Research on the <tex>$\beta{-}$</tex> VAE-DDRRN-BFA Tool Wear Detection Algorithm

With the advent of the industry 4.0 era, the manufacturing industry has become an integral part of the national economy. In machining processes, it is necessary to continuously cut metals using tools. During tool operation, wear occurs continuously. Over time, this wear gradually affects the quality of workpieces, and in severe cases, it may pose the risk of machine tool damage. In this context, tool wear detection has become an indispensable aspect of the manufacturing industry. However, at present, tool wear detection methods face several challenges, including complex model structures, insufficient detection accuracy, low efficiency, and poor generalization capabilities. These factors can impact the effectiveness of tool wear detection, leading to delayed tool replacement and increased industrial losses. This paper addresses the issue of poor generalization capability in existing deep models and proposes a tool wear detection algorithm based on the combination of $\beta{-}$ V AE (Beta Variational Autoencoder) and Deep Dense Recursive Residual Network (DDRRN). Additionally, a Bayesian Fusion Algorithm (BFA) module is integrated at the signal input end to enhance the reliability of tool wear detection. Finally, experimental validation is conducted using the PHM2010 public dataset. The results indicate that the β-VAE-DDRRN-BFA algorithm outperforms other classical methods in the reconstruction of tool wear, and the model's computational efficiency is higher than other methods.