Fault Diagnosis of Circuit Board Chip Based on Infrared Temperature

To solve the problems of defect localization difficulty and poor diagnostic effect of the current circuit board chip-level fault diagnosis methods, this paper proposes a circuit board chip fault diagnosis model composed of one-dimensional convolutional network (1D-CNN) and two-dimensional convolutional network (2D-CNN) in parallel based on the infrared temperature sequence. Firstly, the temperature data of the circuit board chips under different failure modes is collected using an infrared thermal camera. Then, 1D-CNN and 2D-CNN networks are used to extract the temporal and spatial features of the data, respectively, for fault information mining. Finally, the defect characteristics obtained from the dual channels are assigned weights and fused using the attention mechanism, before being fed into Softmax for defect classification. The experiment employed a power supply circuit board for conducting reliability analysis. The experimental results demonstrate that the algorithm analysed in this paper has a diagnostic accuracy of 98.73%. Furthermore, the algorithm outperforms the comparison algorithm in terms of stability and reliability.