Model And Simulation Research On Intelligent Fault Diagnosis Of Medium-low Pressure Gas Regulators

In the process of accelerating urbanization construction in China,the demand for natural gas in small towns increases rapidly,and problems in natural gas construction become increasingly prominent.Therefore,improving the infrastructure construction of urban gas transmission and distribution terminals is the top priority of development.The gas pressure regulating equipment is the most important link of the gas transmission and distribution terminal,and it is also the most critical link to ensure the users to use gas smoothly.With the development of digital technology,the gas pipe network system is more and m ore intelligent,and it is gradually applied to the management of gas pressure regulating equipment.Outlet pressure is not only the regulating object of the gas regulator,but also the important index to monitor the running state of the regulator.When the operating state of the gas regulator changes,the outlet pressure will fluctuate with the passage of time,resulting in the change of energy in different frequency bands.In view of this feature,this paper selects Long Short-Term Memory and Deep Neural Networks that can well process the time series data to establish the fault diagnosis model of medium-low pressure gas regulator,and makes a comparison with the more commonly used Support Vector Machine diagnosis model at present.At the same time,the simulation platform is used to simulate the fault state and observe the variation of outlet pressure under different fault states,so as to make a qualitative analysis of the fault.The main contents are as follows:Firstly,this paper introduces the operation mechanism,evaluation index and common faults of gas regulator in detail.The mathematical model of gas regulator is established by means of equation of mechanical motion,gas state equation and orifice flow equation.Through simulation platform,the mathematical model is transformed into a physical model,and its dynamic characteristics and failure state are simulated.The maximum deviation and transition time were used to evaluate and analyze its dynamic characteristics.Different faults are distinguished b y the characteristics of outlet pressure.The results show that the pressure area of the membrane and the spring stiffness can be increased to improve the performance of the regulator.When the seat diameter is smaller,the regulator has better dynamic performance,while the volume of the low-pressure chamber has little influence on the dynamic performance.The outlet pressure of the gas regulator fluctuates in different pressure range when the signal tube is blocked,the membrane is aged and the spring is tired.When the corresponding pressure fluctuation occurs in the actual operation,this feature can be used to check whether it is the corresponding fault first.To solve the problem of backward fault diagnosis of gas regulator,the surge,ice blocking and normal operation status data of RTZ-50/0.4FQ gas regulator in a certain gas regulator station were taken as the basis.Firstly,Empirical Mode Decomposition method was used to extract the features,and the IMF component energy moment obtained after decomposition was taken as the input data of the fault diagnosis model.Based on this,the fault diagnosis model of Long Short-Term Memory,Deep Neural Networks and Support Vector Machine are established,and the accuracy,precision,recall rate,F1-score and Kappa coefficient and other indicators were used to evaluate the model.By comparison,the performance of Long Short-Term Memory fault diagnosis model is the best,followed by the Deep Neural Networks,and the Support Vector Machine model is slightly lower than the above two models.It shows that the deep learning algorithm is feasible in the field of fault diagnosis of gas regulator and has certain advantages in the diagnosis effect.