Research On Condition Evaluation And Remaining Life Prediction Of Key Parts Of Machine Tool Based On Machine Learning

At present,the manufacturing industry is developing towards the direction of network and intelligence.Machine tool as the basic manufacturing equipment,its normal operation is closely related to the quality of processed products and the on-time delivery of enterprise orders.Under the background of computational intelligence represented by machine learning,the autonomous evaluation of machine tool status and the accurate prediction of remaining life can provide reliable maintenance decisions for enterprises,so as to reduce the maintenance cost caused by emergency maintenance due to machine failure,and enhance the core competitiveness of products.Therefore,in this paper,the key parts of complex machine tools are taken as the research object,and the machine learning method is used to study the condition evaluation and remaining life prediction of key parts of machine tool.Firstly,aiming at the identification of key parts that affect the normal operation of the machine tool,Failure Mode Effect and Criticality Analysis is used to determine that the spindle system fault has the greatest impact on the normal operation of the machine tool subsystems.On this basis,FMECA of the spindle system is carried out,and the bearing fault is identified as the biggest influencing factor through the degree of harm,and the bearing is determined as the key part of the machine tool for research.Secondly,in the feature extraction stage,it is time-consuming and labor-consuming,and relies too much on expert experience,which leads to the slow speed and inaccurate results of spindle bearing condition evaluation.A spindle bearing condition evaluation method based on Convolutional Neural Networks optimized by Sine Cosine Algorithm is proposed,It also realizes the independent selection of learning rate parameters in CNN.Through the CWRU data set,it is verified that the SCA-CNN model can realize the scientific evaluation of bearings.Then,aiming at the problems of complex prediction process and low accuracy of traditional model method for spindle bearing,the remaining life prediction model based on Convolutional Neural Networks and Gated Recurrent Units is constructed.The strong data processing ability of Convolutional Neural Networks is used for feature extraction,and Gated Recurrent Units is used for regression analysis.On this basis,a weighted average noise reduction method is proposed to reduce the deviation error between the prediction curve and the actual life curve,and the effectiveness of the proposed method is verified by PHM2012 data set.Finally,the development of intelligent monitoring system for key parts of machine tool is realized by using JAVA and Front-end development.The functional subsystems such as condition monitoring module,life prediction module and maintenance module are designed,and the feasibility of the model and method proposed in this paper is verified by case operation.