Rolling Bearing Diagnosis Research Of CNC Machine Tools Based On Multi Source Information Fusion

Rolling bearing is one of the main rotating parts in mechanical equipment, but also one of the main sources of mechanical equipment fault. Therefore, research on fault mechanism and diagnosis method of rolling bearing has important theoretical and practical significance to detect and eliminate the faults of rolling bearings, ensure the normal operation of machinery and equipment and improve production efficiency and product quality.In this paper, taking the rolling bearing of NC workbench ball screw shaft as the research object,analyzes the mechanism and common failure forms of the faults of rolling bearings,fault diagnosis method and its adaptability. Considering the randomness of the rolling bearing fault,sensor measurement error, the complexity of the environment and other factors will cause the fault information is random and uncertain characteristics, using the method of external sensors and extracting machine servo information obtained multi-source fault information of rolling bearing, With the help of corresponding signal processing and information fusion method to achieve the data level, feature level and decision level fusion of multi-source information,build a ault diagnosis system of NC machine tool rolling bearing based on multi-source uncertain information fusion. The main research contents are as follows:Starting with the structure features of the rolling bearing, the rolling bearing fault mechanism,failure mode and fault detection method and its adaptability, selects the fault detection method, the type of external sensors and sensor installation of measuring point;according to the relationship between the internal servo information machine with rolling bearing fault state, select and extract the internal part of machine tool servo parameters;build a multi-channel data acquisition system based on NIPCI-6143 multi- channel data acquisition card and Lab VIEW software platform.Considering the nonlinear and non-stationary of rolling bearing fault signals, introduced statistical theory and wavelet packet decomposition method to rolling bearing fault signal analysis in the time domain, frequency domain and time-frequency domain analysis and extract the relevant characteristic values. To effectively eliminate the redundant information, simplify the fault diagnosis process and improve the efficiency of fault diagnosis, selected The characteristic value by rough sets theory, established the training set and test set of fault characteristic value.Introducing SVM method, combination of the data acquisition, feature extraction and selection and pattern recognition, set up a fault diagnosis system of rolling bearing based on multi-source information fusion,realize data level and feature level fusion; In order to optimize the effect of SVM diagnosis, genetic algorithm is introduced to the SVM network parameters were optimized.Considering the randomness of the rolling bearing fault information, and uncertain characteristics, Proposed the method of multi-source information fusion based on cloud model and evidence theory. Study on the method of obtaining parameters of cloud model, obtain evidence matrix through the cloud model. A method is proposed to improve the D-S evidence theory of evidence fusion after weight redistribution fusion. Combination of cloud model and D-S theory of evidence, to achieve recognition of rolling bearing fault model, experiments show that the method is feasible.