Research On Service Performance Monitoring And Fault Diagnosis Of Ball Screw Pair

CNC machine tool has become the strategy equipment and foundation of manufacturing industry,which directly embodies the basic equipment manufacturing technology and the comprehensive economic strength of a country.As an important transmission part of the feed system of CNC,the ball screw has a direct influence on its quality.Along with the increase of the service time,three common failures which are preload loss,pitting and poor lubrication are going to be generated,leading to the decline of the stability,rigidity and vibration resistance of the feed system,and finally resulting in huge economic losses.Therefore,the real time monitoring of the health status and early-warning and diagnosis of ball screw of CNC is of great significance.In this paper,the failure forms and corresponding solutions of ball screw both under the conditions of conventional and high overload are summarized.The relationships between the three failures:preload loss,pitting and poor lubrication and their vibration signals are studied through experiments to obtain the feature vectors that can characterize the failure modes.A signal processing algorithm combining EMD with MSE is proposed for preload loss failure,the HHT transform and the wavelet packet transform energy extraction algorithm are used to analysis pitting failure and poor lubrication failure.BP neural network optimized by genetic algorithm is built in view of the shortcoming of the traditional network which is easy to fall into the local minima.The network is trained by the known fault samples and then the pattern recognition of the unknown fault samples is carried out.The recognition accuracy is up to 90%.Finally,the fault diagnosis system of the ball screw is developed.A DAQ system for vibration signal is built based on the Ball Screw Friction Torque Test Bench,a fault diagnosis program which integrates signal collection,signal analysis and fault diagnosis is written based on MATLAB GUI which significantly improve the efficiency and application range of the fault diagnosis method this paper proposed.