An AI algorithm-based online monitoring method for external leakage of an aircraft hydraulic system

Abstract

This invention belongs to the field of hydraulic system technology and discloses an AI-based online monitoring method for external leakage in aircraft hydraulic systems. It aims to solve the technical problems of poor real-time performance, high false alarm rate, and difficulty in adapting to dynamic operating conditions in existing monitoring methods. The aircraft hydraulic system is the power source for critical functions such as door opening and closing, landing gear retraction and extension, braking, and control surface actuation. External leakage can easily lead to serious safety accidents. Existing monitoring methods have obvious defects, insufficient generalization ability, and inadequate real-time performance. The steps of this method are as follows: First, select multiple leak-free flight parameter data, extract multi-dimensional parameters such as pressure and temperature, and construct a feature vector set after preprocessing; second, based on AI algorithms, construct and train a fuel tank oil volume prediction model under leak-free operating conditions; third, input real-time collected multi-dimensional time-series data into the model and output the predicted oil volume value; finally, compare the predicted value with the measured value from the level sensor, determine the leakage status, sensor failure, or model deviation through the volume difference, and realize online model updates. This invention integrates multi-dimensional parameter features, improves the accuracy of leak identification and anti-interference ability, can output monitoring results in real time and provide early warning of leakage, enhances the robustness of hydraulic system leak monitoring, reduces operation and maintenance costs, and can provide technical support for early warning of leaks in aircraft hydraulic systems, thus having significant engineering application value.

Description

Technical Field

[0001] This invention belongs to the field of hydraulic system technology, and in particular relates to an online monitoring method for external leakage in aircraft hydraulic systems based on AI algorithms. Background Technology

[0002] The aircraft hydraulic system, as the power source for critical functions such as door opening and closing, landing gear retraction and extension, braking, and control surface actuation, directly affects the safety and reliability of the aircraft. Due to its complex structure, extensive piping, and numerous connecting components, the hydraulic system is highly susceptible to external leakage under long-term service and complex flight conditions, influenced by factors such as vibration, temperature changes, seal aging, and assembly deviations. Hydraulic oil leakage not only leads to a drop in system pressure and sluggish actuation response but can also cause serious safety hazards such as fires, and even result in catastrophic accidents involving aircraft destruction and loss of life.

[0003] Currently, domestic and international methods for monitoring leaks in aircraft hydraulic systems mainly employ static threshold monitoring and empirical rule-based diagnosis. However, during different flight phases such as takeoff, cruise, and landing, the pressure, temperature, and load of an aircraft's hydraulic system exhibit strong nonlinear changes. Traditional methods are prone to false alarms or missed alarms under dynamic conditions, making it difficult to accurately identify minute early leaks. In recent years, although some research has attempted to introduce artificial intelligence algorithms such as neural networks and support vector machines, most treat them as black-box models, relying excessively on fault sample data, resulting in weak generalization ability and insufficient robustness under complex and variable conditions. Furthermore, existing technologies are mostly focused on post-fault diagnosis or offline data analysis, lacking an online dynamic monitoring mechanism that can deeply couple hydraulic physical characteristics with real-time operational data, making it difficult to balance real-time monitoring and accuracy.

[0004] Therefore, it is necessary to study a new online monitoring method for external leakage of aircraft hydraulic systems that can adapt to complex dynamic working conditions, has good robustness and real-time performance, and realize a highly reliable online monitoring mechanism during flight. This method can provide technical support for early warning of hydraulic system leakage risks, reduce flight safety hazards and maintenance costs caused by leakage, and has important engineering application value. Summary of the Invention

[0005] This invention addresses the technical problems of existing aircraft hydraulic system leakage monitoring methods, such as poor real-time performance, high false alarm rate, and difficulty in adapting to dynamic operating conditions. It proposes an online monitoring method for external leakage in aircraft hydraulic systems based on AI algorithms, aiming to improve the real-time early warning capability of leakage faults during aircraft flight, reduce operation and maintenance costs, and provide technical support for online monitoring of aircraft hydraulic systems.

[0006] The technical solution of the present invention includes the following steps:

[0007] The first step is sample selection and feature engineering.

[0008] Multiple sortie flight parameter data under leak-free conditions were collected as a baseline sample set, from which multidimensional parameters related to the hydraulic system were extracted, including but not limited to the pressures P1, P2, ..., P of each branch. n and temperatures T1, T2, ..., T m Key parameters were used as samples. The sample data covered different flight phases, operating loads, and environmental conditions to ensure the model fully learned the oil volume fluctuation patterns under leak-free hydraulic system conditions. Sensor jump values ​​were removed, and the samples were normalized to eliminate the influence of dimensions, constructing a feature vector set X=[P]. n , T m ].

[0009] The second step is to build and train a data-driven model based on AI algorithms.

[0010] Using AI algorithms as the core, a data-driven model for predicting the volume of oil in the fuel tank is constructed. When training the initial model, the dynamic changes in the volume of oil in the fuel tank with pressure and temperature during normal system circulation are used as a benchmark. The AI ​​algorithm learns the high-dimensional feature mapping relationship under leak-free operating conditions.

[0011] This step involves inputting the feature vector set into a pre-defined AI model architecture (including but not limited to ensemble learning models, neural network models, or deep learning models), and using optimization algorithms to adjust the model parameters, minimizing the loss function between the predicted volume and the actual volume. The trained model can capture the complex physical relationships and time-varying characteristics within the hydraulic system, outputting a theoretical prediction, V0, of the tank oil volume under leak-free conditions. This AI-based learning approach effectively extracts the nonlinear patterns between multi-dimensional features, significantly improving the model's prediction accuracy and generalization ability.

[0012] The third step is to predict and dynamically monitor the oil volume in the tank under real-time operating conditions.

[0013] The real-time collected aircraft hydraulic system pressures P1, P2, ..., P n Temperatures T1, T2, ..., T m The AI ​​model, trained by inputting multi-dimensional time-series data, performs feature mapping and calculation through the internal logic layer of the model, and outputs the theoretical predicted value V1 of the oil tank volume under the current working condition in real time. This predicted value represents the theoretical value of the oil volume that the hydraulic system should have under real-time working conditions.

[0014] The fourth step is to output the leakage judgment and prediction results.

[0015] The predicted oil volume V1 from the data-driven model is compared with the actual oil volume V2 (obtained by the onboard level sensor), and the volume difference ΔV = V1 − V2 is calculated. The leakage status is then determined based on the difference.

[0016] When |ΔV| = 0 or |ΔV| < ε, the hydraulic system is considered to be normal. ε is a preset hydraulic system error tolerance threshold to avoid false alarms.

[0017] When ΔV>0 and continues to increase or exceeds the threshold, it is determined that there is an external leakage in the hydraulic system. At this time, the leakage amount is output and an alarm is triggered, and the size of ΔV directly reflects the leakage amount.

[0018] When ΔV < 0, meaning the measured value is greater than the theoretical prediction value, it is determined to be a sensor malfunction, measurement anomaly, or model deviation. The data-driven model automatically records this type of characteristic data and feeds it back to the training module for model retraining, thus realizing online updates of the data-driven model.

[0019] The system ultimately outputs the status and judgment results of the hydraulic system, enabling online monitoring and early warning of external leaks in the aircraft's hydraulic system.

[0020] The present invention provides an online monitoring method for external leakage in aircraft hydraulic systems based on AI algorithms, which has the following advantages compared with existing technologies:

[0021] (1) Improved recognition accuracy and anti-interference ability. By using AI algorithm to integrate multi-dimensional parameter features, it effectively learns the nonlinear fluctuation law of hydraulic system and solves the technical problem of high false alarm rate and difficulty in identifying small leaks under dynamic working conditions by traditional methods.

[0022] (2) By quickly mapping the real-time collected time-series data online, the operating status of the hydraulic system can be output in real time; once the hydraulic system leaks externally, it can trigger an alarm and output monitoring results to provide key decision support for the pilot, effectively avoiding serious safety accidents such as pressure instability or control failure caused by leakage, and improving flight safety.

[0023] (3) By determining the positive or negative difference, the system can effectively distinguish between system leakage and sensor failure, and the feedback data can be used for model training and updating, which enhances the robustness of the system and improves flight safety; at the same time, it realizes the active monitoring of hydraulic system leakage, effectively reducing operation and maintenance costs. Attached Figure Description

[0024] Figure 1 This is a flowchart illustrating an online monitoring method for external leakage in an aircraft hydraulic system based on an AI algorithm, as provided by the present invention.

[0025] Figure 2 This is a simplified schematic diagram of an aircraft hydraulic system.

[0026] The markings in the diagram are explained below:

[0027] 1 is a hydraulic pump; 2 is a buffer bottle; 3 is filter 1; 4 is accumulator 1; 5 is accumulator 2; 6 is actuator 1; 7 is directional valve 1; 8 is actuator 2; 9 is directional valve 2; 10 is filter 2; 11 is heat exchanger; 12 is hydraulic oil tank; 13 is throttle valve; 14 is check valve. Detailed Implementation

[0028] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited to this embodiment. In this embodiment, according to Figure 2 The schematic diagram of the aircraft hydraulic system shown is used to establish a system-level hydraulic simulation model; the AI ​​algorithm is exemplified by the random forest algorithm.

[0029] The rated working pressure of the hydraulic system is set to 28 MPa, and the rated speed of hydraulic pump 1 is 3000 r / min; the effective volumes of accumulator 1 and accumulator 2 are set to 1.5L and 1.0L respectively, and the pre-charge pressure is 0.6 times the rated pressure of the system; the initial oil volume of hydraulic oil tank 12 is set to 10L, the density of the hydraulic oil used in the simulation is 850 kg / m³, and the bulk modulus of elasticity is taken as 1.6 × 10⁻⁶. 9 Pa. The initial oil temperature of the system was set to 20℃, and the oil temperature was allowed to vary from 30℃ to 80℃ depending on the operating conditions during the simulation.

[0030] Virtual sensors were deployed at key system nodes to collect data on the hydraulic pump outlet pressure P1, the pressures before and after the accumulator P2-P4, the inlet and outlet pressures of the actuator P5-P8, and the corresponding oil temperatures T1-T3. Simultaneously, the actual oil volume V2 in the tank was obtained in real time from a tank level sensor. The simulation time was set to 2400 seconds, with the actuator's periodic movement frequency ranging from 0.1 to 0.3 Hz, to simulate load variations during different flight phases, including takeoff, cruise, and landing.

[0031] Under leak-free conditions, simulation results show that the oil volume in the tank fluctuates slightly around the initial value, with the maximum fluctuation not exceeding ±0.3L. This fluctuation is mainly caused by temperature changes and the compressibility of the hydraulic system oil. The pressure and temperature data collected under this condition, along with the corresponding oil volume data in the tank, form a leak-free sample set for training the random forest model.

[0032] In the leakage simulation, an external leakage module was introduced into the hydraulic pipeline nodes, and equivalent leakage orifice diameters of 0.5mm, 1mm, and 2mm were set to simulate different degrees of external leakage faults. The actual values ​​of the pressure P1 to P8, hydraulic oil temperature T1 to T3, and oil volume V2 in the hydraulic tank were collected for each branch. The simulation results show that when the leakage orifice diameter is 0.5mm, the oil volume in the tank decreases continuously by about 0.9L within 1200s; when the leakage orifice diameter increases to 0.8mm, the rate of oil volume decrease accelerates significantly.

[0033] The pressure and temperature data collected in real-time simulation are input into the trained random forest model to obtain the predicted oil volume V1 in the tank, which is then compared with the actual value V2 output by the level sensor. Under leak-free conditions, the absolute value of the prediction error is always less than 0.12L; under external leakage conditions, when ΔV exceeds the set threshold of 0.3L, the system can stably determine that a leak has occurred and output the corresponding prediction result.

[0034] In summary, this embodiment verifies the feasibility and effectiveness of the method of the present invention under complex dynamic working conditions through simulation. It can realize real-time online monitoring and early warning of external leakage in aircraft hydraulic systems, providing technical support for flight safety.

Claims

1. A method for online monitoring of external leakage in aircraft hydraulic systems based on AI algorithms, characterized in that, Based on flight parameters, a baseline sample is constructed, and an AI algorithm is used to establish a fuel tank oil volume prediction model. By comparing the difference between the predicted and measured values, the leakage status of the hydraulic system is determined, including the following steps: (1) Sample selection and feature engineering: Flight parameter data under leak-free conditions were collected as a baseline sample set, and multi-dimensional parameters related to the hydraulic system were extracted as features, including pressure P1, P2, ..., P n and temperatures T1, T2, ..., T m The samples are subjected to outlier removal and normalization to construct a feature vector set X=[P]. n , T m ]; (2) Model construction and training: Using AI algorithm as the core, a prediction model of oil tank volume under leak-free conditions is constructed; the model is trained using the benchmark sample set to construct the dynamic feature mapping relationship between pressure, temperature and oil tank volume, and output the predicted value V0 of oil tank volume under leak-free conditions. (3) Online prediction and dynamic monitoring: Real-time collection of pressure and temperature data of the aircraft hydraulic system, input into the trained AI model, and output of the theoretical prediction value V1 of the oil volume in the tank under the current working conditions; (4) Leakage judgment and result output: Compare the theoretical predicted value V1 with the actual value V2 of the oil volume in the tank collected by the liquid level sensor, calculate the oil volume difference ΔV=V1-V2, determine the hydraulic system status based on ΔV and output the monitoring results.

2. The method for online monitoring of external leakage in aircraft hydraulic systems based on AI algorithms according to claim 1, characterized in that: In step (1), the multidimensional parameters include hydraulic pump outlet pressure, accumulator charging pressure, hydraulic system pressure, oil tank pressure, oil tank temperature, and return oil temperature.

3. The method for online monitoring of external leakage in aircraft hydraulic systems based on AI algorithms according to claim 1, characterized in that: The AI ​​algorithms mentioned in step (2) include, but are not limited to, random forest algorithms, neural network algorithms, or deep learning algorithms.

4. The method for online monitoring of external leakage in aircraft hydraulic systems based on AI algorithms according to claim 1, characterized in that: In step (4), the criteria for determining leakage are as follows: when |ΔV|=0 or |ΔV|<ε, the hydraulic system is determined to be normal, where ε is a preset error tolerance threshold; when ΔV continues to increase or exceeds the preset threshold, the hydraulic system is determined to have external leakage, and the leakage amount and alarm information are output; when ΔV<0, it is determined to be a sensor failure, measurement abnormality or model deviation.

5. The method for online monitoring of external leakage in aircraft hydraulic systems based on AI algorithms according to claim 4, characterized in that: When the judgment result is ΔV<0, the data-driven model automatically records the feature data under this working condition and feeds it back to the training module to train and update the AI ​​model online.

6. The online monitoring method for external leakage of an aircraft hydraulic system based on an AI algorithm according to any one of claims 1 to 5, characterized in that: This method enables early warning of hydraulic system leakage risks during aircraft flight by comparing the deviation between theoretical predictions and actual measurements in real time.

Images