Wind Turbine Blades Defect Detection Based on Acoustical Features and Small Sample

In the current study, wind turbine blades(WTBs) defect detection is based on a large amount of data, which caused enormous human and material resources. To address this issue, we proposed a novel acoustic detection method based on small sample. To verify the performance of the proposed method, we went to two wind farms to acquire WTBs signals and analyzed the influence of directions and heights on data acquisition. Besides, we compared MAML-BPNN and Tradition-BPNN in different data-size. The experimental results show that our method can produce reliable results when data-size is only 50. The accuracy reaches 94.1% which the Tradition-BPNN is just 85%. Our method has the advantages of fast training speed and automatic search for global optimal solution. We also found that the performance of MAML-BPNN is better with the rise of data-size, indicating the model is universal in the case of large amounts of data