Real-Time Material Identification Using Light Spectroscopy and Support Vector Machine (SVM)

Abstract

Material identification is vital in diverse industries such as automotive and aerospace, and industrial applications including machining, robotics, and smart manufacturing. Aerospace and automotive sectors deal with machining, drilling, pressing, or grinding of multi-material parts, requiring manual process parameter adjustments based on each material due to various inherent material properties causing delays in setup time resulting in extended throughput times, decreasing production rates and increasing costs. In addition, manual adjustment may lead to a decrease in the quality of the final part. Thus, there is a need for an automated system that can detect the material type in real-time and employ that information to dynamically adjust the machining, drilling, pressing, or grinding parameters. This paper focuses on merging a low-cost light spectroscopy sensor in the wavelength range of 410 nm (UV) to 940nm (IR) and support vector machine (SVM) to facilitate material identification on automated production lines. Various materials including aluminum, acrylonitrile butadiene styrene (ABS), wood, polyvinyl chloride (PVC), plain carbon steel, polyamide (PA), polylactic (PLA), and galvanized plain carbon steel were examined. The findings revealed that, except for PLA and aluminum, all materials achieved very high accuracy, recall, precision, and F1-score of 100%. PLA showed 90% accuracy and recall, along with 100% precision and 94.7% F1-score. Similarly, aluminum attained 95% accuracy and recall, 100% precision, and a 97% F1-score.