Diagnostic of the State of the Cutting Tool of Metal-Cutting Machines using Bidirectional Recurrent Neural Networks with a Long Short-Term Memory *

摘要

The paper presents a method for diagnosing the condition of a cutting tool of metal-cutting machines, which allows determining the amount of tool wear according to information received from force and vibration sensors. To fix the dependence of the amount of wear on the values of vibration and cutting forces, it is proposed to use a diagnostic model based on bidirectional recurrent neural networks with a long short-term memory. The article presents and describes the architecture of such neural networks. We have developed several diagnostic models based on various neural network architectures and conducted a comparative analysis of the accuracy of the obtained models.