Physics informed deep learning in machining equipment

Description

Modern factories want reliable tool-wear estimates and remaining tool life under changing cutting conditions. Project partner Sirris has collected a dataset [1] of cutting-edge images plus accelerometer, microphone, and three-axis force signals from CNC milling and established baselines using basic machine learning models. The project will investigate whether physics-informed machine learning (ML) can be leveraged to obtain results that outperform established baselines.

Recent approaches in machine wear modelling rely on deep learning over images and sensor streams for wear estimation, while machining practice has long used Taylor’s tool-life principle [2] to capture how operating conditions drive wear and life. Physics-informed machine learning [3] shows that adding lightweight physical constraints or priors can improve generalization and trust in scientific and industrial settings.

How

During the project you will develop physics-informed ML models under the guidance of Sirris domain experts. The goal of the project is to improve on existing baselines by developing multimodal models that integrate physics principles into their modelling process.

Prerequisites

1. Solid understanding of Machine Learning and Deep Learning fundamentals
2. High-level coding skills in Python
3. Interest in physics informed modelling and/or mechanical engineering

• [1] De Pauw, Lars, Tom Jacobs, and Toon Goedemé. "MATWI: A multimodal automatic tool wear inspection dataset and baseline algorithms." International Conference on Computer Vision Systems. Cham: Springer Nature Switzerland, 2023.
• [2] Lee, Yong Ju, and Hae-Sung Yoon. "Modeling of cutting tool life with power consumption using Taylor’s equation." Journal of Mechanical Science and Technology 37.6 (2023): 3077-3085.
• [3] Cuomo, Salvatore, et al. "Scientific machine learning through physics–informed neural networks: Where we are and what’s next." Journal of Scientific Computing 92.3 (2022): 88