Dynamic System Simulation Of Gas Turbine And Research On Fault Data Analysis

Gas turbine is a kind of complex thermodynamic system with strong nonlinear,time-varying,and multi-variable coupling.Establishing a perfect mathematical model,a reliable control system and a reasonable fault diagnosis strategy is important contents of gas turbine design.Most of existing gas turbine control system are verified based on the gas turbine mathematical model.The running state of gas turbine can be simulated well by the dynamic system composed of gas turbine model and control system.The design of fault diagnosis method also relies on the gas turbine dynamic system.For different types of gas turbines,the performance indexes are not the same.Therefore,the establishment of mathematical model,the design of control system and the design of fault diagnosis method need to be carried out separately.Signal filtering and threshold comparison are traditional fault diagnosis strategies for gas turbines.Due to the simple logic and deep integration with control system design,this strategy has been widely used.However,with the refinement of gas turbine operation and maintenance,such a simple strategy is no longer able to meet the demanding fault diagnosis requirements.Meanwhile,health management of the whole life cycle is regarded as an effective means to increase the service life of gas turbines.Hence,it is necessary to introduce advanced diagnosis algorithms and prediction methods.Taking a aeroderivative gas turbine as the research object,this paper focuses on the establishment of gas turbine model,control system design and intelligent fault diagnosis method.The main work is as follows:Firstly,combining with gas turbine volumetric method and engineering requirements,a gas turbine mathematical model was built.The classical NewtonRaphson method was used to solve the nonlinear equations.The simulation accuracy of the gas turbine model is verified by comparing with the real test data.The results show that: component-level module,starting module and the steady-state model accuracy basically achieve the verification indexes.Secondly,the function of the control system is divided and designed according to the control module of the gas turbine,the control module of auxiliary equipment,the fault diagnosis and independent protection module of the system.The VSV control module and the core speed control module were designed in detail.The robustness of the control system is verified by combining with the gas turbine mathematical model.Finally,based on gas turbine mathematical model and control system,a brief design idea and a time-space series fault data generation scheme for gas turbine damage propagation module are designed.Combined with the existing data,the classification and prediction algorithms for univariate and multivariate cases were designed from the perspective of time series analysis.The random forest,decision tree and support vector machine were set up for fault diagnosis on a given data set,and significant results were obtained on the dichotomy problem,but not on the multi-classification problem.LSTM model is established to carry out univariate time series analysis of gas turbine fault data,and some results are obtained.The Conv-LSTM model is established to model and predict the fault data sequence from two dimensions of time and space,and remarkable results are obtained.From the perspective of spatio-temporal data analysis,a kind of data-driven life prediction method is provided.