Study on defect detection technology of thermistor wire solder joints in spacecraft based on infrared thermography

Addressing the shortcomings of manual visual inspection, such as poor accuracy, inconsistency, and low efficiency, a detection method for thermistor wire solder joint defects in spacecraft based on active infrared thermography is proposed, and an automated detection system has been developed. The Yolov5l neural network model is utilized to conduct training and testing for defect detection based on infrared images of solder joints. Furthermore, the detection system is optimized by introducing the K-means algorithm and ECA model, achieving good results in detecting common solder joint defects. Experimental results show that the average detection time for a single solder joint using this detection system is 6 seconds, and the average detection accuracy for four types of defects, including uneven thickness, notches, pits, and contamination, reaches 92.3%. This research can provide a reference for the detection of other manually soldered joints.