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 learningassisted (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 acr
