Abstract:Tool wear prediction is essential for increasing production efficiency, improving product quality and reducing manufacturing costs. However, most of the existing studies are either pure experimental or machine learning-assisted (ML) research, which requires numerous expensive and time-consuming wear tests to prepare a sufficiently rich dataset. This limitation hinders the application of ML algorithms in real life monitoring systems, restricting their scope to only academic research. To bridge the gap between research and industry, in this study a novel sequential physics-informed machine learning (PIML) model was developed to predict tool wear with regards to cutting forces, machining parameters and tool geometry. The PIML sequentially integrated the analytical wear-included force model with ML algorithms such as least-squares boosting, random forest and support vector machine. In this respect, initially a thermo-mechanical turning model was developed to calculate the cutting forces by considering the effect of flank wear and edge forces. The accuracy of this model was then improved through the PIML model, achieving 97 % accuracy on the entire training dataset and 94 % accuracy on the unseen test dataset. This facilitated the creation of efficient and reliable training data for another complementary reverse ML model to predict wear length based on cutting forces and machining parameters. Also, the relative significance of different input parameters on the model's predictions was quantified using the Shapley value algorithm, which calculated each feature's contribution to flank wear. According to the obtained results, sequential integration of the mechanistic model with the ML algorithm not only enhanced the prediction accuracy of the model remarkably, but also reduced the need for numerous experimental wear tests. In addition to Steel 1050, the proposed PIML model accurately predicted wear length for Ti6Al4V superalloy, confirming its effectiveness and robustness across various workpiece materials and cutting tools with different geometrical features. These findings indicate the model's versatility and practical applicability in real-world industrial contexts. This highlights the importance of PIML implementation in predictive modeling for enhanced accuracy and reliability, particularly in complex scenarios involving flank wear prediction.

Tool Wear Estimation with a Data-Driven Physics Coupling Approach

[Pancreatitis in a patient with Crohn disease treated with mesalazine and azathioprine].

Hybrid data-driven physics-based model fusion framework for tool wear prediction

A generic tool wear model and its application to force modeling and wear monitoring in high speed milling

A Physics-Assisted Online Learning Method for Tool Wear Prediction

Tool wear estimation and life prognostics in milling: Model extension and generalization

Tool wear prediction based on a fusion model of data-driven and physical models in the milling process

Physics-informed tool wear prediction in turning process: A thermo-mechanical wear-included force model integrated with machine learning

A hybrid tool wear prediction model based on JDA

Research on the Milling Tool Wear and Life Prediction by Establishing an Integrated Predictive Model

Wear Prediction of Tool Based on Modal Decomposition and MCNN-BiLSTM

A milling tool wear predicting method with processing generalization capability

In-process tool condition forecasting based on a deep learning method

Tool Wear Prediction Based on Multi-information Fusion and Genetic Algorithm-optimized Gaussian Process Regression in Milling

Tool wear rate prediction in ultrasonic vibration-assisted milling

Deep Learning Based Tool Wear Estimation Considering Cutting Conditions

Deep Learning Approach for Enhanced Transferability and Learning Capacity in Tool Wear Estimation

Data-model linkage prediction of tool remaining useful life based on deep feature fusion and Wiener process

Tool wear prediction method based on symmetrized dot pattern and multi-covariance Gaussian process regression

Remaining Useful Life Prediction Based on Spindle Load and Cutting Process Parameters in Machining

In-process Tool Wear Prediction System Based on Machine Learning Techniques and Force Analysis