Defect detection and identification for aircraft cable insulation layer based on deep forest.

Due to the special internal environment of aircraft, cable damage is inevitable, which usually starts from insulation layer defects and may cause major economic losses, and even seriously threaten the life of people on board. According to the above issue, a defect detection system for aircraft cable insulation layer based on ultrasonic guided waves (UGWs) is built in this paper. In order to ensure the complete coupling between the sensor and the insulation layer of the aircraft cable, the macro fiber composite (MFC) flexible sensor is employed in the system. Both simulation and experimental results show that this method can simultaneously monitor four different types (abrasion, cut, semi-stripping and full-stripping) of cable insulation layer defects at different locations on the same cable. The reflection signals of different types of defects are extremely similar and difficult to distinguish directly. In this paper, a classification method for four types of defects based on the deep forest method is proposed. This method requires a small sample size, and the classification performance is not affected by network structure and parameters. The recognition accuracy can reach 100%, avoiding the problem of traditional deep learning classification relying on a large number of samples and requiring parameter adjustment. The proposed method in this paper is proven to be able to effectively carry out online monitoring and accurately classify defect types, which has guiding significance for aircraft maintenance personnel to take corresponding measures in time.