Fan Blade Crack Detection Algorithm Based on Multi-Scale Feature Fusion

Abstract

With the rapid development of social economy, energy consumption is growing tremendously so green energy such as wind energy has become widely used, thus promoting the construction of wind turbines. Due to the long-term use of the electro-mechanical unit, the traditional maintenance cost is too high. In order to quickly and accurately detect and maintain the fan blades, based on the intelligent big data from the environment, we propose the convolutional neural network model to solve the problem of low recognition rate due to the lack of feature extraction in the fan blade crack image, and the long short-term memory network (Long Short-Term Memory, LSTM) convolutional neural network model, and the dimensionality reduction of the captured image data, which is beneficial to improve the recognition rate of the picture and reduce the loss rate of the picture through the detection model’s suitable recognition of complex background problems such as target occlusion and overlap. Using LSTM to extract the global context module can effectively improve the target detection accuracy. When this part is added, the detection accuracy will increase by about 3% to 7%. The image position can be accurately captured and the recognition rate is greatly improved through the optimized convolutional neural network, which can provide a reference for future research in other fields.

Keywords

• SCADA database
• neural network
• LSTM feature extraction
• wind turbine blades