Research On Rotating Machinery Operation State Recognition Method Based On Improved D-S Evidence Theory

As an emerging discipline applied in the fields of fault diagnosis,pattern recognition,image fusion and state recognition,decision fusion can obtain more reliable system decision results by integrating the redundancy and complementarity of various information sources.D-S Evidence theory is a classical algorithm for decision fusion,which can achieve good measurement and expression of characteristic information of uncertain systems.It has broad application prospects.In this paper,rolling bearings and motorized spindles are taken as the research objects.According to the mechanical vibration information,the research on the operation state recognition algorithm based on the improved D-S evidence theory is carried out.Aiming at the problems of traditional D-S evidence theory such as BPA acquisition,conflict,and computational complexity,corresponding improvement methods are proposed.The research contents of this paper are as follows:(1)Analyze the typical structure and operation status of rotating machinery components.The basic structure and fault status of typical rotating mechanical components such as bearings and electric spindles are described.According to the structural relationship of electric spindles and the transmission path of vibration signals,the multi-sensor installation position,experimental scheme and data acquisition scheme for the simulation experiment of the unbalance of the electric spindle rotor are determined.(2)Research on the algorithm of obtaining evidence body in D-S evidence theory.The methods of obtaining evidence body based on slope correlation degree,weighting coefficient,SVM model and SVDD model based on the concept of membership degree are discussed respectively to realize the dual fusion of feature level and decision level.Based on a specific example of bearing operation state identification,evidence is obtained to allocate the basic probability of each bearing operation state,and several methods are compared to verify that the evidence body acquisition algorithm based on SVDD model has a high accuracy.Combining the algorithm of obtaining evidence body from the validated PSO-SVDD model with an example of motorized spindle operation status recognition,the analysis found that sensors closer to the fault location have higher recognition accuracy,which is consistent with the actual situation.(3)Research on solutions to evidence conflicts.This paper analyzes the causes of evidence conflict and common solutions,discusses the evidence weight algorithm based on correlation coefficient,and uses it as a comparison algorithm.The Lance distance function and SAC similarity are introduced,and an evidence correction method based on conflict redistribution is proposed.This method considers both the conflict of evidence and the correlation of evidence.The results of bearing operation state recognition show that the results of the two algorithms’ multi evidence synthesis are both effective and reliable,and the newly proposed method has excellent fusion performance and better convergence effect.The improved D-S evidence theory is applied to the multi sensor fusion of the motorized spindle.Through the support and reliability of each sensor,the conflict between different sensors is resolved,and the recognition accuracy is improved.(4)Design the software of rotating machinery operation status recognition system to achieve intelligent data processing.Research approximate algorithms for D-S evidence theory.Aiming at the complexity of computation in evidence volume fusion,an approximation algorithm of focal element control based on Lance distance is proposed.Through specific examples,it is shown that this method can reduce the meaningless computation in the practical application of D-S evidence theory while ensuring the accuracy of decision-making.