基于MobileNet的海流透平机叶片附着物图像诊断

MobileNet-based Diagnosis for Image of Tidal Stream Turbine Blades’Attachments

基于电信号的故障诊断方法适用于海流透平机叶片不平衡故障,却不适用于平衡故障。为解决基于电信号的方法无法对平衡故障进行诊断这一问题,利用轻量级神经网络对不同故障类型的图像进行分类,从而实现海流透平机叶片故障诊断,并且该方法可以减小参数量,降低计算设备的硬件要求。首先,使用绳子模拟附着物,缠绕于海流透平机叶片以模拟海底生物附着过程;然后,采集不同附着程度的图像;其次,对图像数据进行筛选、分类,组建数据集;最后,利用MobileNet对不同附着程度的海流透平机叶片进行诊断。实验结果表明该方法具有良好的效果。

The fault diagnosis method based on electrical signals is suitable for imbalance faults of tidal stream turbine’s blades,but not suitable for balanced faults.In order to solve the problem that the method based on electrical signals can not diagnose balanced faults,the light-weight neural network is adopted to classify the images of different fault types,so as to realize the fault diagnosis of tidal stream turbine’s blades.The method achieves reduction of the amount of parameters and reduce hardware requirements for computing devices.Firstly,ropes are used to simulate attachments and wrap around the blades of the tidal stream turbine to simulate the process of being attached by sea organisms.Then,the images of the tidal stream turbine with different degrees of attachment are collected.Secondly,the image data are screened and classified to form data sets.Finally,MobileNet is adopted to diagnose the faults of tidal stream turbine’s blades.The experimental results show that the method has a good effect.