Fault Diagnosis And Life Prediction Of Fan Gearbox Based On CHMM

The rapid growth of the world economy and population has brought about a sharp increase in the energy gap.However,the traditional coal,oil,natural gas and other conventional non renewable energy will eventually be exhausted,and the concept of sustainable development is increasingly popular.The rational development of clean and high-quality new energy can reduce pollution,reduce the burden on the environment,and realize the harmonious coexistence of man and nature.In the process of developing new energy,wind energy has attracted much attention and got rapid development because of its advantages of non pollution,rich and wide distribution.Wind power generation has become one of the important strategic choices for the sustainable development of electric energy.With the development of intelligent wind power equipment in recent years,how to effectively detect and operate the fan has become an urgent problem in the wind power industry.The gearbox is an important component in the fan drive system,which has complex structure and high failure probability.Its maintenance is difficult,and takes a long time.The gearbox has always been the focus of the maintenance of the fan.In order to improve the reliability of maintenance,it is of great significance to study the fault diagnosis after the accident,the degradation evaluation of equipment performance in the production process,and the overall life prediction.The paper is based on the CHMM model to study the fault detection,performance degradation assessment and life prediction of wind turbine gearboxes,which has good theoretical research and engineering application value.This method can provide a reference for the maintenance of the equipment.At present,for vibration signals with large amount of information,non-stationary,and poor feature repeat-ability,such as wind turbine gearbox vibration signals,continuous hidden markov model is gradually entering researchers’ vision as a more mature analysis and processing method.However,due to the characteristics of this model tending to local maximums,there is still room for improvement in the accuracy of fault diagnosis based on this method.In order to improve the stability of the model and the correctness of the final result,a parameter training method based on the combination of artificial bee colony algorithm and Baum-Welch algorithm is proposed.The artificial bee colony algorithm is used to optimize the initial value of the model,combined with the training parameters of the Baum-Welch algorithm,to improve the accuracy and stability of model fault detection.To eliminate the interference caused by the change of working conditions on the feature extraction of degraded performance state of fan gearbox,a feature transformation method based on interference attribute projection is introduced.The transformed vectors are used as observations to establish a continuous hidden markov model under normal conditions,and the degree of deviation between the output probability of the model and the output probability of the normal state is used as the basis for evaluating the performance degradation of the gearbox.In order to establish a dynamic alarm system,it is proposed to use the limit error method to statistically process the output results to draw a hierarchical alarm line,which provides a reference for subsequent maintenance plan formulation.Support vector machine algorithm can be used to predict the lifetime of a gearbox by learning the characteristics of global optimal solution in small sample,non-linear and high dimension problems.To solve the problem of complex and blindness in parameter selection of support vector machine algorithm,artificial bee colony algorithm is introduced for parameter optimization in this paper.Considering the huge number of lifetime prediction samples and the sharp increase in the workload of feature extraction,the prediction model is established in segments based on the performance degradation assessment of the CHMM model,which reduces the workload and increases the accuracy of the prediction model.Finally,the feasibility of the improved method is verified through the gearbox life test.