基于CNN-Transformer双流融合模型的集成电路故障诊断方法研究OA

With the development trend of high integration and high density of integrated circuits,their internal structures have become increasingly complex.Traditional fault diagnosis models have been unable to meet the needs of efficient and accurate fault diagnosis.To this end,a dual-flow fusion method for integrated circuit fault diagnosis using convolutional neural networks(CNN)and transformers is proposed.This method takes advantage of the local feature extraction ability of the CNN model to accurately extract spatial local features from integrated circuit fault data;at the same time,the self-attention mechanism of the Transformer model extracts long-distance dependencies hidden in the data to deeply mine global features in integrated circuit fault data.By designing a reasonably designed feature fusion module,the features extracted from the two models are effectively fused,making full use of the complementarity of the two models.In order to test the effectiveness of the proposed method,an LM324 operational amplifier was used as a test chip,experimental data was built based on the self-built IC-Fault-2023 dataset,a dedicated test platform was established to carry out experiments,and the proposed method was compared with CNN models,Transformer models,traditional machine learning algorithms.Experimental results show that the fault diagnosis accuracy rate of this fusion model reaches 96.87%,which is 4.72%higher than that of a single CNN model,and the single inference time is only 8.3 ms,meeting the real-time requirements;its recall rate and F1 value are both better than the comparative model,effectively improving the integrated circuit fault diagnosis effect and providing reference for the fault diagnosis of complex integrated circuits.