[T11] Industrial Application

[15P-T11-01] Advanced Eco-Friendly Non Destructive Testing in Powder Metallurgy through AI-Based Acoustic Methodology

*H.-H. Hsiao^{1, 3}, K.-J. Yang², C.-Y. Lien¹, F.-N. Lee⁴, C.-C. Wang², K.-J. Wang³ (1.Auroral Sinter Metals Co., Ltd., Taiwan, 2.Industrial Technology Research Institute, Taiwan, 3.National Taiwan University of Science and Technology, Taiwan, 4.Brilliant Knowbot Machines Co., Ltd., Taiwan)

Keywords:Powder Metallurgy Non Destructive Testing, Audio Recognition, Convolutional Neural Network, Green Manufacturing, Intelligent Manufacturing

This study aims to develop an eco-friendly non destructive testing method for powder metallurgy materials, aiming to improve inspection techniques like magnetic particle inspection, known for their environmental impact. The primary goal is to establish an Intelligent Manufacturing-based AI system for detecting internal cracks in powder metallurgy components after sintering. Acoustic data is collected using an electric hammer and microphone after sintering and processed to create spectrograms. An autoencoder extracts features, followed by cluster analysis to select samples for destructive testing. By integrating audio recognition technology and a convolutional neural network, unique sound features of sintered materials are identified, achieving over 98% accuracy in identifying specific types of powder metallurgy products. This eco-friendly testing method aligns with green manufacturing and showcases AI's potential in powder metallurgy manufacturing. This study demonstrates a significant advancement in enhancing quality control and propelling the powder metallurgy industry towards environmentally friendly manufacturing practices.