基于优化双向长短时记忆网络的刀具磨损状态识别

Tool Wear State Recognition Based on Optimization Bidirectional Long Short Term Memory Network

准确可靠地对刀具磨损状态进行监测和识别,有助于保证加工质量和加工效率。为提高刀具磨损状态识别精度,提出一种优化双向长短时记忆网络(NGO-BiLSTM)的刀具磨损状态识别新方法。NGO-BiLSTM核心思想就是通过北方苍鹰优化算法(NGO)对BiLSTM网络超参数进行自适应优化选取,从而解决BiLSTM网络超参数取值不同导致识别结果不稳定这一问题,进而提高BiLSTM的识别性能。通过刀具磨损状态识别实例对所提方法的有效性进行验证,结果表明:所提方法提高了识别精度,在5种评价指标上也是优于其它几种方法。

Accurate and reliable monitoring and recognition of the tool wear state are helpful to ensure machining quality and efficiency.In order to improve the accuracy of tool wear state recognition,a new method of tool wear state recognition based on optimization bidirectional long short term memory(NGO-BiLSTM)is proposed which can solve the problem of unstable recognition results caused by different values of the BiLSTM network super parameters.The core idea of NGO-BiLSTM is to carry out adaptive optimization selection of the BiLSTM network super parameters through northern goshawk optimization(NGO)algorithm,so as to improve the recognition performance of BiLSTM.The effectiveness of the proposed method is verified by an example of tool wear state identification.The results show that the proposed method improves the identification accuracy and is superior to other methods in five evaluation indices.