A fault diagnosis method for aero-engine fuel metering valve sticking

Abstract

The present application belongs to the field of fault diagnosis of aero-engine, and provides a fault diagnosis method for a stuck fuel metering valve (SFMV) of an aero-engine. A method for simulating the stuck SFMV fault is designed to solve the problem that the stuck SFMV fault data is difficult to obtain. Meanwhile, the fault data is converted into vibration data type to facilitate fault diagnosis. A switching DBN network model is provided to diagnose the stuck SFMV fault under different working conditions. In each working condition, the corresponding DBN parameters are used to accurately extract the feature information in the SFMV fault data, so that the SFMV under different working conditions is diagnosed, and the accuracy of the SFMV fault diagnosis under different working conditions is improved.

Description

Technical Field

[0001] This invention relates to the field of fault diagnosis of aero-engines, and in particular to a fault diagnosis method for a stuck fuel metering valve in an aero-engine. Background Technology

[0002] An aircraft engine is a device that provides thrust to an aircraft; its working principle is as follows: Figure 1 As shown: Air enters compressor 2 through intake duct 1, is compressed in compressor 2, and then enters combustion chamber 3 where fuel is injected and ignited. The high-pressure, high-temperature gas generated by combustion drives compressor 2 and turbine 4 to rotate, generating rearward thrust. Finally, the high-pressure, high-temperature gas is discharged from exhaust nozzle 5. Therefore, the energy required for aero-engines to compress air and generate thrust comes from the chemical energy released by the combustion of fuel and air. The greater the fuel flow rate, the more energy is released. Therefore, the controllability of fuel flow rate is related to whether the aero-engine can output the desired thrust. At the same time, the fuel flow rate is adjusted by controlling the opening of the fuel metering valve. However, impurities carried in the fuel, such as fibers, sand particles, and metal shavings, may enter the gap between the metering valve core and the sleeve, increasing the friction between the valve core and the sleeve, causing the metering valve to be obstructed or even jammed. Ultimately, this results in uncontrollable fuel flow, the engine's inability to output the desired thrust, and even dangerous operating conditions such as overheating, over-revving, and surge. Therefore, studying how to address the sticking of the fuel metering valve (SFMV) is of great significance for improving the reliability of aero engines.

[0003] Currently, research on SFMV mainly focuses on how to prevent its occurrence. For example, Z. Liu, Y. Tang, Z. Wang, and N. Ma, in "The contamination harm and control of engine fule-metering unit," Aviation Maintenance, vol. 60, pp. 59–60, 2003, proposed improving fuel quality, and J. Yu, in "Fuel contaminated and maintenance," Aeronautical Science and Technology, vol. 6, pp. 41–42, 2007, proposed installing filters in the fuel system. However, the sources of impurities in fuel are diverse, making it impossible to completely avoid SFMV. For instance, mechanical wear can constantly generate new metal debris, which can also cause jamming. Therefore, in addition to using preventative measures to avoid SFMV, it is also necessary to design fault diagnosis strategies for SFMV to diagnose whether SFMV has occurred in real time.

[0004] Existing fault diagnosis methods can be divided into model-based methods and data-based methods. "Hu Changhua, Xu Hualong. Analysis and Design of Fault Diagnosis and Fault-Tolerant Control of Control Systems [M]. Beijing: National Defense Industry Press, 2000.", "Zhou Donghua, Ye Yinzong. Modern Fault Diagnosis and Fault-Tolerant Control [M]. Beijing: Tsinghua University Press, 2000.", and "Jiang Bin, Mao Zehui, Yang Hao, et al. Fault Diagnosis and Fault Adjustment of Control Systems [M]. Beijing: National Defense Industry Press, 2009" all employ model-based fault diagnosis methods. Model-based methods first require establishing an accurate mathematical model of the research object and determining whether a fault has occurred by judging the difference between the model output and the object's output. However, the dynamic model of a metering valve contains uncertain fluid dynamics and unknown disturbances, making it difficult to accurately model the valve's motion. On the other hand, data-based fault diagnosis methods require collecting state data of the research object, extracting implicit feature information, and judging whether a fault has occurred based on this feature information.

[0005] In this invention, we employ a data-based fault diagnosis method to diagnose metering valve jamming. The references to "Hinton GE, Salakhutdinov R R. Reducing the dimensionality of data with neural networks[J].science, 2006, 313(5786):504-507" and "Hinton GE, Osindero S, Teh YW. A fast learning algorithm for deep belief nets[J].Neural computation, 2006, 18(7):1527-1554" indicate that DBNs have strong feature extraction capabilities, thus they can be used to extract feature information from fault data to diagnose whether metering valve jamming has occurred and the degree of jamming. However, the following problems still exist: DBN requires a sufficient amount of fault data to extract feature information. However, the fuel flow rate in the metering valve is relatively high, and the gap between the valve core and the sleeve is small, making it difficult for impurities in the fuel to enter between the valve core and the sleeve. Therefore, metering valve jamming rarely occurs, making it difficult to obtain sufficient fault data. In addition, artificially injecting jamming malfunctions is also unacceptable, as it will put the aircraft engine in a dangerous operating state and cause damage to the engine.

[0006] Furthermore, since SFMV fault data is not a vibration data type, DBN cannot directly use this data. Therefore, a data conversion method is needed to convert SFMV fault data into a vibration data type. Finally, SFMV may occur under different operating conditions, which expands the scope of SFMV fault data and makes it difficult for DBN to extract feature information from the fault data.

[0007] In summary, to achieve data-driven SFMV fault diagnosis, we need a sufficient amount of SFMV fault data, a method to convert SFMV fault data into vibration data types, and fault diagnosis methods for SFMV under different operating conditions. Summary of the Invention

[0008] To address the above problems, this invention proposes a fault diagnosis method for fuel metering valve jamming in aero-engines, including the acquisition of SFMV fault data and a DBN-based SFMV diagnosis method. First, considering that SFMV rarely occurs and that artificially injecting jamming faults can damage aero-engines, this invention proposes a method to simulate SFMV, thereby acquiring fault data of various jamming degrees. To solve the problem that DBN cannot directly use SFMV fault data, this invention proposes a data conversion method to convert SFMV fault data into vibration data type. Finally, for SFMV under different operating conditions, this invention proposes a Switch DBN (SDBN) fault diagnosis method. The method of this invention accurately extracts feature information from SFMV fault data under each operating condition using corresponding DBN parameters, thereby diagnosing SFMV under different operating conditions.

[0009] In aircraft engines, the movement of metering valves is controlled by electronic controllers, such as... Figure 3 As shown. The electronic controller uses a PID control law:

[0010]

[0011] Where e(t) = r(t) - y(t), r(t) is the expected opening of the metered valve, and y(t) is the actual opening of the metered valve. K p K I and K D The parameters are PID parameters. The electronic controller adjusts the metering valve opening through a high-speed solenoid valve and a control chamber. The dynamic mathematical model of the metering valve opening is shown in equation (2):

[0012]

[0013] Where P is the control chamber oil pressure, Ω is the metering valve opening, and F f F2 and P are the frictional forces between the valve core and the sleeve. e M2 and Fs These represent the initial pressure, forward pressure, mass, and steady-state hydraulic pressure of the metering valve, respectively. F2, P e M2 and F s Both are constants, therefore Ω is determined by P and F. f control.

[0014] SFMV occurs because contaminants in the fuel enter the gap between the valve core and the sleeve, causing F... f The increase hinders the movement of the metering valve. According to equation (2), this is equivalent to F f If P remains constant, it decreases. Since P is controlled by the output q of the electronic controller, this invention simulates SFMV by changing q and thus changing P. The rewritten equation (2) is shown below:

[0015]

[0016] Here, 'a' represents the feedforward term. By changing the value of 'a', 'q' is changed, ultimately simulating SFMV. By assigning different values ​​to 'a', different degrees of jamming faults can be simulated.

[0017] The technical solution of this invention is as follows:

[0018] A fault diagnosis method for stuck fuel metering valve in aero-engines includes the following steps:

[0019] Step 1: Obtain SFMV fault data for fuel metering valve sticking;

[0020] Based on the principle of controlling the movement of metering valves through electronic controllers in existing aero engines, by changing the output q of the electronic controller, the oil pressure in the control chamber is changed, SFMV faults are simulated, and aero engine speed data is collected as fault data.

[0021] SFMV occurs during the change of the metering valve opening Ω. According to equation (3), when the desired opening of the metering valve increases, q is positive, causing Ω to gradually increase, the fuel flow rate to increase, and the speed of the aero-engine to increase. When the desired opening of the metering valve increases, the feedforward term a of the electronic controller is negative during SFMV simulation, causing the output q of the electronic controller to decrease, which hinders the increase of the metering valve opening Ω. When the desired opening of the metering valve decreases, the feedforward term a of the electronic controller is positive during SFMV simulation, causing the output q of the electronic controller to increase, which hinders the decrease of the metering valve opening Ω.

[0022] Because the degree of jamming of the metering valve is continuous, i.e., F f Since the speed is continuously changing, the value of the feedforward term should also be continuous when simulating a jamming fault. However, under similar feedforward terms, the corresponding speed data differ very little, and the degree of jamming is almost the same. For example... Figure 4As shown, during the process of speed increase, when the speed percentage is 75%, three values are respectively assigned to a: 0, -0.4, and -0.404. Among them, the speed curves of a being -0.4 and -0.404 are almost the same. The feedforward term is a continuous value, and several specific feedforward term values are set. The feedforward term values within the range of r on both sides of it are replaced with specific feedforward term values, so as to use several specific feedforward term values to represent all degrees of jamming faults; obtain fault data under different working conditions; the speed of the aeroengine is represented by n, and the speed data is divided into two processes: speed increase and speed decrease; during the speed increase n + process, the feedforward term a of the electronic controller is a negative value, represented by a + ; during the speed decrease n - process, the feedforward term a of the electronic controller is a positive value, represented by a - ;

[0023] During the speed increase process, the speed interval [0, n max is divided into s + 1 small intervals, that is, s + 1 working conditions, represented by indicating the interval where the speed increases from to , i1, j1 ∈ {1, 2,..., s} and 1 ≤ i1 < j1 < s; m feedforward term values of different sizes are set to represent all degrees of jamming faults, represented by , l1 ∈ {1, 2,..., m}, and (s + 1)*m groups of different fault data are obtained<00​​​​​​​​​​​​​​​​​​​​​​​​​​​​Where [t1, t2] refers to the value fed forward during the speed increase process. The period with the greatest impact; Mean represents the impact on rotational speed data. [t1, t2] take the average value; Equation (4) represents the rotational speed data [t1, t2] are transformed into eigenvalues ​​T;

[0030] When the rotational speed decreases, equation (4) is rewritten as equation (5):

[0031]

[0032] Where [t1, t2] refers to the feedforward value during the speed reduction process. The period with the greatest impact, Equation (5) represents the time when the rotational speed is reduced. [t1,t2] is transformed into eigenvalues ​​T;

[0033] Substituting the eigenvalue T into formula (6), we obtain the number of sampling points N in a single period. T :

[0034] N T =[f s *T] (6)

[0035] Among them, f s The sampling frequency is [ ], where f represents the frequency. s *T is an integer;

[0036] N T Substituting into formula (7), we obtain the fault data matrix s. uv :

[0037]

[0038] u = 0, 1, ..., p-1, v = 0, 1, ..., N T -1

[0039] Where A0, C, f s f r f n p are constants; matrix s uv Convert to a one-dimensional array, let u=0,1,…,p-1, d=[b0,b1,…,b p-1 ], thus obtaining one-dimensional fault data d;

[0040] Adding Gaussian white noise wgn, the fault data s' of the vibration data type is obtained, as shown in equation (8):

[0041] s'=d+wgn (8)

[0042] After converting the data type, a switching DBN model is proposed for SFMV under different operating conditions. A network parameter selection module is added before the DBN model. When diagnosing fault data, the network parameter selection module assigns the corresponding network parameters to the DBN model according to the speed range in which the fault data is located.

[0043] The switching DBN model was trained using fault data from different speed ranges to generate the corresponding network parameters w(n). i →n j ), w(n i →n j ) indicates that when the speed range is n i →n j The network parameters used by the switching DBN model at that time;

[0044] After the switching DBN model is trained, SFMV fault diagnosis is performed. The speed range in which the SFMV occurs is determined based on the throttle lever angle. For different speed ranges, the switching DBN model selects the corresponding network parameters. The fault data of the vibration type is input into the switching DBN model. Finally, the switching DBN outputs the degree of SFMV fault, i.e., the feedforward term a.

[0045] The interval range r is 0.02.

[0046] Furthermore, experiments were conducted using the established simulation model. Multiple sets of speed data were obtained by varying the value of the feedforward term and different speed ranges, and the data was processed. The processed fault data was then used to train the SDBN, and finally, tests were performed to verify the accuracy of fault diagnosis.

[0047] The beneficial effects of this invention are:

[0048] 1) To address the difficulty in obtaining data on fuel metering valve jamming faults, a method for simulating fuel metering valve jamming faults is proposed for the first time.

[0049] 2) Since the original fault data is not suitable for direct analysis, it is converted into a vibration data type, which makes fault diagnosis more convenient.

[0050] 3) To address the difficulty in diagnosing SFMV under different operating conditions, a switching DBN network model is designed to diagnose jamming faults under different operating conditions, thereby improving the accuracy of SFMV fault diagnosis under different operating conditions. Attached Figure Description

[0051] Figure 1 This is a structural diagram of an aircraft engine; in the diagram: 1-Air intake; 2-Compressor; 3-Combustion chamber; 4-Turbine; 5-Exhaust nozzle;

[0052] Figure 2 This is a structural diagram of an aircraft engine fuel control system.

[0053] Figure 3 This is a block diagram of the fuel metering valve control.

[0054] Figure 4 These are the speed response curves with feedforward terms of 0, -0.4, and -0.404 during the speed increase process.

[0055] Figure 5(a) shows the speed response curves under different feedforward terms during the speed increase process.

[0056] Figure 5(b) shows the speed response curves under different feedforward terms during the speed reduction process.

[0057] Figure 6 It is data(0.75n) max →n max a + =-0.40) corresponds to the fault data of the vibration data type.

[0058] Figure 7 It is the network structure of DBN.

[0059] Figure 8 This is a schematic diagram of a switching DBN.

[0060] Figure 9(a) shows the rotational speed starting from 0.75 rpm. max Increase to n max During the process, the fault data of the vibration data type under the action of zero feedforward term.

[0061] Figure 9(b) shows the rotational speed starting from 0.75 rpm. max Increase to n max During the process, the feedforward term is the fault data of the vibration data type under the action of -0.40.

[0062] Figure 9(c) shows the rotational speed starting from 0.75 rpm. max Increase to n max During the process, the feedforward term is the fault data of the vibration data type under the action of -0.45.

[0063] Figure 9(d) shows the rotational speed starting from 0.75 rpm. max Increase to n max During the process, the feedforward term is the fault data of the vibration data type under the action of -0.47.

[0064] Figure 10(a) shows the rotational speed from n max Reduced to 0.75n max During the process, the fault data of the vibration data type under the action of zero feedforward term.

[0065] Figure 10(b) shows the rotational speed from n max Reduced to 0.75n max During the process, the feedforward term is the fault data of the vibration data type under the action of 0.45.

[0066] Figure 10(c) shows the rotational speed from n max Reduced to 0.75n max During the process, the feedforward term is the fault data of the vibration data type under the action of 0.48.

[0067] Figure 10(d) shows the rotational speed from n max Reduced to 0.75n max During the process, the feedforward term is the fault data of the vibration data type under the action of 0.50. Detailed Implementation

[0068] The specific design and implementation method of the present invention will be described in detail below with reference to the accompanying drawings and embodiments.

[0069] This implementation method uses MATLAB and Amesim to build as follows: Figure 2 The model shown is of an aircraft engine fuel control system, in which the fuel regulator module was built in Amesim software. The specific structure is as follows: Figure 3 As shown. The model of the aero-engine and fuel regulator was provided by the Aero-Engine Research Institute and constructed based on experimental data; therefore, the model can realistically simulate the operation of an aero-engine. Figure 2 As can be seen, the aero-engine control system is a dual closed-loop system. The outer loop input is the desired engine speed, and the output is the actual engine speed. The inner loop input is the desired fuel flow rate, which is converted into the desired opening degree of the metering valve, and the output is the actual fuel flow rate. Therefore, the inner loop is the main focus of the study.

[0070] Step 1: Obtain fault data

[0071] This invention simulates a jamming fault by changing the output of an electronic controller;

[0072]

[0073] By changing the value of the feedforward term 'a' and setting different speed ranges, multiple sets of jamming fault data can be obtained, as shown in Table 1, which represents the SFMV data during the speed increase process. In the simulation model, the speed is represented by... Indicated. Where n cur n represents the actual rotational speed. max This indicates the maximum speed, so the speed range is [0%, 100%].

[0074] Table 1. SFMV data during the speed increase process.

[0075]

[0076]

[0077] As shown in Table 2, in this embodiment, the speed increase process and the speed decrease process are divided into six intervals.

[0078] Table 2 Division of Speed ​​Range

[0079]

[0080] As shown in Figure 5, the left figure shows the speed range from 75% to 100%, with feedforward terms set to 0, -0.40, -0.45, and -0.47. The right figure shows the speed range from 100% to 75%, with feedforward terms set to 0, 0.45, 0.48, and 0.50.

[0081] Step 2: Design an SFMV fault diagnosis solution

[0082] First, the fault data is converted into vibration data. Using formula (7) and algorithm 1, the speed data of a = -0.40 during the speed range of 75% → 100% is converted into vibration data. The specific steps are as follows: first, extract the speed data of the time period [2s, 5s] in the speed response curve, then extract the feature value T through formula (4), and finally use algorithm 1 to obtain the fault data s'.

[0083]

[0084]

[0085] Final fault data as follows Figure 6 As shown.

[0086] Then, a switching DBN model is designed. The DBN network structure is as follows: Figure 7 As shown, the system consists of an input layer, a hidden layer, and an output layer. The input layer takes fault data of the vibration type as input. The hidden layer consists of multiple Restricted Boltzmann Machines (RBMs) to extract features from the fault data. The output layer consists of a Backpropagation (BP) network to classify the features of the fault data and output the corresponding degree of jamming fault.

[0087] This invention trains the DBN model using fault data from different speed ranges. After training, the DBN has corresponding network parameters for fault data in different speed ranges, namely w(n). i →n j This completes the design of the switching DBN model. The working process of the switching DBN is as follows: Figure 8As shown, the speed range is first determined based on the throttle lever angle. For different speed ranges, the DBN selects the corresponding network parameters. Finally, the fault data of the vibration data type is input into the DBN to diagnose the fault degree, i.e., the value of the feedforward term.

[0088] Step 3: Perform SFMV fault diagnosis based on simulation data;

[0089] The fault data of fuel metering valve sticking were obtained using a joint simulation model of MATLAB and Amesim. In this invention, the fault diagnosis effect of the switching DBN is illustrated using the speed increase process from 75% to 100% and the speed decrease process from 100% to 75% as examples. During the speed increase process, the feedforward terms were set to 0, -0.40, -0.45, and -0.47; during the speed decrease process, the feedforward terms were set to 0, 0.45, 0.48, and 0.50.

[0090] The following rotational speed data were obtained through simulation: data(75%→100%, 0), data(75%→100%, -0.40), data(75%→100%, -0.45), data(75%→100%, -0.47), data(100%→75%, 0), data(100%→75%, 0.45), data(100%→75%, 0.48), data(100%→75%, 0.50). The rotational speed response curves are shown in Figures 5(a) and 5(b).

[0091] Next, the rotational speed data is converted into vibration data. The conversion results of data (75% → 100%) are shown in Figure 9, and the conversion results of data (100% → 75%) are shown in Figure 10. Then, a switching DBN is trained. First, the four feedforward terms during the rotational speed increase process are replaced with labels 1, 2, 3, and 4 from smallest to largest, and the same applies during the rotational speed decrease process. Next, the switching DBN is trained using fault data of the vibration data type and the corresponding labels. Finally, the trained network is tested using fault data.

[0092] Table 3 shows the fault diagnosis results for the speed increase and speed decrease processes. The accuracy of the training set for both processes is 100%, and the accuracy of the test set is above 99%, which demonstrates the effectiveness of the switching DBN model.

[0093] Table 1. Rotational speed 0.75 rpm max →n max process and n max →0.75n max Fault diagnosis results of the process

[0094]

[0095] The embodiments described above merely illustrate the implementation method of the present invention, but should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention.

Claims

1. A fault diagnosis method for stuck fuel metering valve in an aero-engine, characterized in that, The steps include the following: Step 1: Obtain SFMV fault data for fuel metering valve sticking; Based on the principle of controlling the movement of metering valves through electronic controllers in existing aero engines, by changing the output q of the electronic controller, the oil pressure in the control chamber is changed, SFMV faults are simulated, and aero engine speed data is collected as fault data. SFMV occurs during the change of the metering valve opening Ω. When the desired opening of the metering valve increases, during SFMV simulation, the feedforward term a of the electronic controller is negative, causing the output q of the electronic controller to decrease, thus hindering the increase of the metering valve opening Ω. When the desired opening of the metering valve decreases, during SFMV simulation, the feedforward term a of the electronic controller is positive, causing the output q of the electronic controller to increase, thus hindering the decrease of the metering valve opening Ω. The feedforward term is a continuous value. Several specific feedforward terms are set, and these specific feedforward terms replace the feedforward terms within the range r on both sides, thus using these specific feedforward terms to represent all degrees of jamming faults; fault data under different operating conditions are acquired; the engine speed is represented by n, and the speed data is divided into two processes: speed increase and speed decrease; during the speed increase n... + During the process, the feedforward term 'a' of the electronic controller is negative, and 'a' is used as the basis for this process. + This indicates that as the rotational speed decreases by n... - During the process, the feedforward term 'a' of the electronic controller is positive, and 'a' is used as the basis for this process. - express; During the process of increasing the rotational speed, the rotational speed range [0, n max is divided into s + 1 small intervals, that is, s + 1 working conditions. Use to represent the interval where the rotational speed increases from to . i1, j1 ∈ {1, 2,..., s} and 1 ≤ i1 < j1 < s; Set m feedforward term values of different sizes to represent all degrees of jamming faults. Use to represent, l1 ∈ {1, 2,..., m}, and obtain (s + 1) * m groups of different fault data During the process of speed reduction, the speed range [0, n max is divided into h + 1 small intervals, and is used to represent the interval where the speed decreases from to , where i2, j2 ∈ {1, 2,..., h} and 1 ≤ i2 < j2 < h; g feedforward term values of different magnitudes are set and represented by , where l2 ∈ {1, 2,..., g}, and (h + 1) * g groups of fault data Step 2: Use a DBN-based stuck fault diagnosis method to diagnose fault data; First, the fault data is processed into vibration data type, as shown in equations (1), (2), (3), (4), and (5): When the speed increases, the fault data processing is as follows; Where [t1, t2] refers to the value fed forward during the speed increase process. The period with the greatest impact; Mean represents the impact on rotational speed data. [t1, t2] take the average value; Equation (1) means taking the rotational speed data [t1, t2] are transformed into eigenvalues ​​T; When the rotational speed decreases, equation (1) is rewritten as equation (2): Where [t1, t2] refers to the feedforward value during the speed reduction process. The period with the greatest impact, Equation (2) represents the time when the rotational speed is reduced. Transform into eigenvalue T; Substituting the eigenvalue T into formula (3), we obtain the number of sampling points N in a single period. T : N T =[f s *T](3) where, f s The sampling frequency is [ ], where f represents the frequency. s *T is an integer; N T Substituting into formula (4), we obtain the fault data matrix s uv : Where A0, C, f s f r f n p are constants; matrix s uv Convert to a one-dimensional array, let d = [b0, b1, ..., b p-1 ], to obtain one-dimensional fault data d; add Gaussian white noise wgn to obtain fault data s' of vibration data type, as shown in equation (5): After converting the data type s'=d+wgn(5), a switching DBN model is proposed for SFMV under different working conditions. A network parameter selection module is added before the DBN model. When diagnosing fault data, the network parameter selection module assigns the corresponding network parameters to the DBN model according to the speed range of the fault data. The switching DBN model was trained using fault data from different speed ranges to generate the corresponding network parameters w(n). i →n j ), w(n i →n j ) indicates that when the speed range is n i →n j The network parameters used by the switching DBN model at that time; After the switching DBN model is trained, SFMV fault diagnosis is performed. The speed range in which the SFMV occurs is determined based on the throttle lever angle. For different speed ranges, the switching DBN model selects the corresponding network parameters. The fault data of the vibration type is input into the switching DBN model. Finally, the switching DBN outputs the degree of SFMV fault, i.e., the feedforward term a.

2. The fault diagnosis method for fuel metering valve jamming in aero-engines according to claim 1, characterized in that, The interval range r is 0.02.

Images