Research On Key Technology Of The Ball Screw Performance Degradation Assessment

Manufacturing industry is the mainstay of the national economy.More specifically,highend equipment manufacturing is the pillar industry related to national development and social progress.The ball screw is an indispensable mechanical transmission device in manufacturing equipment such as CNC machine tools.The ball and raceway of ball screw are subjected to continuous contact stress on the surface during the machining process,resulting in irreversible damage including fatigue,wear,and even corrosion,leading to the performance of ball screw gradually degrading to failure.It will directly affect the reliability of manufacturing equipment and even lead to safety accidents if the degradation of ball screw is not monitored.Therefore,accurate degradation assessment of the ball screw is not only very essential for improving the operational reliability and machining quality of manufacturing equipment but also very important for manufacturing intelligence and automation.However,due to the complex service environment and multiple working conditions,the measured data are affected by both degradation and working conditions,making the research on the degradation assessment of ball screw becomes more challenging.Studying the degradation assessment methods of the ball screw under various influencing factors such as multiple working conditions or missing labels has become the core problem of health management.Therefore,some crucial issues about the degradation assessment of the ball screw are studied in this paper.Details are as follows.(1)The accelerated degradation test bench of the ball screw is designed and constructed,which can simulate the whole degradation process of ball screw from new to failure in a relatively short time.The run-to-failure test of ball screw is designed and carried out.Multiple accelerometers are mounted on different positions of test bench to collect degradation data of the ball screw under different working conditions.The collected data is the data basis for the following degradation assessment researches.(2)For the degradation recognition of ball screw under multiple working conditions,a recognition method based on deep parameter transfer learning is proposed.Deep residual network Res Net-18,which has been trained on Image Net,is transferred as a source model.By initializing the target recognition model with the network structure and parameters of the source model,a deep transfer learning model with better performance than randomly initialized network is established.The recognition results indicate that the proposed deep parameter transfer learning method is able to eliminate the interference of working conditions and accurately recognize the degradation of ball screw under complex and multiple working conditions.(3)Aiming at recognizing the degradation of ball screw with unlabeled data,a method based on improved deep domain adaptive network is proposed.In this work,a onedimensional convolutional neural network(CNN)is constructed as basic recognition model,while multi-layer domain adaptation module and domain classifier adaptation module are added to the CNN to construct the improved deep domain adaptive network.The multi-layer domain adaptation module and domain classifier adaptation module facilitate the onedimensional CNN to learn domain invariant features,thus completing transfer learning from the source domain to the target domain.The test results indicate that the proposed method is able to accurately recognize the degradation of the ball screw under new working condition with unlabeled data by transferring the existing knowledge of other working conditions.(4)Owing to the requirement of remaining useful life(RUL)prediction in the performance degradation stage of ball screw,a RUL prediction method is proposed.Firstly,the degradation model of ball screw is established by analyzing its degradation failure mechanism,which is the exponential wiener process model.Then,the health indicator of ball screw is constructed based on the reconstruction error of the denoising autoencoder,which has excellent monotonicity and tendency.Finally,particle filtering is utilized to incorporate the degradation model and health indicator information for parameter update and RUL prediction of ball screw.Results show that the proposed method can predict the RUL of ball screw with high accuracy.(5)Aiming at the lack of industrial application research on the degradation assessment technology of ball screw,a practical system for degradation assessment of ball screw is established.The system is mainly composed of data acquisition,data transmission,the server side and the user side.Function modules of degradation assessment and RUL prediction are developed based on the proposed methods.The system can identify the degradation state and predict the RUL of the ball screw,and provide corresponding maintenance suggestions for users.The constructed practical system is of great significance to improve the operation safety of the ball screw on manufacturing equipment.