Impact Factor 2025:
0.9
CiteScore 2025:
SJR 2025:
Prof Richard (Chunhui) Yang
Prof Richard Yang is the Interim Dean, School of Engineering, Director, Centre for Advanced Manufacturing Technology and Head of Discipline, Mechanical and Mechtronic Engineering, Faculty of Engineering, Computing and Science, Western Sydney University. His research fields include Mechanical Engineering, Manufacturing Engineering, advanced Manufacturing, Additive Manufacturing/3D Printing, Digital Twin, Industry 4.0, Mechanical Design, Engineering Numerical Methods, Multiscale Modelling Techniques, etc.
2026-10-23
2026-08-17
2026-06-18
Abstract—On the way to Industry 4.0, the digitization, networking and automation of industrial plants are the core challenges for manufacturing companies. These developments are accompanied by a rapid increase in the amount of available data, which often remains unused, resulting in waste in the sense of lean production. One reason for this is the great effort involved in the subsequent analysis of large amounts of data. In this context, predictive maintenance is a promising means of benefiting from data in terms of early wear prediction. In order to implement predictive maintenance, the approach presented here uses machine learning methods to generate a model for wear and plant status detection and, based on this, an algorithm for wear prediction. Only current signatures of production facilities are used for this. These signatures are available in every electrical system, have a high information content and can be measured with minimal effort and expense. Following the CRISP-DM methodology, a short-time Fourier transform is applied to the continuously acquired current signatures in order to extract features. In the modeling phase, recurrent neural networks are trained with these features. To create the right conditions, the current signatures are generated with a test setup for wear simulation, which is also used for the evaluation and verification of the developed models and algorithms. Especially in the area of critical wear, the trained recurrent neural network models provide correct classifications with an accuracy of over 95 percent. The developed algorithm for predictive maintenance therefore delivers reliable wear forecasts so that maintenance can be planned at an early stage. Finally, the models and algorithms are implemented and tested in a developed embedded system to perform wear detection and prediction at the machines edge in almost real-time.