Research On Fault Diagnosis Method Of Gas Turbine Blade Fracture

As a large-scale complex rotating power mechanical equipment,gas turbine has the advantages of high thermal efficiency,fast start-up and variable operating speed and low noise.It has been widely used in air power,power generation system,oil pipeline and other fields.As the core component of the gas turbine,the blade works under the harsh conditions of high pressure,high speed and high temperature for a long time,and the probability of failure is very high.Blade fracture is one of the typical failure modes.Once the gas turbine blade breaks,not only the performance of the whole machine will be degraded,but also the broken blades that fly out at high speed will damage the rotor components such as the subsequent blades and the static components such as the casing.In turn,secondary failures such as rubbing,shaft jamming,or even fire are caused,which seriously threatens the safe and reliable operation of the gas turbine.Monitoring the state of the gas turbine blades and analyzing the state of the blades in real time are important means to ensure the secure and reliable operation of the gas turbine.The paper focuses on the following researches on the fracture failure of gas turbine compressor rotor blades:(1)In order to extract the characteristic parameters of the gas turbine blade fracture fault,the vibration response of the stator blade under the excitation force of the trailing edge of the normal and fractured cascade was calculated which based on the mechanism of the blade wake excitation force and the transmission law of the blade vibration response under the complex path of the thin-walled casing.The characteristic parameters of blade fracture fault based on mechanism were proposed based on the above calculation results.At the same time,according to the change of the rotor centroid position caused by the rotor blade fracture,the time domain,frequency domain and time-frequency domain features were extracted to form the multi-dimensional vibration characteristic parameters of the gas turbine blade fracture fault mixed domain based on the vibration data;(2)In order to realize blade fault early warning of gas turbine,a blade fault early warning method based on single classification support vector machine(OCSVM)was proposed by taking mechanism characteristic parameters as input parameters;Simultaneously combined with the deep belief neural network(DBN),a blade fault early warning method based on DBM-OCSVM was proposed by taking the multidimensional vibration characteristics of mixed domain as input parameters,where the DBM network reduces the dimensionality of the input high-dimensional vibration characteristics,and the OCSVM realizes the blade state early warning;(3)In order to accurately identify the rotor blade fracture faults for the faulty gas turbine blades,the improved heuristic segmentation algorithm was used to analyze the variation law of blade vibration response which eliminated the influence of similar characteristics caused by airflow parameter changes or operating condition disturbances on the recognition results,so as to accurately identify blade fracture fault;(4)In order to locate the location of gas turbine blade fracture faults,a blade fracture position identification parameters were proposed to realize the location of blade fracture faults based on the closed-loop feedback control principle of the gas turbine.The effectiveness of the method in this paper was verified by the case data of a certain type of gas turbine blade fracture fault.