of materials, essential information when manufacturing metallic vehicle components.

Potential outcomes

This project will lead to more efficient optical microscopy processes for developing materials and assessing quality during vehicle production. While manual characterization of images can be a lengthy and tedious operation, automated characterization using AI can be carried out up to 10 times faster, reducing R&D overhead and time-to-market for new products.

This project has also improved data analysis capabilities. Regular optical microscopy can be done on only a few samples at a time and often involves looking at only 1 specific element of the microstructure. "With the added automated components, optical microscopy can generate much more data and allow a more detailed analysis of the structure and components in a sample's microstructure," says Marc-Olivier Gagné, project lead and research officer at the NRC's Automotive and Surface Transportation Research Centre in Saguenay (Quebec).

These improved processes can be applied to a number of industries, but the focus of the NRC team's work was vehicle parts manufacturing. Because all vehicles contain metallic components, they require advanced materials and manufacturing processes in order for them to meet the requirements for their specific purpose, requirements related to mechanical properties, durability and aesthetics, for example. Characterization is a reliable, efficient and economically viable process and an indispensable tool for the automotive industry.

On the path to success

At the 2022 North American Die Casting Association ( NADCA ) conference, the association recognized the work of the NRC team with the best congress paper of the year award for their article on this work: Local microstructure-properties model for HPVDC Aural™-2 using image analysis and machine learning.

Phase 2 is now underway looking at using more advanced AI to find practical applications for these emerging technology solutions that will benefit the transportation equipment manufacturing industry. Phase 2 will ultimately allow users to easily generate large data sets of images tailored to specific needs instead of using experimental data, which requires longer overall characterization procedures.