Research On Three-spool Gas Turbine Multi-fault On-line Monitoring And Hierarchical Diagnosis

With the transformation of gas turbine fault diagnosis technology from simple manual inspection,scheduled maintenance and post-event maintenance to conditionbased maintenance,online condition monitoring and real-time fault diagnosis of key components with intelligent algorithms and advanced reasoning technology ha s become a new research focus.Gas turbine contains several key components,but the number of measuring points on the general gas turbine is not enough to support independent monitoring and diagnosis for each key component,which may cause fault of any component to trigger a false alarm of the diagnostic system for other components.Therefore,it is of great significance to establish a complete set of fault diagnosis framework and formulate reasonable fault detection logic to realize on-line monitoring and fault diagnosis of multiple faults of gas turbines.In this paper,the research on multi-fault online monitoring and hierarchical diagnosis of three-shaft gas turbine is carried out.The main work of the paper is as follows:A mathematical simulation model for gas turbine fault diagnosis is established.The model is improved on the basis of the traditional lumped parameter model.The multi-flame tube structure is designed,and the turbine rotation and mixing effects are considered.The new simulation model with a circumferentially distributed structure is formed,which model can accurately describe the effect of local combustion chamber fault on exhaust temperature distribution.The maximum error of the design condition is 1.86%.When the output power of the gas turbine is greater than 50% of the rated power,the relative deviation between the model simulation value and the actual value is within ±0.5%.The model provides data support and verification platform for the follow-up fault diagnosis algorithm research.In addition,a multiinput and multi-output equilibrium manifold expansion model is obtained through system identification,which provides a system model with real-time computing capability for the development of Kalman filter-based on-line monitoring and hierarchical diagnosis technology.For sensor faults,a fault diagnosis method based on filter array and unsupervised learning is proposed,which realizes the detection,isolation and estimation of multi-sensor faults.The experimental results show that the average value of the minimum detectable fault of the sensor is about the output value 0.46% of the gas turbine under the rated operating condition.Aiming at actuator faults,a fault diagnosis method based on state parameter expansion is proposed,which realizes accurate estimation of actuator faults.The experimental results show that the average error of the estimation result of the gradual fault is 1.87%,and the average error of the estimation result of the abrupt fault is 2.22%.Finally,by combining the two fault diagnosis methods and designing reasonable fault detection and isolation logic,the first-layer framework of the gas turbine multi-fault online monitoring and hierarchical diagnosis system is formed.The first-layer framework can realize the fault detection and isolation of sensors,actuators and gas turbine body.For gas turbine combustion chamber faults,an energy distribution coefficient that reflects the performance change of the combustion chamber is defined.By combining the inverse calculation of turbine exhaust temperature with the improved Kalman filter array,the influence of various interference factors on the estimation of energy distribution coefficient is suppressed,and the sensitive detection and accurate location of local faults in combustion chamber are realized.Further,a logical classifier is designed by analyzing the variation law of the energy distribution coefficient under different fault conditions,which constitutes the second-level framework of the gas turbine multi-fault online monitoring and hierarchical diagnosis system.On the basis of the first layer framework,the faults of gas turbine body are further classified in detail.The detection and isolation of combustion chamber faults and compressor/turbine faults are realized.A fault diagnosis method based on SRCKF array is proposed for gas turbine compressor and turbine faults.By constructing a model array,the underdetermined fault diagnosis problem is well-posed,and the accurate estimation of the compressor and turbine abrupt faults is realized.The simulation results show that the accuracy of different gas path component fault estimations are more than 97.5%.A fault detection method based on isolation factor is proposed to detect and isola te the abrupt fault of compressor and turbine.Finally,the two methods constitutes the third-layer framework of the gas turbine multi-fault online monitoring and hierarchical diagnosis system,which realizes the accurate detection,isolation and estimatio n of compressor and turbine faults.Based on the principle of hierarchical classification,this paper establishes a gas turbine multi fault on-line monitoring and hierarchical diagnosis system with three-layer frame structure.The system transforms the multi fault classification problem into local two classification problem by layer by layer,and realizes the detection,isolation and estimation of gas turbine sensor,actuator,combustion chamber,compressor and turbine faults.The simulation results prove t hat the proposed system can meet the needs of multi fault on-line monitoring and diagnosis of gas turbine.