| As a renewable and clean energy,wind energy is more and more popular.In recent years,with the development of wind power technology,the utilization rate of wind energy is also greatly increased.However,due to the harsh environment and complex structure,the operation and maintenance cost of wind turbine is much higher than that of traditional power generation.Therefore,it is of great significance to monitor the condition of the fan by means of condition monitoring,diagnosis and prediction to reduce the cost of operation and maintenance and ensure the normal operation of the fan.Therefore,this paper takes the transmission chain with high failure rate as the research object,and studies the composite diagnosis method based on multi-dimensional feature extraction of vibration signal.(1)A brief description of the structural characteristics and working principles of wind turbines is given.The rolling bearings and gears in the transmission chain are the key research objects,and their main structures,failure forms,and causes of failure are introduced to provide a theoretical basis for the realization of fault diagnosis.(2)Aiming at the problem that the working condition of wind turbine drive chain is complex and the fault features are submerged in noise signals,a one-dimensional timedomain feature extraction and fault diagnosis algorithm based on multi-scale mathematical morphology and correlation analysis is proposed.Firstly,a new type of "W" structural element,which is more similar to the fault vibration signal,is proposed to get more fault feature information;Then,the adaptive peak energy algorithm is introduced to determine the morphological scale and element parameters;Finally,the correlation analysis is used to identify the fault.The experimental results show that the algorithm has a good diagnosis effect for most faults and a certain generalization ability,but the description of some complex faults only through one-dimensional time domain features is not comprehensive,which leads to high similarity between the target fault and other faults and poor anti-interference ability.(3)In order to solve the problem that one-dimensional time-domain feature is not comprehensive and easy to be interfered,a fault diagnosis algorithm based on twodimensional gray image feature extraction is proposed.Firstly,one-dimensional vibration signal is transformed into two-dimensional gray image;Secondly,the local binary pattern(LBP)is used to extract the local texture features of the gray image,and the gray gradient co-occurrence matrix(GGCM)is used to extract the global features of the gray image;Finally,support vector machine is used for fault diagnosis.The experimental results show that the description of fault features is more comprehensive,the extracted features have large discrimination and strong anti-interference ability,but its generalization ability is poor.(4)In order to better complement the advantages of the above methods and achieve more reliable fault diagnosis,a composite diagnosis method based on Improved D-S evidence theory is proposed.Firstly,the diagnosis results of the two single methods are presented in the form of probability,which is used as the basic confidence probability(BPA)value of D-S evidence theory;Then,a new conflict measure operator non coincidence degree conflict operator is introduced to fuse the BPA of the two methods,so as to improve the accuracy of the algorithm fusion,and finally get the composite diagnosis results.Finally,the data of Case Western Reserve University bearing center,transmission chain fault simulation experiment platform and actual fan operation data are used to verify the algorithm,which shows the effectiveness and advantages of the composite diagnosis algorithm. |
