Fault monitoring device and method for aviation servo actuation test bench

Abstract

The invention provides a fault monitoring device and method for an aviation servo actuation test bench, and belongs to the technical field of aircraft detection.The fault monitoring device comprises a hydraulic test system and a control system, the hydraulic test system is used for simulating faults of a servo actuation system and collecting data through various sensors, and the control system is used for controlling the hydraulic test system; and the control system is used for controlling the hydraulic test system to reach required monitoring conditions and collecting monitoring data, is provided with a deep learning model, carries out training through collected fault simulation data, and is used for comprehensively analyzing and effectively detecting various fault types after training. According to the invention, the reliability and safety of the servo system are obviously improved, and the method is of great significance for ensuring the performance of an aircraft and carrying out preventive maintenance.

Description

Technical Field

[0001] The invention belongs to the technical field of aircraft detection, and relates to a fault monitoring device and method for an aviation servo actuation test bench. Background Art

[0002] In the modern aviation industry, servo actuators play a vital role. Servo actuators are widely used in flight control, landing gear operation and other critical flight systems. In order to ensure flight safety and reliability, the actuators must be fully and carefully monitored and diagnosed before aircraft assembly. Fault monitoring and simulation are of great significance for identifying potential problems, preventing accidents, reducing maintenance costs and improving the overall performance of the system.

[0003] In order to achieve efficient fault monitoring, a variety of sensors are needed for comprehensive data collection. These sensors include pressure sensors, temperature sensors, acceleration sensors, and position sensors. Working together, a variety of sensors can provide rich data information, which helps to fully understand the operating status of the servo actuator system. By analyzing this data, subtle anomalies and trends can be identified, so as to detect faults early. In practical applications, an intelligent fault monitoring system can be built by combining multiple sensors with machine learning technology. This system can not only monitor the working status of the servo actuator in real time, but also continuously learn and optimize the data to improve the accuracy and efficiency of fault detection. For example, when the pressure sensor detects an abnormal fluctuation, the system can combine the data of the temperature sensor to determine whether it is a normal reaction caused by temperature changes or a potential fault signal. The data of the acceleration sensor and position sensor can help determine the specific location and nature of the fault, providing precise guidance for maintenance personnel. Summary of the invention

[0004] The present invention provides an aviation servo actuation test bench fault monitoring device and method, which is used for monitoring the aircraft servo actuation system before assembly. The present invention monitors the hydraulic servo actuation system through multiple sensors such as vibration, displacement, load, power, pressure, flow and temperature, and intelligently judges faults in combination with machine learning methods.

[0005] The technical solution adopted by the present invention is as follows:

[0006] A fault monitoring device for an aviation servo actuation test bench, comprising a hydraulic test system and a control system, wherein the hydraulic test system is used to simulate a servo actuation system fault and collect data through various sensors, and the control system is used to control the hydraulic test system to achieve the required monitoring conditions, collect monitoring data and output monitoring results.

[0007] The hydraulic test system comprises an oil tank 1, a filter 2, a hydraulic pump 3, three solenoid valves, three adjustable flow valves, a pilot electromagnetic overflow valve 10, an accumulator 11, a servo valve 12, a safety valve 13, a load 14 and a monitoring sensor group; the filter 2 is connected to the oil outlet end of the oil tank 1, the hydraulic pump 3 is connected to the oil outlet end of the filter 2, the oil outlet end of the hydraulic pump 3 is connected to the adjustable flow valve A7 and the servo valve 12 in sequence, the servo valve 12 is connected to the two sides of the piston of the actuator cylinder in the servo actuator system 25 to be monitored through two pipelines, the servo valve 12 is used to accurately control the movement position and speed of the actuator cylinder in the servo actuator system 25, its oil outlet end is connected to the oil tank 1, and the hydraulic oil after use flows back to the oil tank 1, the two sides of the actuator cylinder piston are connected with pipelines, and the solenoid valve A4 and the adjustable flow valve B8 are arranged on the pipelines, the load 14 weighs 250kg, and acts on the top of the actuator cylinder in the servo actuator system 25 when in use, and is used to simulate the leakage fault in the actuator cylinder; the hydraulic pump 3 and the adjustable flow valve A There are two branches on the pipeline between the adjustable flow valve A7 and the servo valve 12, one of which is connected with the solenoid valve B5 and the adjustable flow valve C9 in sequence, the oil outlet of the adjustable flow valve C9 is connected to the oil tank 1, and is used to simulate the hydraulic pump leakage fault, and the other branch is provided with a pilot electromagnetic overflow valve 10, the oil outlet of which is connected to the oil tank 1, the pilot electromagnetic overflow valve 10 is normally open, and is used to accurately adjust the overflow pressure in the hydraulic test system; there are also two branches on the pipeline between the adjustable flow valve A7 and the servo valve 12, one of which is provided with a safety valve 13, the oil outlet of the safety valve 13 is connected to the oil tank 1, the safety valve 13 is normally open, and is used to protect the hydraulic test system from damage caused by excessive pressure, and the other branch is connected with the solenoid valve C6 and the accumulator 11 in sequence, and is used to store and release hydraulic energy to balance the system pressure and supplement the flow, and the solenoid valve C 6 is normally open; each solenoid valve in the hydraulic test system is used as a pipeline switch to determine the connection and closing of the pipeline, and each adjustable flow valve is used to adjust the liquid flow of the pipeline; the monitoring sensor group includes multiple flow meters 15, multiple pressure gauges 16, multiple pressure sensors 17, multiple temperature sensors 18, liquid level sensors 19, power sensors 20, vibration sensors 21, optical fiber strain sensors 22, force sensors 23, and displacement sensors 24, wherein: a liquid level sensor 19 and a temperature sensor 18 are set at the oil tank 1, respectively used to monitor the liquid level and temperature of the hydraulic oil in the oil tank 1; a power sensor 20, a vibration sensor 21 and a temperature sensor 18 are set at the hydraulic pump 3, respectively used to monitor the power consumption, vibration and temperature of the hydraulic pump 3; pressure gauges 16 are set at both ends of the adjustable flow valve A 7 for reading the pressure value; pressure sensors 17 and temperature sensors 18 are set on the two pipelines connecting the servo valve 12 and the actuator in the servo actuation system 25; a flow meter 15 is set on the pipeline where the solenoid valve A 4 and the adjustable flow valve B 8 are set;In the servo actuation system 25, an optical fiber strain sensor 22, a force sensor 23 and a displacement sensor 24 are arranged on the actuating cylinder, wherein the optical fiber strain sensor 22 is used to monitor the stress change caused by leakage in the actuating cylinder, the force sensor 23 is used to monitor the weight of the load 14 on the actuating cylinder, and the displacement sensor 24 is used to monitor the movement displacement of the actuating cylinder. ;

[0008] The control system is used to control the hardware in the hydraulic test system to simulate various servo actuation system failures, and at the same time collect monitoring data from the monitoring sensor group in the hydraulic test system. The control system is equipped with a deep learning model, which uses the monitoring data of simulated failures as input and the failure type as output for training. After training, it is used to comprehensively analyze and effectively detect various failure types. After training, the accuracy of the deep learning model is greater than 95%.

[0009] The deep learning model first pre-processes the collected data and divides it into time series data and spatial data, and inputs the time series data into the LSTM network, while the spatial data is input into the CNN network for processing; the LSTM network includes two LSTM layers, the first LSTM layer extracts preliminary features of the time series data, which involve time dependencies and dynamic changes such as flow changes and power fluctuations; then, the second LSTM layer further processes these features to enhance the understanding of time series patterns to capture more complex long-term and short-term time dependencies; the CNN network includes a convolutional layer 1, a pooling layer, a convolutional layer 2, and a flattening layer, the convolutional layer 1 first extracts local features from the spatial data, such as pressure distribution and stress changes, and then reduces the dimension of the feature map through the pooling layer to retain the main Features and reduce computational complexity. Then, the convolution layer 2 further extracts high-level spatial features to enhance the recognition ability of complex spatial patterns. Finally, the flattening layer flattens the two-dimensional convolution output into one-dimensional data to prepare for the subsequent processing of the fully connected layer. The output features of the LSTM network and the CNN network are spliced ​​through the fusion step to form a comprehensive feature vector. This feature vector is input to the fully connected layer 1 for high-dimensional feature mapping and integration, and the Dropout layer is applied during the training process to prevent over-fitting. The fully connected layer 2 further processes the fused features and extracts the final high-level features. The output layer is a classification output layer, which is used to map the feature vector to different classification labels to detect servo actuator system faults. If necessary, a regression output layer can be added to predict the severity of the fault or other continuous values.

[0010] The servo actuation system faults that can be simulated and detected by the fault monitoring device include hydraulic pump leakage, solenoid valve delay and leakage in the actuator cylinder. Other faults can also be added, such as actuator cylinder shaking, hydraulic pump 3 shaking, etc.

[0011] The deep learning model is not limited to LSTM network and CNN network, and can also be other algorithms, such as Transformer.

[0012] Furthermore, the monitoring sensor in the fault monitoring device may also include a current sensor and the like.

[0013] The present invention also provides a fault monitoring method for an aviation servo actuation test bench, which is implemented based on the above-mentioned fault monitoring device. The fault monitoring method comprises the following steps:

[0014] Step 1, determine the data set: connect the actuator to the fault monitoring device, use the fault monitoring device to simulate various faults, and collect monitoring data from various sensors such as vibration, displacement, load, power, pressure, flow and temperature as fault data. In addition, collect the monitoring data of the actuator under normal working conditions as non-fault data, organize these data into time series, and mark the normal state and fault data in different situations. Normalize the data to form a data set as the input item for training the deep learning model, and use non-fault or fault type as the training output item. Among them, the simulated servo actuator system faults include hydraulic pump leakage fault, solenoid valve delay fault and actuator internal leakage fault. The three fault data collection methods are as follows:

[0015] When simulating a hydraulic pump leakage fault, the solenoid valve A4 in the hydraulic test system is closed and the solenoid valve B5 is opened. By controlling the solenoid valve A4, the adjustable flow valve C9 is set to multiple different leakage flow values ​​in sequence to cover various possible operating conditions and leakage conditions. The leakage flow value is monitored by the flow meter 15 on the branch. During the simulation process, the data of each sensor is collected as the hydraulic pump leakage fault data for training and verifying the deep learning model.

[0016] When simulating the solenoid valve delay failure, the flow path setting in the hydraulic test system is the same as when simulating the hydraulic pump leakage failure. The delay of the control signal of the solenoid valve B5 is used for simulation. Multiple different delay options are set to cover various situations, and the solenoid valve B5 is simulated in turn. Under each delay setting, the data of the flow meter 15 of the branch where the solenoid valve B5 is located after it is started and closed is recorded as the solenoid valve delay failure data for training and verifying the deep learning model.

[0017] When simulating the internal leakage fault of the actuator, the solenoid valve A4 is opened, the solenoid valve B5 is closed, the load 14 is carried on the actuator and moves up and down, and the adjustable flow valve B8 is set to multiple different leakage flow values ​​to cover various situations. Under each flow setting, when the internal leakage fault occurs, the data of each sensor is collected as the internal leakage fault data of the actuator to train and verify the deep learning model.

[0018] Step 2, determine the training model: divide the data set into training set and validation set in a ratio of 3:1, set the training threshold, train the deep learning model, and return to training when the training result is less than the threshold. When it is greater than the threshold, the training is completed. After the training is completed, the deep learning model is input into the validation set data for verification, so that the accuracy of the deep learning model after training is greater than 95%.

[0019] Step 3, fault monitoring: connect the servo actuator system 25 to be monitored to the hydraulic test system and monitor the data of each monitoring sensor during the working process in real time, input the data into the deep learning model, and finally determine whether there is a fault and output the fault type to realize fault monitoring, and provide intelligent decision-making support for the maintenance and fault prevention of the servo actuator system 25.

[0020] Beneficial effects of the invention: The invention significantly improves the reliability and safety of the servo system, which is of great significance for ensuring aircraft performance and performing preventive maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a fault monitoring device for the servo actuator test bench;

[0022] Figure 2 This is the schematic diagram of the deep learning model;

[0023] Figure 3 Training and fault monitoring process for deep learning models;

[0024] In the figure: 1 oil tank; 2 filter; 3 hydraulic pump; 4 solenoid valve A; 5 solenoid valve B; 6 solenoid valve C; 7 adjustable flow valve A; 8 adjustable flow valve B; 9 adjustable flow valve C; 10 pilot electromagnetic overflow valve; 11 accumulator; 12 servo valve; 13 safety valve; 14 load; 15 flow meter; 16 pressure gauge; 17 pressure sensor; 18 temperature sensor; 19 liquid level sensor; 20 power sensor; 21 vibration sensor; 22 optical fiber strain sensor; 23 force sensor; 24 displacement sensor; 25 servo actuator system. DETAILED DESCRIPTION

[0025] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. The drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0026] This embodiment provides a fault monitoring device for an aviation servo actuation test bench, such as Figure 1As shown, the fault monitoring device includes a hydraulic test system and a control system, wherein the hydraulic test system is used to simulate servo actuation system faults and collect data through a variety of sensors, including hydraulic pump leakage, solenoid valve delay and actuator cylinder leakage faults, and the control system is used to control the hydraulic test system to achieve the required monitoring conditions, collect monitoring data and output monitoring results.

[0027] The hydraulic test system comprises an oil tank 1, a filter 2, a hydraulic pump 3, three solenoid valves, three adjustable flow valves, a pilot electromagnetic overflow valve 10, an accumulator 11, a servo valve 12, a safety valve 13, a load 14 and a monitoring sensor group; the filter 2 is connected to the oil outlet end of the oil tank 1, the hydraulic pump 3 is connected to the oil outlet end of the filter 2, the oil outlet end of the hydraulic pump 3 is connected to the adjustable flow valve A7 and the servo valve 12 in sequence, the servo valve 12 is connected to the two sides of the piston of the actuator cylinder in the servo actuator system 25 to be monitored through two pipelines, the servo valve 12 is used to accurately control the movement position and speed of the actuator cylinder in the servo actuator system 25, its oil outlet end is connected to the oil tank 1, and the hydraulic oil after use flows back to the oil tank 1, the two sides of the actuator cylinder piston are connected with pipelines, and the solenoid valve A4 and the adjustable flow valve B8 are arranged on the pipelines, the load 14 weighs 250kg, and acts on the top of the actuator cylinder in the servo actuator system 25 when in use, and is used to simulate the leakage fault in the actuator cylinder; the hydraulic pump 3 and the adjustable flow valve A There are two branches on the pipeline between the adjustable flow valve A7 and the servo valve 12, one of which is connected with the solenoid valve B5 and the adjustable flow valve C9 in sequence, the oil outlet of the adjustable flow valve C9 is connected to the oil tank 1 for simulating the hydraulic pump leakage fault, and the other branch is provided with a pilot electromagnetic overflow valve 10, the oil outlet of which is connected to the oil tank 1 for accurately adjusting the overflow pressure in the hydraulic test system; there are also two branches on the pipeline between the adjustable flow valve A7 and the servo valve 12, one of which is provided with a safety valve 13, the oil outlet of the safety valve 13 is connected to the oil tank 1 for protecting the hydraulic test system from damage caused by excessive pressure, and the other branch is connected with the solenoid valve C9 in sequence. 6 and accumulator 11, used to store and release hydraulic energy to balance system pressure and supplement flow; each solenoid valve in the hydraulic test system is used as a pipeline switch to determine the connection and closing of the pipeline, and each adjustable flow valve is used to adjust the liquid flow of the pipeline; the monitoring sensor group includes multiple flow meters 15, multiple pressure gauges 16, multiple pressure sensors 17, multiple temperature sensors 18, liquid level sensors 19, power sensors 20, vibration sensors 21, optical fiber strain sensors 22, force sensors 23, and displacement sensors 24, among which: a liquid level sensor 19 and a temperature sensor 18 are set at the oil tank 1, which are respectively used to monitor the liquid level and temperature of the hydraulic oil in the oil tank 1; a power sensor 20, a vibration sensor 21 and a temperature sensor 18 are set at the hydraulic pump 3, which are respectively used to monitor the power consumption, vibration and temperature of the hydraulic pump 3; pressure gauges 16 are set at both ends of the adjustable flow valve A 7 for reading the pressure value; pressure sensors 17 and temperature sensors 18 are set on the two pipelines connecting the servo valve 12 and the actuator in the servo actuation system 25; a solenoid valve A is set 4 and the pipeline of the adjustable flow valve B 8 are provided with a flow meter 15;In the servo actuation system 25, an optical fiber strain sensor 22, a force sensor 23 and a displacement sensor 24 are arranged on the actuating cylinder, wherein the optical fiber strain sensor 22 is used to monitor the stress change caused by the leakage in the actuating cylinder, the force sensor 23 is used to monitor the weight of the load 14 on the actuating cylinder, and the displacement sensor 24 is used to monitor the movement displacement of the actuating cylinder. Table 1 shows the physical quantities and units monitored by some of the monitoring sensors in the monitoring sensor group:;

[0028] Table 1

[0029] sensor Measuring physical quantities unit Shock Sensor Jitter mms Displacement Sensors length mm Load Cells force N Flow Sensors flow L / min Pressure Sensors pressure bar Temperature Sensor temperature ℃

[0030] The control system is used to control the hardware in the hydraulic test system to simulate various servo actuation system failures. The hardware controlled by the control system includes 3 hydraulic pumps, 3 solenoid valves and 3 adjustable flow valves, etc. At the same time, the monitoring data of the monitoring sensor group in the hydraulic test system is collected. The control system is equipped with a deep learning model, which uses the monitoring data of simulated faults as input and the fault type as output for training. After training, it is used to comprehensively analyze and effectively detect various fault types. After training, the accuracy of the deep learning model is greater than 95%.

[0031] The deep learning model, such as Figure 2As shown, the collected data is first pre-processed and divided into time series data and spatial data, and the time series data (such as flow meter 15 data) is input into the LSTM network, and at the same time, the spatial data (such as optical fiber strain sensor 22 data) is input into the CNN network for processing; the LSTM network includes two LSTM layers, the first LSTM layer extracts preliminary features of the time series data, which involve time dependencies and dynamic changes such as flow changes and power fluctuations; then, the second LSTM layer further processes these features to enhance the understanding of time series patterns to capture more complex long-term and short-term time dependencies; the CNN network includes convolution layer 1, pooling layer, convolution layer 2 and flattening layer, convolution layer 1 (Conv2D Layer 1) first extracts local features from the spatial data, such as pressure distribution and stress changes, and then reduces the dimension of the feature map through the pooling layer (MaxPooling), retains the main features and reduces the computational complexity, then, convolution layer 2 (Conv2D Layer 2) Further extract high-level spatial features to enhance the recognition ability of complex spatial patterns. Finally, the flattening layer flattens the two-dimensional convolution output into one-dimensional data to prepare for the subsequent processing of the fully connected layer; the output features of the LSTM network and the CNN network are spliced ​​through the fusion step to form a comprehensive feature vector; this feature vector is input to the fully connected layer 1 (Dense Layer 1) for high-dimensional feature mapping and integration, and the Dropout layer is applied during the training process to prevent over-fitting; the fully connected layer 2 (Dense Layer 2) further processes the fused features and extracts the final high-level features; the output layer is a classification output layer (Dense, Softmax activation), which is used to map the feature vector to different classification labels to detect servo actuator system faults; if necessary, a regression output layer (Dense, linear activation) can also be added to predict the severity of the fault or other continuous values.

[0032] This embodiment also provides a method for monitoring faults in an aviation servo actuator test bench, which is implemented based on the above-mentioned fault monitoring device. Figure 3 As shown, the fault monitoring method comprises the following steps:

[0033] Step 1, determine the data set: connect the actuator to the fault monitoring device, use the fault monitoring device to simulate various faults, and collect monitoring data from various sensors such as vibration, displacement, load, power, pressure, flow and temperature as fault data. In addition, collect the monitoring data of the actuator under normal working conditions as non-fault data, organize these data into time series, and mark the normal state and fault data in different situations. Normalize the data to form a data set as the input item for training the deep learning model, and use non-fault or fault type as the training output item. Among them, the simulated servo actuator system faults include hydraulic pump leakage fault, solenoid valve delay fault and actuator internal leakage fault. The three fault data collection methods are as follows:

[0034] When simulating a hydraulic pump leakage fault, the solenoid valve A4 in the hydraulic test system is closed and the solenoid valve B5 is opened. By controlling the solenoid valve A4, the adjustable flow valve C9 is set to 6 different leakage flow values ​​in sequence: 0.1L / min, 0.14L / min, 0.18L / min, 0.2L / min, 0.3L / min and 0.4L / min. The leakage flow value is monitored by the flow meter 15 on the branch. During the simulation process, the data of each sensor is collected as the hydraulic pump leakage fault data to train and verify the deep learning model to ensure that the data covers various possible operating conditions and leakage conditions.

[0035] When simulating the solenoid valve delay failure, the flow path setting in the hydraulic test system is the same as that when simulating the hydraulic pump leakage failure. The delay of the control signal of the solenoid valve B5 is used for simulation. A total of 5 different delay options are set as 0.1 seconds, 0.5 seconds, 1 second, 2 seconds and 3 seconds, and the solenoid valve B5 is simulated in turn. Under each delay setting, the data of the flow meter 15 of the branch where the solenoid valve B5 is located after it is started and closed is recorded as the solenoid valve delay fault data for training and verifying the deep learning model.

[0036] When simulating the internal leakage of the actuator, the solenoid valve A4 is opened, the solenoid valve B5 is closed, the load 14 is carried on the actuator and moves up and down, and the adjustable flow valve B8 is set to 12 different leakage flow values: 0.1L / min, 0.12L / min, 0.14L / min, 0.16L / min, 0.18L / min, 0.2L / min, 0.22L / min, 0.24L / min, 0.26L / min, 0.28L / min, 0.3L / min and 0.4L / min. Under each flow setting, when the internal leakage occurs, the data of each sensor is collected as the internal leakage data of the actuator to train and verify the deep learning model.

[0037] Step 2, determine the training model: divide the data set into training set and validation set in a ratio of 3:1, set the training threshold, train the deep learning model, and return to training when the training result is less than the threshold. When it is greater than the threshold, the training is completed. After the training is completed, the deep learning model is input into the validation set data for verification, so that the accuracy of the deep learning model after training is greater than 95%.

[0038] Step 3, fault monitoring: connect the servo actuator system 25 to be monitored to the hydraulic test system and monitor the data of each monitoring sensor during the working process in real time, input the data into the deep learning model, and finally determine whether there is a fault and output the fault type to realize fault monitoring, and provide intelligent decision-making support for the maintenance and fault prevention of the servo actuator system 25.

[0039] Table 2 shows the results of fault monitoring of a servo actuator system using this embodiment. There are 20 faults actually verified and 21 normal working items. Among them, 18 faults were correctly monitored and 2 were wrong. There are 20 normal working items correctly monitored and 1 was wrong.

[0040] Table 2

[0041]

[0042] The above-described embodiments merely express the implementation methods of the present invention, but they should not be understood as limiting the scope of the present invention. It should be pointed out that for those skilled in the art, several modifications and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.

Claims

1. A fault monitoring device for an aviation servo actuation test bench, characterized in that: The fault monitoring device includes a hydraulic test system and a control system, wherein the hydraulic test system is used to simulate the servo actuator system fault and collect data through various sensors, and the control system is used to control the hydraulic test system to achieve the required monitoring conditions, collect monitoring data and output monitoring results.

2. The fault monitoring device for an aviation servo actuation test bench according to claim 1, characterized in that: The specific structure of the fault monitoring device is as follows: The hydraulic test system comprises an oil tank (1), a filter (2), a hydraulic pump (3), three solenoid valves, three adjustable flow valves, a servo valve (12), a load (14) and a monitoring sensor group; the filter (2) is connected to the oil outlet of the oil tank (1), the hydraulic pump (3) is connected to the oil outlet of the filter (2), the oil outlet of the hydraulic pump (3) is connected to the adjustable flow valve A (7) and the servo valve (12) in sequence, the servo valve (12) is connected to the two sides of the piston of the actuator cylinder in the servo actuator system (25) to be monitored through two pipelines, and the servo valve (12) is used to accurately control the movement position and speed of the actuator cylinder in the servo actuator system (25). The oil outlet is connected to the oil tank (1), and the hydraulic oil after use flows back to the oil tank (1). Pipelines are connected to both sides of the piston of the actuator cylinder, and a solenoid valve A (4) and an adjustable flow valve B (8) are arranged on the pipeline. When the load (14) is used, it acts on the top of the actuator cylinder in the servo actuator system (25) to simulate the leakage failure in the actuator cylinder. A branch is arranged on the pipeline between the hydraulic pump (3) and the adjustable flow valve A (7), and the solenoid valve B (5) and the adjustable flow valve C (9) are connected in sequence on the branch. The oil outlet of the adjustable flow valve C (9) is connected to the oil tank (1); the monitoring sensor group includes multiple flow meters (15), multiple pressure gauges (16), and multiple A pressure sensor (17), a plurality of temperature sensors (18), a liquid level sensor (19), a power sensor (20), a vibration sensor (21), an optical fiber strain sensor (22), a force sensor (23), and a displacement sensor (24), wherein: a liquid level sensor (19) and a temperature sensor (18) are provided at the oil tank (1) to monitor the liquid level and temperature of the hydraulic oil in the oil tank (1) respectively; a power sensor (20), a vibration sensor (21) and a temperature sensor (18) are provided at the hydraulic pump (3) to monitor the power consumption, vibration and temperature of the hydraulic pump (3) respectively; and an adjustable throttle valve A (7) is provided at both ends. A pressure gauge (16) is provided; pressure sensors (17) and temperature sensors (18) are provided on the two pipelines connecting the servo valve (12) and the actuator cylinder; a flow meter (15) is provided on the pipeline provided with the electromagnetic valve A (4) and the adjustable flow valve B (8); an optical fiber strain sensor (22), a force sensor (23) and a displacement sensor (24) are provided on the actuator cylinder in the servo actuator system (25), wherein the optical fiber strain sensor (22) is used to monitor stress changes caused by leakage in the actuator cylinder, the force sensor (23) is used to monitor the weight of the load (14) on the actuator cylinder, and the displacement sensor (24) is used to monitor the movement displacement of the actuator cylinder; The control system is used to control the hydraulic test system to simulate the servo actuator system failure and collect monitoring data at the same time. The control system is equipped with a deep learning model, which uses the monitoring data of the simulated failure as input and the failure type as output for training. After training, it is used to comprehensively analyze and effectively detect various failure types.

3. The fault monitoring device for an aviation servo actuation test bench according to claim 2, characterized in that: Another branch is also provided on the pipeline between the hydraulic pump (3) and the adjustable flow valve A (7), on which a pilot electromagnetic overflow valve (10) is provided, and its oil outlet end is connected to the oil tank (1). The pilot electromagnetic overflow valve (10) is normally open and is used to accurately adjust the overflow pressure in the hydraulic test system; two branches are provided on the pipeline between the adjustable flow valve A (7) and the servo valve (12), one of which is provided with a safety valve (13), and the oil outlet end of the safety valve (13) is connected to the oil tank (1). The safety valve (13) is normally open and is used to protect the hydraulic test system from damage caused by excessive pressure, and the other branch is connected in sequence to the electromagnetic valve C (6) and the accumulator (11) for storing and releasing hydraulic energy, and the electromagnetic valve C (6) is normally open.

4. The fault monitoring device for an aviation servo actuation test bench according to claim 2, characterized in that: The deep learning model first pre-processes the collected data and divides them into time series data and spatial data, inputs the time series data into the LSTM network, and at the same time, inputs the spatial data into the CNN network for processing; the LSTM network includes two LSTM layers, the first LSTM layer extracts preliminary features of the time series data, and the second LSTM layer further processes these features to enhance the understanding of the time series pattern; the CNN network includes a convolution layer 1, a pooling layer, a convolution layer 2 and a flattening layer, the convolution layer 1 first extracts local features from the spatial data, the convolution layer 2 further extracts high-level spatial features, and the flattening layer flattens the two-dimensional convolution output into one-dimensional data; the output features of the LSTM network and the CNN network are spliced ​​through a fusion step to form a comprehensive feature vector; This feature vector is input to the fully connected layer 1 for high-dimensional feature mapping and integration, and the Dropout layer is applied during the training process to prevent over-fitting. The fully connected layer 2 further processes the fused features and extracts the final high-level features. The output layer is a classification output layer, which is used to map feature vectors to different classification labels to detect servo actuator system faults.

5. The fault monitoring device for an aviation servo actuation test bench according to claim 4, characterized in that: The accuracy of the deep learning model is greater than 95%.

6. The fault monitoring device for an aviation servo actuation test bench according to claim 2, characterized in that: The servo actuation system faults that can be simulated and detected by the fault monitoring device include hydraulic pump leakage, solenoid valve delay and leakage in the actuation cylinder.

7. A method for monitoring faults of an aviation servo actuator test bench, implemented based on the fault monitoring device according to any one of claims 1 to 6, characterized in that: The fault monitoring method comprises the following steps: Step 1, determine the data set: connect the actuator to the fault monitoring device, use the fault monitoring device to simulate various faults, and collect monitoring data as fault data. In addition, collect monitoring data of the actuator in normal working state as non-fault data, normalize the data to form a data set as input for training the deep learning model, and use non-fault or fault type as training output; Step 2, determine the training model: divide the data set into a training set and a validation set in a ratio of 3:1, set a training threshold, train the deep learning model, and return to training if the training result is less than the threshold. If it is greater than the threshold, the training is completed. After the training is completed, the deep learning model is input into the validation set data for verification; Step 3, fault monitoring: connect the servo actuator system (25) to be monitored to the hydraulic test system and monitor the data of each monitoring sensor in the working process in real time, input the data into the deep learning model, and finally determine whether there is a fault and output the fault type.

8. The method for monitoring faults of an aviation servo actuation test bench according to claim 7, characterized in that: In step 1, the simulated servo actuation system faults include hydraulic pump leakage fault, solenoid valve delay fault and actuator cylinder leakage fault. The three fault data collection methods are specifically as follows: When simulating a hydraulic pump leakage fault, the solenoid valve A (4) in the hydraulic test system is closed and the solenoid valve B (5) is opened. By controlling the solenoid valve A (4), the adjustable flow valve C (9) is sequentially set to a plurality of different leakage flow values. During the simulation process, the data of each sensor is collected; When simulating a solenoid valve delay fault, the solenoid valve A (4) in the hydraulic test system is closed and the solenoid valve B (5) is opened. The simulation is performed using the delay of the control signal of the solenoid valve B (5). A plurality of different delay options are set. Under each delay setting, the data of the flow meter (15) of the branch where the solenoid valve B (5) is located is recorded after the solenoid valve B (5) is started and closed. When simulating an internal leakage fault in the actuator, the solenoid valve A (4) is opened, the solenoid valve B (5) is closed, a load (14) is placed on the actuator and the actuator moves up and down, and the adjustable flow valve B (8) is set to a plurality of different leakage flow values. Under each flow setting, when an internal leakage fault occurs, data from each sensor is collected.

Images