A method for continuous and high-precision diagnosis of gas turbine engine gas path faults

Abstract

The present invention proposes a method for continuous high-precision diagnosis of gas turbine engine gas path faults, which belongs to the technical field of gas turbine engine fault diagnosis; the steps include: establishing a nonlinear component-level model of the engine; capturing the dynamic effects of the engine transient process; the output of the nonlinear component-level model of the engine is an estimate of the engine observation parameter; collecting the engine observation parameter measurement value through the sensor; and then using the solver to iteratively update the degradation factor, so that the residual between the observation parameter estimate output by the fault diagnosis model and the actual engine observation parameter measurement value obtained by the sensor on the engine is minimized. The present invention captures the dynamic effects of the transient process at continuous moments through the gas path measurement parameters based on the time series, thereby realizing continuous high-precision diagnosis of the health status of the gas turbine engine. The present technology can provide a new continuous high-precision diagnosis method for gas turbine engines under steady-state and transient conditions.

Description

Technical Field

[0001] The invention belongs to the technical field of gas turbine engine fault diagnosis, and in particular relates to a continuous high-precision diagnosis method for gas turbine engine gas path faults. Background Art

[0002] In recent years, gas turbine engine gas path fault diagnosis has received increasing attention. Gas path fault diagnosis is crucial to ensure the safe, economical and reliable operation of gas turbine engines. Existing research focuses on steady-state conditions, and steady-state gas path fault diagnosis cannot achieve continuous diagnosis of the health status of gas turbine engines. In addition, engineering practice has been able to record and save monitoring data throughout the entire working process of gas turbine engines, but the lag in basic theoretical research on transient gas path fault diagnosis has resulted in a large amount of transient monitoring data during flight processes being unable to be effectively utilized. However, there are dynamic effects in the transient process of gas turbine engines, and the dynamic effects seriously affect the accuracy of fault diagnosis. Therefore, the existing online diagnosis of gas turbine engine gas path faults is mainly based on the gas path parameters of steady-state conditions during the cruise phase, which cannot meet the continuous monitoring needs of the engine health status during flight. The engine may encounter sudden failures during transient processes such as takeoff and climb, and the existing steady-state gas path fault diagnosis system will cause delays in health status diagnosis. Therefore, it is necessary to explore the influence of dynamic effects in transient processes to achieve continuous and high-precision diagnosis of gas path faults and improve the emergency response capability of gas turbine engines. Continuous and high-precision diagnosis of gas path faults under transient operating conditions of gas turbine engines is an urgent problem to be solved. Therefore, it is crucial to propose a fault diagnosis algorithm that can continuously capture the actual health status, especially to accurately diagnose sudden faults in transient processes, for gas turbine engine health monitoring. Summary of the invention

[0003] Technical issues to be solved:

[0004] In order to overcome the shortcomings of the prior art, the present invention provides a method for continuous and high-precision diagnosis of gas turbine engine gas path faults. The fault diagnosis captures the dynamic effects of the transient process of the gas turbine engine based on the time-series gas path measurement parameters to support continuous and high-precision diagnosis of gas path faults, thereby overcoming the limitation of the prior art that it is impossible to continuously and high-precision diagnose the health status of the engine.

[0005] The technical solution of the present invention is: a method for continuous high-precision diagnosis of gas turbine engine gas path faults, wherein the gas path faults are applicable to both steady-state and transient gas path faults; the method is characterized in that the specific steps are as follows:

[0006] Step 1: Establish a nonlinear component-level model of the engine;

[0007] Step 2: Collect engine observation parameter measurement values ​​through sensors;

[0008] Step 3: Capturing the dynamic effects of the engine transient process in the engine nonlinear component level model;

[0009] Step 4: The engine nonlinear component-level model output is an estimate of the engine observation parameters;

[0010] Step 5: Iteratively update the degradation factor X through the solver so that the observed parameter estimate Z output by the fault diagnosis model is Predict The actual engine observation parameter measurement value Z obtained by the sensor on the engine Acual The residual error between them is minimal.

[0011] A further technical solution of the present invention is: in the step 1, a Newton-Raphson iteration method is used to iterate the engine nonlinear component level model.

[0012] A further technical solution of the present invention is: in step 2, the sensor is located on the engine.

[0013] A further technical solution of the present invention is: in the step 3, the method for capturing the dynamic effects of the engine transient process in the engine nonlinear component level model is:

[0014] (1) Based on continuous data, the rotor speed at adjacent moments is obtained, and then the rotor acceleration can be obtained. The unbalanced work at any moment can be known from the rotor acceleration, moment of inertia and speed. When considering the unbalanced work, the turbine work is always equal to the compressor work plus the unbalanced work and the power consumption of other auxiliary equipment, so as to update the constraints of fault diagnosis.

[0015] (2) Based on the engine body temperature T at the previous moment m , calculate the gas temperature and engine body temperature after considering the heat soak effect;

[0016] (3) Based on the first-order inertia theory, the time lag effect of sensors and actuators is considered.

[0017] A further technical solution of the present invention is: specific method steps for considering unbalanced work:

[0018] Since the engine rotor speed N is monitored in time series, the rotor acceleration is obtained by the rotor speed deviation in a finite time through equation (1);

[0019]

[0020] Among them, t is the monitoring time;

[0021] In this case, the unbalanced work SP is calculated from the rotor acceleration, rotor speed N and shaft inertia I by equation (2):

[0022]

[0023] Then, the shaft work balance between the axes is obtained by equation (3); this equation is applicable to both steady-state and transient conditions, where SP is zero under steady-state conditions; therefore, when unbalanced work is considered under both steady-state and transient conditions, the shaft work balance condition is met;

[0024] TW=SP+CW+AW (3)

[0026] Among them, TW is the work done by the turbine, CW is the work done by the compressor, and AW is the power consumed by the auxiliary equipment.

[0027] A further technical solution of the present invention is: specific method steps considering the heat soak effect:

[0028] The heat transfer between the airflow and the engine body is obtained by equation (4);

[0029] Q=U ht ·A ht (T g -T m )·(e -Δt / τ -1) (4)

[0030] Among them, Q is the heat transfer, U ht is the heat transfer coefficient, A ht is the effective heat transfer area, T g is the gas temperature in the current step without considering the heat soak effect, Δt is the time step, and τ is the time constant;

[0031] Heat transfer coefficient U ht Obtained by equation (5);

[0032]

[0033] Where FC is the convective heat transfer coefficient, k m Thermal conductivity of the material, l eff is the effective heat exchange length;

[0034] The time constant τ is obtained by equation (6);

[0035]

[0036] Among them, W m is the equivalent mass, c m is the specific heat of the component material;

[0037] Average temperature T m Obtained by equation (7);

[0038]

[0039] Enthalpy drop of gas ΔH g It can be obtained from equation (8); when the gas pressure and enthalpy are known, the gas temperature considering the heat soaking effect in the current step is obtained;

[0040]

[0041] Among them, W g is the air mass flow rate.

[0042] A further technical solution of the present invention is: specific method steps considering the time lag effect:

[0043] During transient operation, there is a time lag between the engine sensors and actuators, which is represented by a first-order inertia link;

[0044]

[0045] Where τ is the time constant, Y(s) is the input value with delay and B(s) is the input value without delay.

[0046] A further technical solution of the present invention is: in step 5, a Newton-Raphson method is selected to establish an iterative solver.

[0047] A further technical solution of the present invention is: in step 5, the performance simulation and fault diagnosis processes are called in the same iterative loop;

[0048] (1) The ratio of the characteristic parameter of the degraded component to the characteristic parameter of the healthy component is used to characterize the degradation degree of each characteristic parameter in the component, that is, the degradation factor X;

[0049] (2) The flight altitude, Mach number, and fan inlet conditions are obtained through the sensors in step 2;

[0050] (3) In the nonlinear component-level model of the engine in step 1, the convergence criteria are divided into two categories according to the aerodynamic and thermal relationship between the engine principle and each component: a set of convergence criteria is obtained from the gas path measurement, including T4, T5, T9, and T 10 The measured and estimated values ​​meet the threshold conditions; another set of convergence criteria requires meeting the flow balance, shaft power balance and the design value of the nozzle area at the design point;

[0051] (4) During the iteration process, the root mean square error RMSE defined in equation (10) is selected and 1E-5 is used as the threshold to evaluate the convergence:

[0052]

[0053] Where n is the number of measurement parameters.

[0054] Beneficial Effects

[0055] The beneficial effects of the present invention are as follows: the present invention proposes a continuous high-precision diagnosis method for steady-state and transient gas path faults of a gas turbine engine, which captures the dynamic effects of transient processes at continuous moments through time-series-based gas path measurement parameters; wherein the measurement value is time-dependent, and can consider unbalanced work, heat soak effects of gas and engine body, and time lag effects in continuous time steps. This enables continuous high-precision diagnosis of the health status of the gas turbine engine. This technology can provide a new continuous high-precision diagnosis method for gas turbine engines under steady-state and transient conditions. The method is suitable for industrial gas turbines, turbojet engines, turbofan engines, turboprop engines, etc.

[0056] The method proposed by the present invention takes into account the unbalanced work, heat soak effect and time lag effect, and can continuously and accurately diagnose the degree of engine fault under transient conditions. The proposed method can diagnose sudden faults in transient processes within 0.1582 seconds with a maximum relative error of 0.0059%. Therefore, the present invention can realize continuous and high-precision diagnosis of gas turbine engine gas path faults under steady-state and transient conditions.

[0057] The present invention supplements the research on gas turbine engine gas path fault diagnosis methods under transient operating conditions, helps us understand gas turbine engine gas path fault diagnosis in many ways, improves the online health monitoring capability of gas turbine engines, and is beneficial to the safety, availability and reliability of gas turbine engines, providing theoretical and technical support for the construction of gas turbine engine safe operation guarantee capabilities.

[0058] Based on the description in the embodiments, the average diagnostic error of the method proposed by the present invention is 0.0009%, which is better than the existing method. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 The turbofan engine components involved and their section numbers;

[0060] Figure 2 Schematic diagram of existing fault diagnosis methods;

[0061] Figure 3 Existing steady-state model-based fault diagnosis flow chart;

[0062] Figure 4 Schematic diagram of a continuous high-precision diagnosis method for steady-state and transient gas path faults in a gas turbine engine;

[0063] Figure 5 The changing trend of oil supply plan and unbalanced work in transient process;

[0064] Figure 6 Estimation of engine degradation factors during transient processes;

[0065] Figure 7 The average relative error of the transient process degradation factor;

[0066] Figure 8 The influence of heat soak effect on exhaust temperature;

[0067] Fig. 9 Estimation of engine degradation factors during transient processes;

[0068] Fig.10 The average relative error of the transient process degradation factor;

[0069] Fig.11 Relative error of the transient process degradation factor;

[0070] Fig.12 Comparison of fault diagnosis accuracy among three examples;

[0071] Fig.13 Relative error of the transient process degradation factor;

[0072] Fig.14 Comparison of fault diagnosis accuracy of four examples.

[0073] Fig.15 Flowchart of a method for continuous and high-precision diagnosis of gas turbine engine gas path faults.

[0074] Explanation of the reference numerals: 1. Environmental conditions, 33. Fan culvert outlet, 3. Fan culvert outlet, 4. Low-pressure compressor outlet, 5. High-pressure compressor outlet, 51. Air flow section for cooling of high-pressure turbine inlet guide vanes, 52. Air flow section for cooling of low-pressure turbine inlet guide vanes, 6. Combustion chamber outlet, 7. High-pressure turbine inlet, 8. Low-pressure turbine inlet, 9. Low-pressure turbine inlet, 10. Low-pressure turbine outlet, 35. Cubic tail nozzle outlet, 12. Cubic tail nozzle outlet. DETAILED DESCRIPTION

[0075] The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, but should not be construed as limiting the present invention.

[0076] This embodiment provides a method for continuous high-precision diagnosis of steady-state and transient gas path faults in a gas turbine engine, comprising the following steps:

[0077] S1. Use Newton-Raphson iteration method to establish engine nonlinear component level model;

[0078] S2. The sensor located on the engine collects the measured values ​​of the engine observation parameters;

[0079] S3. The dynamic effects of the engine transient process are captured in the model through the following three methods: (a) Based on continuous data, the rotor speed at adjacent moments is obtained, and then the rotor acceleration can be obtained. The unbalanced work at any moment can be known from the rotor acceleration, moment of inertia and speed. When considering unbalanced work, the turbine work is always equal to the compressor work plus the unbalanced work and the power consumption of other auxiliary equipment to update the constraints of fault diagnosis; (b) Based on the engine body temperature Tm at the previous moment, the gas temperature and engine body temperature after considering the heat soak effect are obtained; (c) Based on the first-order inertia theory, the time lag effect of sensors and actuators is considered. In the method proposed in the present invention, the measured value is time-dependent, and the unbalanced work, heat soak effect and time lag effect can be considered in continuous time steps.

[0080] S4. The model output is the estimated value of the engine observation parameters;

[0081] S5. Select the Newton-Raphson method to establish an iterative solver and iteratively update the degradation factor X to minimize the observed parameter estimate Z output by the fault diagnosis model Predict Compared with the actual engine sensor observation parameter measurement value Z Actual The residual between .

[0082] Further, the S3 includes:

[0083] S31. Since the engine rotor speed N is monitored in time sequence, the rotor acceleration can be obtained by the rotor speed deviation in a finite time through equation (1).

[0084]

[0085] Where t is the monitoring time.

[0086] In this case, the unbalanced work SP can be calculated from the rotor acceleration, rotor speed N and shaft inertia I by using equation (2).

[0087]

[0088] Then, the shaft work balance between the axes is obtained by equation (3). This equation is applicable to both steady-state and transient conditions, where SP is zero under steady-state conditions. Therefore, when unbalanced work is considered under both steady-state and transient conditions, the proposed method can always meet the shaft work balance condition.

[0089] TW=SP+CW+AW (3)

[0090] Among them, TW is the work done by the turbine, CW is the work done by the compressor, and AW is the power consumed by the auxiliary equipment.

[0091] S32. During transient operation, changes in the gas temperature in the gas turbine engine will affect the temperature of the gas turbine engine body. This phenomenon is called the heat soak effect. The method of the present invention takes the heat soak effect into account in the continuous diagnosis of gas path faults at any time during the transient process of the gas turbine engine. The heat transfer between the gas flow and the engine body is obtained by equation (4).

[0092] Q=U ht ·A ht (T g -T m )·(e -Δt / τ -1) (4)

[0093] Where Q is the heat transfer, U ht is the heat transfer coefficient, A ht is the effective heat transfer area, T g is the gas temperature in the current step, T m is the bulk temperature in the previous step, Δt is the time step, and τ is the time constant.

[0094] Heat transfer coefficient U ht Obtained by equation (5);

[0095]

[0096] Where FC is the convective heat transfer coefficient, k m Thermal conductivity of the material, l eff is the effective heat transfer length.

[0097] The time constant τ is obtained by equation (6);

[0098]

[0099] Among them, W m is the equivalent mass, c m is the specific heat of the component material.

[0100] Average temperature T m Obtained by equation (7);

[0101]

[0102] Enthalpy drop of gas ΔH g It can be obtained by equation (8). When the gas pressure and enthalpy are known, the gas temperature considering the heat soaking effect in the current step can be obtained.

[0103]

[0104] Among them, W g is the air mass flow rate.

[0105] S33. During transient operation, there is a time lag between the engine sensors and actuators, which can be represented by a first-order inertia link.

[0106]

[0107] Where τ is the time constant, Y(s) is the input value with delay and B(s) is the input value without delay.

[0108] Further, the S5 includes:

[0109] The performance simulation and troubleshooting processes are called in the same iteration loop, rather than nested iterations.

[0110] S51 uses the ratio of the characteristic parameter of the degraded component to the characteristic parameter of the healthy component to characterize the degradation degree of each characteristic parameter in the component, that is, the degradation factor X.

[0111] S52. Flight altitude, Mach number and fan inlet conditions are obtained through sensors in S2.

[0112] S53. In the S1 engine model, the inlet conditions of the fan are obtained through the air inlet. According to the engine principle and the aerodynamic and thermal relationship of each component, the convergence criteria can be divided into two categories: a set of convergence criteria is obtained from the gas path measurement, including T4, T5, T9, and T 10 The measured and estimated values ​​meet the threshold conditions. Another set of convergence criteria requires meeting the flow balance, shaft work balance, and nozzle area design value at the design point.

[0113] S54. During the iteration process, the root mean square error (RMSE) defined in formula (10) is selected, and 1E-5 scientific notation, representing 0.00001, is used as the threshold to evaluate convergence.

[0114]

[0115] Where n is the number of measurement parameters.

[0116] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is further described in detail below in conjunction with specific implementation methods.

[0117] The implementation mode of the present invention takes the fault diagnosis of the gas path components of a certain type of high bypass ratio turbofan engine as an example, wherein the actual turbofan engine is replaced by a nonlinear component-level model, and the iteration method of this component-level model is the Newton-Raphson method.

[0118] Figure 1The turbofan engine structure and its cross-sectional numbering shown include an air inlet 1, a fan, a low-pressure compressor, a high-pressure compressor, a combustion chamber, a mixing model of high-pressure compressor bleed air, a high-pressure turbine, a low-pressure turbine, an external tail nozzle and an internal tail nozzle.

[0119] The design parameters of the turbofan engine are shown in Table 1. The physical quantities measured in the turbofan engine for fault diagnosis are listed in Table 2.

[0120] Table 3 summarizes the degradation factors associated with the turbofan engine. "Healthy State 1" refers to a high bypass split exhaust turbofan engine that has completed 6000 flight cycles, where "Healthy State 2" refers to half the degradation level of "Healthy State 1". "Healthy State 2" applies to the engine degradation level before the sudden failure, where "Healthy State 1" refers to the engine degradation level after the sudden failure.

[0121] Table 1 Turbofan engine design point parameters

[0122]

[0123] Table 2 Physical quantities measured in turbofan engines

[0124]

[0125] Table 3 Turbofan engine degradation factors

[0126]

[0127]

[0128] The iterative variables of the method proposed in the present invention are the ten degradation factors listed in Table 3. Figure 3 The fuel supply in is the actual fuel quantity of the engine. The convergence standard is Figure 3 The detailed process is explained as follows: the flight altitude, Mach number and inlet conditions are obtained through the sensors on the engine. Then, the fan intake conditions are obtained through the inlet duct model. The fan pressure ratio PR is calculated according to formula (7): FAN,BP , where P 33 Obtained through gas path measurement. Given the inlet conditions, rotor speed, and pressure ratio of the fan module in the engine model, the inlet conditions of the low-pressure compressor and the outer duct can be determined through the fan module.

[0129] PR FAN,BP =P 33 / P2 (7)

[0130] Where P2 is the intake pressure after taking intake loss into account.

[0131] Low pressure compressor pressure ratio PR LPCIt is obtained by formula (8), where P4 is the air path measurement and P3 can be calculated by the aforementioned fan module. Then, when the pressure ratio, rotor speed and inlet conditions are known, the low-pressure compressor module is calculated. It is worth noting that the flow obtained in the low-pressure compressor module is used to update the internal flow at the outlet of the fan module, and the fan external flow is further obtained based on the fan inlet flow, thereby determining the bypass ratio. Correspondingly, the compression work of the fan is also updated according to the new bypass ratio.

[0132] PR LPC =P4 / P3 (8)

[0133] High pressure compressor pressure ratio PR HPC It can be obtained by formula (9), where P5 and P4 are gas path measurement values. When the pressure ratio, rotor speed and inlet conditions are known, the outlet operating conditions can be calculated using the high-pressure compressor module.

[0134] PR HPC =P5 / P4 (9)

[0135] Since the outlet working conditions of the high-pressure compressor module are known, the outlet parameters of the combustion chamber module can also be calculated when the fuel flow rate is known. Furthermore, the mixing model is used to calculate the inlet conditions of the high-pressure turbine. The high-pressure turbine drop pressure ratio PRHPT is obtained by formula (10), where P9 is the gas path measurement value and P7 is known from the mixing model after the combustion chamber. At this time, the high-pressure turbine outlet parameters can be calculated through the high-pressure turbine module.

[0136] PR HPT =P7 / P9 (10)

[0137] The inlet conditions of the low-pressure turbine module can be obtained through the mixing model after the high-pressure turbine. LPT It can be obtained from formula (11) that P9 and P 10 is the gas path measurement value. Finally, the inner / outer duct and nozzle modules are used to calculate the inner and outer exhaust parameters respectively.

[0138] PR LPT =P9 / P 10 (11)

[0139] There are 11 convergence criteria in the diagnostic algorithm, using Figure 3 The right hexagonal box in the figure shows the convergence criteria. The convergence criteria can be divided into two categories: a set of convergence criteria is obtained from gas path measurement, including T4, T5, T9, and T 10The measured and estimated values ​​meet the threshold conditions. Another set of convergence criteria requires meeting the flow balance, shaft power balance, and nozzle area design value at the design point. The shaft power of the low-pressure and high-pressure shafts in the hexagonal box on the right must ensure shaft power conservation, that is, when using a steady-state fault diagnosis model, the turbine work must always be equal to the compressor work plus the auxiliary equipment power consumption. The prior art ignores the unbalanced work under transient conditions, which may lead to diagnostic errors. Another assumption of the prior art is that in the steady-state fault diagnosis model, the heat soak effect and time lag effect under transient conditions are ignored. In this case, the accuracy of the diagnosis may be affected.

[0140] The present invention aims to diagnose the health of a gas turbine engine using time series data under steady-state and transient conditions. Since a gas turbine engine exhibits fast dynamic response characteristics, the dynamic effect cannot be ignored.

[0141] The schematic diagram of the method proposed by the present invention is as follows Figure 4 As shown, based on the speed at the previous moment and the current moment, the unbalanced work at the current moment is obtained; based on the engine body temperature T at the previous moment m , the gas temperature and engine body temperature after considering the heat soak effect are obtained; based on the first-order inertia theory, the time lag effect of the sensor and the actuator is considered. Obviously, in the method proposed by the present invention, the measurement is time-dependent, in which the unbalanced work, heat soak effect and time lag effect need to be considered in the continuous time step. Figure 4 Although two triangle points under transient conditions are selected to illustrate Figure 4 Although this is a new approach in , the proposed method is also applicable to steady-state conditions.

[0142] Specifically, based on continuous data, the rotor speed at adjacent moments is obtained, and then the rotor acceleration can be obtained. The unbalanced work at any moment can be known from the rotor acceleration, moment of inertia and speed. When considering unbalanced work, the turbine work is always equal to the compressor work plus the unbalanced work and the power consumption of other auxiliary equipment to update the constraints of fault diagnosis. The engine body temperature at the next moment is calculated using the estimated value of the engine body temperature at the previous moment, so that the heat soak effect under transient conditions is considered in the diagnostic model. In addition, the time lag effect of transient conditions can be characterized by a first-order inertia link. Finally, since the measurement and diagnosis are continuous, sudden faults can also be accurately diagnosed at any time point during the transient operation process.

[0143] In order to verify the diagnostic accuracy of the gas turbine engine timing fault diagnosis method proposed in the present invention in steady state and transient processes, the following four simulation case verification studies were carried out. At the same time, in order to verify the advancedness of the method proposed in the present invention, comparative tests were conducted with the prior art. In all case studies, a laptop with an Intel(R) i7 CPU@2.90GHz and 16GB RAM was used to evaluate the fault diagnosis time.

[0144] The goals of these four examples are as follows:

[0145] Example 1: This example aims to evaluate the fault diagnosis effect of existing methods under transient conditions without considering the engine heat soak effect.

[0146] Example 2: In this example, the heat soak effect is taken into account in the gas path measurement parameters. The purpose of this example is to study the accuracy of the existing method for transient engine fault diagnosis under the condition of considering the heat soak effect, and then to compare it with the method proposed in the present invention.

[0147] Example 3: This example aims to illustrate the comparison and advantages of the diagnostic results of the proposed method over the existing method in Example 2.

[0148] Case Study 4: The previous three cases tested the diagnosis results under constant fault level during transient operation, while this case aims to prove that the proposed method can accurately diagnose sudden faults during transient operation. The first three cases have a constant fault level of "healthy state 1", as shown in Table 3. In case study 4, we inject a fault level labeled "healthy state 2" between [0-3) seconds, a sudden fault at 3.0 seconds, and "healthy state 1" between [3-15] seconds.

[0149] Example 1: Existing method - transient process gas path fault diagnosis without considering heat soak effect

[0150] Existing methods do not consider the residual power in fault diagnosis under transient conditions. When performing fault diagnosis on heavy-duty industrial gas turbines, the accuracy obtained by this method may be acceptable because the moment of inertia of heavy-duty industrial gas turbines is relatively large. However, this assumption will affect the diagnostic accuracy of other types of gas turbine engines.

[0151] Figure 5 The above shows the accelerated fuel supply plan of a turbofan engine during transient operation with a time step of 0.1 seconds. During transient operation, the compressor work between the shafts is not equal to the coaxial turbine work. Figure 5 As shown in , during transient operation, the unbalanced work between the compressor and turbine of both the low-pressure shaft and the high-pressure shaft is close to 320kW. Figure 5As can be seen from the following, the maximum relative deviations between the compressor work and the turbine work on the low-pressure shaft and the high-pressure shaft are 5.3% and 3.0%, respectively. Therefore, if the residual power is ignored, the relative deviation will propagate to the fault diagnosis results. It can be foreseen that the greater the residual power, the lower the diagnostic accuracy. The average calculation time for diagnosis using the existing method is 0.2071s. The diagnostic results based on the existing method are shown in Figure 2. Figure 6 As shown in the figure. It can be seen that the residual power significantly affects the accuracy of fault diagnosis. With the transient acceleration process, the error of fault diagnosis continues to increase until the maximum diagnostic error occurs at 3s. Then, the estimation error of the degradation factor decreases as the residual power decreases. In this case, the existing method will cause fluctuations in the diagnosis results, which may cause false alarms. In addition, the faster the fuel flow changes, the larger the residual power corresponds, and the larger the diagnostic error. The average estimation error of all 10 degradation factors in the transient process is shown in Figure 7 Although the average maximum estimation error of the 10 degradation factors is around 1.5%, FAN.E The maximum estimation error of the transient operation process at 2.5s is 6.5852%, which seriously affects the credibility of fault diagnosis.

[0152] In summary, the above diagnostic results may be acceptable if the residual power can be ignored. This generally occurs in transient processes where the fuel flow changes relatively slowly. In other cases, the existing methods will cause significant fluctuations in the diagnostic results. Therefore, when recording each set of measurement values, the existing diagnostic methods cannot continuously monitor the engine operating conditions with high accuracy. It is even more unsuitable for monitoring sudden faults in transient processes.

[0153] Example 2: Existing methods - transient process gas path fault diagnosis considering heat soak effect

[0154] Gas turbine engines not only face shaft power imbalance under transient conditions, but also heat exchange between the gas and the engine body. Figure 8 The time variation of exhaust temperature during transient operation with and without considering the heat soak effect is shown. Compared with ignoring the heat soak effect, the heat soak effect will cause a delay in the exhaust temperature change, such as Figure 8 If the engine performs a rapid acceleration / deceleration maneuver, the engine degradation factor estimated by the existing method will be significantly affected.

[0155] In the transient process of 15 seconds, the average diagnosis time of the existing method is 0.2083s. Fig. 9It can be seen that the estimated degradation factor has a larger deviation from the actual healthy state compared with Example 1. In transient process gas path fault diagnosis considering the heat soak effect, in addition to the HPT efficiency degradation factor, the heat soak effect will aggravate the diagnostic error of the existing method. The residual power will cause the HPT efficiency degradation to be overestimated, while the heat soak effect will cause the HPT efficiency degradation to be underestimated. Fig.10 is a summary of the average estimation errors of the 10 degradation factors. When the heat soak effect is considered in transient process gas path fault diagnosis, the maximum average error of the 10 degradation factors obtained by the existing method increases from 1.4240% in Example 1 to 5.5853% in Example 2. In addition, X FAN.E During the entire transient operation, the maximum error is 13.2919% at 3.0s in Example 2. Ignoring the heat soak effect during the transient process will significantly affect the accuracy of fault diagnosis.

[0156] The results show that under transient operating conditions, the existing steady-state fault diagnosis system will produce significant estimation errors. In addition, the fluctuation of fault diagnosis results may cause false alarms. If the diagnosis system frequently issues false alarms, it will seriously affect the credibility of the fault diagnosis system.

[0157] Example 3: The method proposed by the present invention - transient process gas path fault diagnosis under constant degradation

[0158] In this example, the method proposed in the present invention is used for continuous diagnosis of transient gas path faults, and the heat soak effect is taken into account in the gas path faults at any time during the transient process of the gas turbine engine. Fig.11 The relative errors of the 10 degradation factors in the transient process are shown. It is worth noting that Fig.11 X in FAN.E The maximum relative error is 0.0066% at 11.4s. Fig.12 From the comparison results of the three examples in , it can be seen that the method proposed in the present invention can estimate the degradation factor more accurately than the existing method. Table 4 summarizes the diagnostic results of the three examples. The calculation time of Example 3 is 0.1619s, which is slightly better than Example 2. The average diagnostic error of the diagnostic method proposed in the present invention is 0.0009%, which is better than the existing method, among which the average error of Example 2 is 1.3754%, and the maximum relative errors of the existing method and the present method in the entire transient operation process are 13.2919% and 0.0066%, respectively. The results show that the proposed timing fault diagnosis method is superior to the existing method in terms of calculation time and diagnostic accuracy.

[0159] Table 4 Comparison of diagnostic results of three examples

[0160]

[0161] Example 4: Method proposed by the present invention - transient process sudden fault diagnosis

[0162] During operation, aircraft engines may face sudden failures caused by foreign matter ingestion. In addition, bird strikes may occur during the transient or quasi-steady-state operation of takeoff and landing, causing sudden engine failures. Therefore, it is necessary to verify the diagnostic performance of the proposed method under sudden failures under transient conditions. Figure 5 During the transient operation on , it is assumed that a sudden fault occurs at 3.0s. The operating state suddenly degenerates from "operating state 2" to "operating state 1", as shown in Table 3. Fig.13 The relative error of the diagnostic results of this method under transient process sudden faults is given. Fig.13 It can be seen that the method proposed in the present invention can accurately capture sudden faults in transient processes. Fig.14 The results of the 10 degradation factors in all four examples are compared. It can be seen that the relative errors of all degradation factors in the case of sudden failure in Example 4 remain at the same level as in Example 3, and the maximum relative errors of the 10 degradation factors are all less than 0.004%.

[0163] Table 5 gives the diagnostic results of all four examples. The average computation time of example 4 is only 0.15820s. It is worth noting that the computation time of example 4 is similar to that of example 3. The sudden fault does not affect the computational efficiency of this method. From the perspective of diagnostic accuracy, the average error and maximum error of all 10 degradation factors are 0.0009% and 0.0059% respectively during the entire transient process. FAN.E The maximum error occurs at 3.6s, while the sudden failure occurs at 3.0s, which means that the sudden failure will not affect the diagnostic accuracy of the method proposed in the present invention.

[0164] Table 5 Comparison of diagnostic results of four examples

[0165]

[0166] The results show that this method can use time series data to continuously and accurately diagnose the health status of the engine in steady-state and transient processes, and can accurately estimate sudden faults even in transient states.

[0167] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention without departing from the principles and intent of the present invention.

Claims

1. A method for continuous high-precision diagnosis of gas turbine engine gas path faults, wherein the gas path faults include steady-state and transient gas path faults; characterized in that The specific steps are as follows: Step 1: Establish a nonlinear component-level model of the engine; Step 2: Collect engine observation parameter measurement values ​​through sensors; Step 3: Capturing the dynamic effects of transient processes in the nonlinear component-level model of the engine; Step 4: The engine nonlinear component-level model output is an estimate of the engine observation parameters; Step 5: Iteratively update the degradation factor X through the solver so that the observation output of the fault diagnosis model Parameter estimates The actual engine observation parameter measurement value obtained by the sensor on the engine The residual between them is the smallest; In the step 3, the method for capturing the dynamic effects of the engine transient process in the engine nonlinear component level model is as follows: Specific steps to consider unbalanced work: Due to engine speed N It is monitored in time series, and the rotor acceleration is obtained by the rotor speed deviation in a finite time through equation (1); in, t To monitor the moment; In this case, the unbalanced power SP The rotor acceleration and rotor speed N and shaft inertia I Calculated by formula (2): Then, the shaft work balance between the axes is obtained by equation (3); this equation is applicable to both steady-state and transient conditions, where SP It is zero under steady-state conditions; therefore, when unbalanced work is considered under both steady-state and transient conditions, the condition for shaft work balance is satisfied; in, TW To do work for the turbine, CW To do work for the compressor, AW Consumes power for auxiliary equipment; (2) Based on the engine body temperature at the previous moment , calculate the gas temperature and engine body temperature after considering the heat soak effect; the specific method steps for considering the heat soak effect are: The heat transfer between the airflow and the engine body is obtained by equation (4); in, Q It is heat exchange. is the heat transfer coefficient, is the effective heat transfer area, is the gas temperature in the current step without considering the heat soak effect, is the engine body temperature in the previous step, is the time step, is the time constant; Heat transfer coefficient Obtained by equation (5); in, FC is the convective heat transfer coefficient, Thermal conductivity of the material, is the effective heat exchange length; Time constant Obtained by equation (6); in, is the equivalent mass, is the specific heat of the component material; Average temperature Obtained by equation (7); Enthalpy drop of gas It can be obtained from equation (8); when the gas pressure and enthalpy are known, the gas temperature considering the heat soaking effect in the current step is obtained; in, is the air mass flow rate; (3) Based on the first-order inertia theory, the time lag effect of sensors and actuators is considered; the specific method steps for considering the time lag effect are: During transient operation, there is a time lag between the engine sensors and actuators, which is represented by a first-order inertia link; in, is the time constant, is the input value with delay and is the input value without delay.

2. The method for continuous high-precision diagnosis of gas turbine engine gas path faults according to claim 1, characterized in that: In the step 1, the Newton-Raphson iteration method is used to iterate the engine nonlinear component level model.

3. The method for continuous high-precision diagnosis of gas turbine engine gas path faults according to claim 1, characterized in that: In step 2, the sensor is located on the engine.

4. The method for continuous high-precision diagnosis of gas turbine engine gas path faults according to claim 1, characterized in that: In step 5, the Newton-Raphson method is selected to establish an iterative solver.

5. The method for continuous high-precision diagnosis of gas turbine engine gas path faults according to claim 1, characterized in that: In step 5, the performance simulation and fault diagnosis processes are called in the same iterative loop; (1) The ratio of the characteristic parameter of the degraded component to the characteristic parameter of the healthy component is used to characterize the degradation degree of each characteristic parameter in the component, that is, the degradation factor X ; (2) The flight altitude, Mach number, and fan inlet conditions are obtained through the sensors in step 2; (3) In the nonlinear component-level model of the engine in step 1, the convergence criteria are divided into two categories according to the engine principle and the aerodynamic and thermal relationship of each component: one set of convergence criteria is obtained from the gas path measurement, including the T4, T5, T9 and T10 measured values ​​and estimated values ​​satisfying the threshold conditions; the other set of convergence criteria requires satisfying the flow balance, shaft power balance and the nozzle area design value at the design point; (4) During the iteration process, the root mean square error defined in equation (10) is selected RMSE , using 1E-5 as the threshold to evaluate convergence: in, n is the number of measurement parameters.

Images