Vibration-based fault diagnosis of the natural gas compressor using adaptive stochastic resonance realized by Generative Adversarial Networks

Abstract The compressor as an important energy transmission equipment is widely used in the natural gas pipelines to pressurize natural gas, which is prone to fail due to the characteristics of high pressure, flammability and corrosion of the natural gas. The faulty compressors may cause natural gas leakage, transmission interruption, and even explosion. Vibration monitoring and fault diagnosis of the natural gas compressor is an effective technology to ensure its security. Due to the strong noise interference, the traditional vibration diagnostic methods are difficult to obtain better effects. Stochastic resonance (SR) is a useful technology that can make full use of the noise components contained in the vibration signals to enhance the weak fault features to complete fault diagnosis. However, the applications of SR in fault diagnosis are faced with many problems, such as the difficulty of parameter selection and the need for intensive data. To solve them, an adaptive generation model of SR parameters is established by using Generative Adversarial Networks (GAN), an adaptive SR-GAN method is accordingly proposed for completing the fault diagnosis of a natural gas compressor. The simulation experiment and field data analysis are adopted to verify the effectiveness of the proposed method. The results indicate that the proposed method can improve diagnostic accuracy by 2.07% compared to the traditional adaptive SR method realized by multilayer perceptron neural network, and improve the signal-to-noise ratio by 2.25 dB at most compared with the other three methods. Moreover, the proposed method still has higher accuracy under the condition of small sample sizes. The proposed method can expand the sample when the sample size is smaller, so that the model can contain completely the operating conditions to improve the accuracy of fault diagnosis. It provides a new idea for the vibration monitoring of natural gas compressors.