基于DHNN的风电机组齿轮箱故障诊断

Fault Diagnosis of Wind Turbines Gearbox Based on DHNN

运用具有联想记忆功能的离散Hopfield神经网络(DHNN)对风电机组齿轮箱的故障进行诊断,选用时域和频域的5个故障特征指标作为评价因子,利用MATLAB工具箱建立一个可以对风电机组齿轮箱的3种故障进行诊断的DHNN模型,并将该模型用于北方某风电场的实测数据的故障诊断,验证模型的泛化能力.仿真结果表明,DHNN的诊断结果准确率高、收敛速度快,具有很好的实用性.

Discrete Hopfield neural network(DHNN)with an associative memory function was used to diagnose gearbox faults of wind turbines. Five fault feature indexes in domains of both time and frequency are treated as evaluation factors. A DHNN model capable of diagnosing three types of gearbox faults of wind turbines was established using MATLAB toolbox. And the model was applied to the diagnosis of the real data collected from a wind farm in northern China to test its generalization ability. The simulation results show that DHNN has high diagnosis accuracy ,fast convergence speed as well as high practicality.