Convolutional Neural Network Based Feature Fusion for Fault Diagnosis of Chemical Dynamic Equipment

Chemical power equipment operates under harsh conditions such as high temperatures, high pressures, or corrosive environments, making them susceptible to wear, fatigue, and corrosion. These issues can lead to performance deterioration or even catastrophic failures, posing significant risks to the safety of individuals, their property, and overall social stability. Addressing this critical concern, the study delves into the use of deep learning technologies for the safety diagnosis of chemical and dynamic equipment. A convolutional neural network-based diagnostic model is proposed, capable of extracting one-dimensional signal attributes and two-dimensional features from wavelet time-frequency diagrams. The model innovatively fuses these two feature types, markedly enhancing the accuracy of fault diagnosis in chemical dynamic equipment. Such an improvement in safety diagnostics is critical for preventing accidents and ensuring operational integrity. Furthermore, the intelligent design of the pooling and global average pooling layers substantially reduces the model's parameter count, mitigating the risk of overfitting. This strategic reduction optimizes the model's ability to generalize, solidifying its utility and reliability in real-world applications within the chemical industry.