Fault Diagnosis of Chemical Process System Based on 1DKACNN-BiGRU-ECA

With the development trend of increasingly large and complex modern industrial systems, the volume of chemical process data is growing exponentially. Deep learning techniques can automatically extract essential features from historical data, demonstrating significant potential in chemical fault diagnosis. In this paper, we propose the 1DKACNN-BiGRU-ECA model for chemical process fault diagnosis by integrating KolmogorovArnold Networks (KAN) and one-dimensional Convolutional Neural Networks (CNN) for the first time. The model achieves an average accuracy of 98.85% using the Tennessee Eastman Process dataset. All evaluation metrics indicate superior performance compared to other models. To demonstrate the model’s effectiveness in real-world production environments. This verifies the model’s excellent generalization performance and efficacy in chemical system fault diagnosis.