Fault Detection of Organic Heat Carrier Boilers Based on SeqGAN-CNN

Abstract

Abstract Organic heat carrier boilers are extensively used in industrial applications. However, the low accuracy of current fault detection methods, which primarily rely on statistical analysis and traditional machine learning methods, has become a significant challenge. This paper proposes a SeqGAN-CNN-based approach for detecting faults in organic heat carrier boilers. Firstly, SeqGAN is used to augment the dataset, solving the problem of limited fault sample data for training the CNN model. Then, the differential encoding method is utilized to transform one-dimensional monitoring data into two-dimensional grayscale images, resolving the issue that the CNN model’s input is generally multi-dimensional images while this study’s dataset is one-dimensional. Finally, the CNN model is trained and tested, and a simulation comparison is conducted with common industrial boiler fault detection methods. The experimental results demonstrate that the proposed method achieves higher accuracy, indicating that this method is accurate and effective for detecting faults in organic heat carrier boilers.