基于深度强化学习变分模态分解的风机叶片故障诊断方法OA

To address the problem that wind turbine blades are prone to defects such as cracks and wear under complex operation conditions,and to overcome the limitations of traditional vibration signal analysis methods,this paper proposes a fault diagnosis method for wind turbine blades based on adaptive variational mode decomposition optimized by deep reinforcement learning.In this method,deep reinforcement learning is introduced to construct an adaptive optimization framework for variational mode decomposition parameters.The dynamic optimization of the mode number and penalty factor is realized through a discrete-continuous hybrid action space and a dual-objective reward function constrained by the average envelope entropy and the variance of envelope entropy.On this basis,sensitive modes are screened according to envelope entropy,the time-domain,frequency-domain and entropy features are fused,and the blade fault classification is completed using a support vector machine.The research results show that the proposed method can reduce the average envelope entropy of variational mode decomposition to 2.21,control its variance within the range of 0.21-0.24,and achieve a recognition accuracy of blade faults as high as 96.4％.The research conclusions provide a reference for improving the accuracy and reliability of fault diagnosis for wind turbine blades.