Aircraft control surface oscillation fault monitoring method based on multi-scale wavelet transform

Abstract

A method for monitoring aircraft control surface oscillation faults based on multi-scale wavelet transform is proposed. The autopilot calculates the control surface command δ using a PID control law based on the input flight path command. desired Based on the nonlinear model and control commands, the predicted rudder deflection value δ is generated. est The predicted deflection value of the control surface is compared with the actual deflection value δ collected by the sensor module. mea The residual signal R is generated by subtraction. n The residual signal is decomposed into several sub-band signals using wavelet decomposition. Based on wavelet energy calculation of the sub-bands that may contain control surface oscillation fault signals, the occurrence of a fault is determined by judging the magnitude of an energy comparison threshold, which is determined through Monte Carlo simulation. This invention separates the command signal by generating residuals and performs wavelet transform processing on the residual signals. This reduces the interference of noise and irrelevant frequency signals on monitoring performance. Furthermore, by using the wavelet energy of the sub-bands as a fault criterion, it can effectively monitor small-amplitude oscillation signals, significantly improving the monitoring coverage of oscillation faults.

Description

Technical Field

[0001] This invention relates to a technology in the field of aircraft manufacturing, specifically a method for monitoring aircraft control surface oscillation faults based on multi-scale wavelet transform. Background Technology

[0002] For aircraft using fly-by-wire flight control systems, the exceptionally complex operating environment during flight can lead to severe uncommanded oscillations during takeoff, landing, or cruise. Oscillation faults in the main flight control system control surfaces, such as ailerons, elevators, and rudders, are primarily caused by pseudo-sinusoidal command signals generated by electronic components in fault modes. This results in corresponding periodic vibrations on the main flight control system control surfaces, leading to control surface oscillation faults. The impact of control surface oscillation faults on the aircraft is mainly manifested in its effect on the airframe structure and flight handling characteristics. Currently, commonly used algorithms for control surface oscillation monitoring include: sensor signal-based methods, data-driven methods, and model-based methods. Sensor signal-based methods require hardware redundancy; data-driven methods are sensitive to noise and struggle to simultaneously cover all frequency bands; and model-based methods require building specific models based on flight principles and component parameters. Summary of the Invention

[0003] To address the aforementioned shortcomings of existing technologies, this invention proposes a method for monitoring aircraft control surface oscillation faults based on multi-scale wavelet transform. By generating residuals to separate command signals and performing wavelet transform processing on the residual signals, the interference of noise and irrelevant frequency signals on monitoring performance is reduced. Furthermore, by using sub-band wavelet energy as a fault criterion, it can effectively monitor small-amplitude oscillation signals and significantly improve the monitoring coverage of oscillation faults.

[0004] This invention is achieved through the following technical solution:

[0005] This invention relates to a method for monitoring aircraft control surface oscillation faults based on multi-scale wavelet transform, comprising:

[0006] Step 1: Receive control surface commands generated by the autopilot ;

[0007] The aforementioned control commands are calculated by the autopilot using a PID control law based on the input trajectory command.

[0008] Step 2: Generate predicted control surface deflection values ​​based on the nonlinear model and control surface commands. The predicted deflection value of the control surface is compared with the actual deflection value collected by the sensor module. Difference generates residual signal The residual signal is decomposed into several sub-band signals by wavelet decomposition.

[0009] The predicted control surface deflection value is obtained by constructing a nonlinear electro-hydraulic servo actuation system dynamic model based on the working principle and parameter information of the electro-hydraulic servo actuation system, and predicting it according to the input control surface command. This nonlinear electro-hydraulic servo actuation system dynamic model includes: an autopilot, an electro-hydraulic servo valve unit, a symmetrical hydraulic cylinder unit, a load and control surface unit, and a sensor unit. Specifically: the autopilot generates a servo valve drive signal by comparing the desired position with the actual feedback position based on the control surface deflection command information of the aircraft; the electro-hydraulic servo valve unit converts current into mechanical force through a torque motor based on the servo valve drive signal, drives the valve core to move, and adjusts the flow rate of high-pressure oil to the hydraulic cylinder; the pressure difference between the two hydraulic cylinders in the symmetrical hydraulic cylinder unit due to the different flow rates pushes the piston rod connected to the control surface load, drives the control surface, and converts the hydraulic flow into a force acting on the control surface; the load and control surface unit is subjected to hydraulic thrust and completes the control surface deflection under the action of internal hydraulic load and external aerodynamic load; finally, the sensor unit detects the control surface deflection in real time and feeds it back to the autopilot.

[0010] The sub-band signal is generated by using the predicted and actual deflection values ​​of the control surface to form a residual signal. After sampling at a frequency of 40Hz, the discrete residual signal is captured through a time window with a length of 8 steps. The signal is then decomposed by wavelet to obtain 8 sub-band signals with frequency distributions from low to high.

[0011] Step 3: Perform wavelet energy calculation on the sub-bands that may contain rudder surface oscillation fault signals, and determine whether a fault has occurred by judging the magnitude of the energy comparison threshold. The judgment threshold is determined by Monte Carlo simulation.

[0012] The wavelet energy calculation selects all sub-frequency bands within the frequency range of 0.5-10Hz, calculates the sum of wavelet energy over 8 time steps based on the wavelet amplitude of the sub-frequency bands, and uses whether the sum of wavelet energy exceeds a threshold determined by Monte Carlo simulation as a criterion for the occurrence of a fault.

[0013] The Monte Carlo simulation refers to performing simulations under fault-free normal operating conditions and recording the maximum wavelet energy in the 0.5-10Hz frequency band of the residual signal after each simulation. After accumulating 1000 simulation samples, the mean and standard deviation are calculated. The Z critical value is queried according to the confidence level, the confidence interval is calculated, and the upper bound of the confidence interval is used as the judgment threshold of wavelet energy.

[0014] The simulation refers to: based on the constructed electro-hydraulic servo actuation system, applying hydraulic differential pressure to the model... and actuator damping coefficient Uncertainty is introduced into the parameter settings to improve the robustness and applicability of the model.

[0015] This invention relates to an aircraft control surface oscillation fault monitoring system based on multi-scale wavelet transform, which implements the above-mentioned method. The system includes: a signal preprocessing module, a feature extraction module, and an oscillation fault diagnosis module. The signal preprocessing module receives the desired control surface deflection angle output by the autopilot and the actual control surface deflection angle measured by the sensor. Based on a pre-built nonlinear actuator model and the desired control surface deflection angle, it generates a predicted value for the control surface deflection angle. The difference between the predicted value and the sensor measurement value is used to obtain a residual signal. The feature extraction module performs multi-scale wavelet decomposition on the processed residual signal and calculates the wavelet energy of sub-frequency bands that may contain fault signals. The fault diagnosis module compares the wavelet energy of each sub-frequency band with the corresponding judgment threshold. If the threshold is exceeded, a control surface oscillation problem is considered to exist in the servo actuator system.

[0016] Technical effect

[0017] This invention performs multi-scale wavelet decomposition on the residual signal and calculates the wavelet energy of the frequency bands where control surface oscillation signals may be distributed in the decomposition results. By comparing the wavelet energy with a judgment threshold, the occurrence of a fault is diagnosed. Compared with existing technologies, this invention significantly improves the fault coverage of traditional amplitude oscillation counting methods, and the performance of the improved control surface oscillation fault monitor is verified through Monte Carlo simulation. This scheme improves monitoring accuracy while ensuring the practicality and reliability of the system, providing strong protection for the safety of fly-by-wire flight control systems. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the invention;

[0019] Figure 2 This is a schematic diagram of the hydraulic servo actuator of the present invention;

[0020] Figure 3 This is a structural block diagram of the rudder surface oscillation monitoring device of the present invention;

[0021] Figure 4 This is a logic diagram of the rudder surface oscillation monitoring device of the present invention;

[0022] Figure 5 A schematic diagram of sample data for determining the simulation threshold in Monte Carlo simulation;

[0023] Figure 6 This is a schematic diagram illustrating the residual curve and monitoring status when a solid control surface oscillation fault occurs.

[0024] In the figure: (a) from top to bottom are the measured and estimated values ​​of the control surface deflection under solid failure conditions, the generated residual signal and the curve of the control surface oscillation monitor to determine whether to issue a warning; (b) from top to bottom are the wavelet energy curves of sub-band 1 and sub-band 2 after wavelet decomposition.

[0025] Figure 7 This is a schematic diagram illustrating the residual curve and monitoring status when a liquid control surface oscillation fault occurs. Detailed Implementation

[0026] This embodiment relates to a method for monitoring aircraft control surface oscillation faults based on multi-scale wavelet transform. By comparing the predicted signal with the sensor-acquired signal to generate residuals, the residuals are processed for data analysis and fault diagnosis to identify control surface oscillation faults, thereby improving the coverage of control surface oscillation faults while ensuring real-time performance.

[0027] like Figure 1 The diagram illustrates possible sources and modes of the aircraft control surface oscillation fault. The fly-by-wire flight control system includes a flight control computer and control surface servo actuators; in other embodiments, the fly-by-wire system may include more or fewer flight control electronics.

[0028] The flight control computer includes a monitoring channel and a control channel. The monitoring channel is used to detect various faults in the servo circuit, and the detection of control surface oscillation faults is also completed in the monitoring channel. The control channel can calculate control surface commands based on the input signals according to the control law and output them to the servo actuator control electronics to control the aircraft control surfaces.

[0029] Control surface oscillation faults are primarily caused by malfunctions in electronic or mechanical components, such as the generation of spurious signals (e.g., sine waves). Based on the different manifestations of the fault, they can be divided into two categories:

[0030] Liquid-type fault: The oscillation signal is superimposed on the original control signal.

[0031] Solid-state failure: The oscillation signal completely replaces the original control signal.

[0032] These two types of oscillation signals are transmitted to the control surfaces through the servo control loop, causing control surface oscillation. The sources of control surface oscillation include: the analog input / output interface (ADC), the hydraulic piston rod position sensor, the servo current amplifier, the control surface sensor, and the flight control computer. These components generate oscillation pseudo-command signals, which are transmitted to the control surfaces through the actuator servo control loop, resulting in control surface oscillation faults.

[0033] like Figure 2 The diagram illustrates the working principle of a hydraulic servo actuator: The first-order derivative of the lever displacement of the servo actuator is obtained. This refers to the actuator rod displacement rate. After receiving the input electrical signal, the servo valve changes the valve core displacement, controlling the flow rate within the chamber, thus generating pressures at hydraulic chambers A and B respectively. and The hydraulic pressure is calculated by subtracting the two values ​​to obtain the hydraulic pressure difference. , The surface area of ​​the actuator piston, multiplied by the pressure difference, equals the force acting on the piston rod. This force pushes the piston rod to produce displacement, causing the control surface to deflect. The actual rod displacement is... Actual servo actuator surface deflection The transformation relationship can be obtained by interpolation and table lookup, and is generally related to the actual installation angle.

[0034] like Figure 3 As shown in the figure, this embodiment relates to an aircraft control surface oscillation fault monitoring system based on multi-scale wavelet transform, which includes: a control surface servo control loop and a control surface oscillation monitor. The control surface servo control loop is the actuation system of the fly-by-wire flight control civil aircraft system when performing trajectory tracking, and is shown here in a simplified flowchart. The function of the control surface oscillation monitor is to determine the fault occurrence based on the autopilot's expected deflection signal and the control surface sensor's collected signal when the actuator or sensor experiences a control surface oscillation fault during the operation of the control surface servo control loop.

[0035] The control surface servo control loop includes: an autopilot, an electro-hydraulic servo valve unit, a symmetrical hydraulic cylinder unit, a load and control surface unit, and a sensor unit. Specifically: the autopilot compares the desired position with the actual feedback position based on the control surface deflection command information from the aircraft, generating a servo valve drive signal; the electro-hydraulic servo valve unit, based on the servo valve drive signal, converts current into mechanical force via a torque motor, driving the valve core to move and adjusting the flow rate of high-pressure oil to the hydraulic cylinder; the symmetrical hydraulic cylinder unit, based on the pressure difference between the two hydraulic cylinders, pushes the piston rod connected to the control surface load, driving the control surface and converting the hydraulic flow into a force acting on the control surface. The load and control surface unit, subjected to hydraulic thrust, completes the control surface deflection under the action of internal hydraulic load and external aerodynamic load; finally, the sensor unit detects the control surface deflection in real time and feeds it back to the autopilot.

[0036] The control surface oscillation monitor includes a signal preprocessing module, a feature extraction module, and an oscillation fault diagnosis module. The signal preprocessing module receives the desired control surface deflection angle output by the autopilot and the actual control surface deflection angle measured by the sensor. Based on a pre-built nonlinear actuator model and the desired control surface deflection angle, it generates a predicted value for the control surface deflection angle and subtracts the predicted value from the sensor measurement value to obtain a residual signal. The feature extraction module performs multi-scale wavelet decomposition on the processed residual signal and calculates the wavelet energy of sub-frequency bands that may contain fault signals. The fault diagnosis module compares the wavelet energy of the sub-frequency bands with the corresponding judgment thresholds for each frequency band; if the threshold is exceeded, a control surface oscillation problem is considered to exist in the servo actuation system.

[0037] The control surface oscillation fault sources considered in this invention are oscillation faults in the stick displacement sensor signal and oscillation faults in the servo actuator input signal. Furthermore, noise interference experienced by the sensor during operation, disturbances received during normal aircraft operation, and model uncertainties of the servo actuator are also considered.

[0038] like Figure 4 As shown, when the actuator receives the control command to deflect the control surface, it deflects the control surface and uses sensors to measure the actual stick displacement. At the same time, the monitor calculates the estimated stick displacement based on the control command and the pre-built actuator dynamics model.

[0039] The residual signal is generated based on the estimated and measured values. Wavelet decomposition is then performed on the residual signal, which decomposes the signal into low-frequency and high-frequency components at different levels. Sub-band signals with frequency components around 0.5-10Hz are selected, and wavelet energy calculation is performed on these signals. The magnitude of the wavelet energy within the frequency band is compared with a pre-set threshold to determine the fault situation and decide whether to trigger an early warning or continue operation.

[0040] like Figure 5 The image shows sample data for determining the simulation threshold in Monte Carlo simulations. It records the maximum wavelet energy of the two sub-bands during each simulation, demonstrating the distribution of wavelet energy. The initial simulation conditions are: normal system operation, no rudder surface oscillation fault, and the sensor is affected by white noise with a variance of 0.0005. Model uncertainty is determined by setting the model hydraulic differential. Within the random range [16, 30] and the actuator damping coefficient [6.8,10], with 1000 simulations.

[0041] Calculate the sample mean based on the statistical results. , and standard deviation , Using the upper bound of the 99.9999% confidence space as the monitoring threshold, the final determination was made. , As the wavelet energy threshold for the two sub-bands.

[0042] After such Figure 6 The specific experiment shown illustrates how, under the condition of a civil aircraft executing trajectory tracking commands, a solid-state or liquid-state fault of 1-10Hz is injected into the sensor module. When a solid-state oscillation fault occurs, the residual curve and monitoring results are illustrated. The control surface oscillation type is a solid-state fault, injected into the sensor output signal to simulate a sensor fault. The fault injection time is 40 seconds, the amplitude is 1 mm, and the fault frequency is 5Hz. Using this method, the oscillation fault in the servo actuation loop can be detected within three sampling steps after the fault occurs.

[0043] like Figure 6 (a) and Figure 6 As shown in (b), when a solid fault occurs in the sensor after 40 seconds, the measured and estimated values ​​of the control surface deflection deviate. Since the solid fault is replaced by an oscillating signal, the residual and wavelet energy can both reflect whether an oscillating fault has occurred. The energy in both frequency bands exceeds the threshold, and the control surface oscillation monitor triggers an early warning. The fault signal is set from 0 to 1 after 40 seconds. It can be seen that the control surface oscillation monitoring method that introduces wavelet energy calculation on the basis of the model-based method can effectively detect the oscillation faults in the system.

[0044] like Figure 7 The image shows an example of the residual curve and monitoring status when a liquid oscillation fault occurs. The oscillation type of the control surface is a liquid fault. A fault is injected into the sensor output signal to simulate a sensor fault. The fault injection time is 40 seconds, the amplitude is 1 mm, and the fault frequency is 5 Hz.

[0045] like Figure 7 As shown in (a), from top to bottom, the measured and estimated values ​​of the control surface deflection under liquid failure conditions are compared with the generated residual signal and the curve used by the control surface oscillation monitor to determine whether a warning is needed; Figure 7 As shown in (b), from top to bottom, are the wavelet energy curves of sub-band 1 and sub-band 2 after wavelet decomposition.

[0046] like Figure 7 As shown, when a sensor liquid fault occurs after 40 seconds of operation, the measured and estimated values ​​of the control surface deflection deviate. Since the amplitude of the liquid fault is small at this time, it is difficult to determine based on the residual signal alone. However, the wavelet energy amplitude of frequency band 2 significantly exceeds the judgment threshold. The rudder surface oscillation monitor triggers an early warning, and the fault signal is set from 0 to 1 in 40 seconds. This shows that the rudder surface oscillation monitoring method that introduces wavelet energy calculation on the basis of the model-based method can effectively improve the fault coverage of fault detection.

[0047] Table 1 shows the minimum fault amplitude that the control surface oscillation fault monitor can detect, demonstrating the detection capability of the minimum fault amplitude.

[0048] Table 1

[0049] Fault location Fault type Frequency range Noise intensity Operating conditions Minimum monitoring amplitude actuator liquid 1-10Hz 0.0005 Tracking 0.09mm sensor liquid 1-10Hz 0.0005 Tracking 0.05mm

[0050] Since solid-state fault detection is independent of amplitude, only liquid-state faults are considered here. Through simulation experiments, it was found that when the frequency of the oscillation signal is 0.5-10Hz, and considering white noise interference with a variance of 0.0005, the oscillation fault monitoring device designed in this invention can detect servo actuator faults with a minimum amplitude of 0.09mm and sensor faults with a minimum amplitude of 0.05mm when a fault occurs during aircraft trajectory tracking, and has good fault monitoring coverage.

[0051] Compared with existing technologies, this invention introduces multi-scale wavelet decomposition in the feature extraction module, which greatly improves the robustness of the monitor to noise and interference. In the fault diagnosis module, wavelet energy is introduced as a fault diagnosis indicator, which can effectively detect early small-amplitude oscillations and has a higher coverage of faults than traditional methods.

[0052] The above-mentioned specific implementation can be partially adjusted in different ways by those skilled in the art without departing from the principles and purpose of the present invention. The scope of protection of the present invention shall be based on the claims and shall not be limited by the above-mentioned specific implementation. All implementation schemes within its scope shall be subject to the constraints of the present invention.

Claims

1. A method for monitoring aircraft control surface oscillation faults based on multi-scale wavelet transform, characterized in that, include: Step 1: The autopilot calculates the control surface commands based on the input trajectory commands using a PID control law. ; Step 2: Generate predicted control surface deflection values ​​based on the nonlinear model and control surface commands. The predicted deflection value of the control surface is compared with the actual deflection value collected by the sensor module. Difference generates residual signal The residual signal is decomposed into several sub-band signals by wavelet decomposition. Step 3: Perform wavelet energy calculation on the sub-bands that may contain rudder surface oscillation fault signals, and determine whether a fault has occurred by judging the magnitude of the energy comparison threshold. The judgment threshold is determined by Monte Carlo simulation. The sub-band signal is generated by using the predicted and actual deflection values ​​of the control surface to form a residual signal. After sampling at a frequency of 40Hz, the discrete residual signal is captured through a time window with a length of 8 steps. The signal is then decomposed by wavelet to obtain 8 sub-band signals with frequency distributions from low to high. The wavelet energy calculation selects all sub-frequency bands within the frequency range of 0.5-10Hz, calculates the sum of wavelet energy within 8 time steps based on the wavelet amplitude of the sub-frequency band, and uses whether the sum of wavelet energy exceeds the threshold determined by Monte Carlo simulation as the criterion for the occurrence of a fault. The Monte Carlo simulation refers to performing simulations under fault-free normal operating conditions and recording the maximum wavelet energy in the 0.5-10Hz frequency band of the residual signal after each simulation. After accumulating 1000 simulation samples, the mean and standard deviation are calculated. The Z critical value is queried according to the confidence level, the confidence interval is calculated, and the upper bound of the confidence interval is used as the judgment threshold of wavelet energy.

2. The method for monitoring aircraft control surface oscillation faults according to claim 1, characterized in that, The predicted control surface deflection value is obtained by constructing a nonlinear electro-hydraulic servo actuation system dynamic model based on the working principle and parameter information of the electro-hydraulic servo actuation system, and predicting it according to the input control surface command. This nonlinear electro-hydraulic servo actuation system dynamic model includes: an autopilot, an electro-hydraulic servo valve unit, a symmetrical hydraulic cylinder unit, a load and control surface unit, and a sensor unit. Specifically: the autopilot generates a servo valve drive signal by comparing the desired position with the actual feedback position based on the control surface deflection command information of the aircraft; the electro-hydraulic servo valve unit converts current into mechanical force through a torque motor based on the servo valve drive signal, drives the valve core to move, and adjusts the flow of high-pressure oil to the hydraulic cylinder; the symmetrical hydraulic cylinder unit pushes the piston rod connected to the control surface load according to the pressure difference between the two hydraulic cylinders, drives the control surface, and converts the hydraulic flow into a force acting on the control surface. The load and control surface unit are subjected to hydraulic thrust, and the control surface deflection is completed under the action of internal hydraulic load and external aerodynamic load; finally, the sensor unit detects the control surface deflection in real time and feeds it back to the autopilot.

3. The method for monitoring aircraft control surface oscillation faults according to claim 1, characterized in that, The simulation refers to: based on the constructed electro-hydraulic servo actuation system, applying hydraulic differential pressure to the model... and actuator damping coefficient Uncertainty is introduced into the parameter settings to improve the robustness and applicability of the model.

4. A multi-scale wavelet transform-based aircraft control surface oscillation fault monitoring system according to any one of claims 1-3, characterized in that, include: The system comprises a signal preprocessing module, a feature extraction module, and an oscillation fault diagnosis module. The signal preprocessing module receives the desired control surface deflection angle output by the autopilot and the actual control surface deflection angle measured by the sensor. Based on a pre-built nonlinear actuator model and the desired control surface deflection angle, it generates a predicted value for the control surface deflection angle. The difference between the predicted value and the sensor measurement value yields the residual signal. The feature extraction module further divides the system into three modules: a signal preprocessing module, a feature extraction module, and an oscillation fault diagnosis module. The extraction module performs multi-scale wavelet decomposition on the processed residual signal and calculates the wavelet energy of the sub-frequency bands that may contain fault signals. The fault diagnosis module compares the wavelet energy of the sub-frequency bands with the judgment thresholds corresponding to each frequency band. If the threshold is exceeded, it is considered that there is a rudder surface oscillation problem in the servo actuation system.

Images