EVTOL health management and control method and system

Abstract

The invention discloses an eVTOL health management and control method and system, and relates to the technical field of data analysis, and the method comprises the steps: dividing equipment operation parameters into a large-scale data set and a small-scale data set; performing feature extraction on the small-scale data set to obtain statistical feature information and engineering feature information, and performing fusion processing to obtain fused feature information; performing rule reasoning based on the fused feature information to obtain an initial fault isolation decision, and adjusting the initial fault isolation decision based on an artificial intelligence lightweight small model; performing equipment fault diagnosis by using the large-scale data set based on an artificial intelligence large model; performing equipment degradation feature extraction on the large-scale data set, and determining a maintenance plan based on equipment degradation feature information and an equipment fault diagnosis result; and performing eVTOL health state management and control based on the maintenance plan and the target fault isolation decision. According to the method, a more accurate maintenance plan can be obtained, and the safety and economy of eVTOL operation are improved.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, and in particular to a health management method and system for eVTOL. Background Technology

[0002] The increasing global urbanization rate has brought about a crisis in transportation efficiency. As an emerging mode of transportation, eVTOL (eVTOL) has seen rapid development due to its pollution-free, low-noise, and safe characteristics. Traditional helicopter crashes are mostly caused by power failure, while eVTOLs use distributed propulsion, employing multiple small propellers to provide lift and thrust. This redundant distributed propulsion improves the fault tolerance of the power system, significantly enhancing the safety and reliability of eVTOLs while also significantly reducing overall noise. The control of distributed propulsion relies on fly-by-wire flight control for autopilot. Through the integration of distributed electric propulsion and autopilot technologies, eVTOLs have become a crucial driving force for building future integrated transportation networks. To ensure the safety and reliability of eVTOL operation, health management is necessary. Currently, health status is typically analyzed manually using statistical data to formulate maintenance strategies. However, this method relies heavily on the professional expertise of personnel, making it difficult to guarantee the reliability of fault repair decisions. Furthermore, with the ever-increasing scale of data, manual data analysis is prone to errors and omissions, and its efficiency is low, failing to meet the high requirements of current eVTOL health management. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides a health management method and system for eVTOL, which can obtain more accurate maintenance plans, reduce false alarm rates, and improve the safety and economy of eVTOL operation.

[0004] To address the aforementioned technical problems, this invention provides a health management method for eVTOL, the method comprising: Obtain the equipment operating parameters of the electric vertical takeoff and landing (eVTOL) aircraft, and divide the equipment operating parameters into large-scale datasets and small-scale datasets; Feature extraction is performed on the small-scale dataset to obtain statistical feature information and engineering feature information, and the statistical feature information and engineering feature information are fused to obtain fused feature information; Based on the fused feature information, rule reasoning is performed to obtain an initial fault isolation decision, and the initial fault isolation decision is adjusted based on a lightweight artificial intelligence model to obtain a target fault isolation decision. Based on the large-scale artificial intelligence model, the equipment fault diagnosis is performed using the large-scale dataset to obtain equipment fault diagnosis results; Equipment degradation features are extracted from the large-scale dataset to obtain equipment degradation feature information, and a maintenance plan is determined based on the equipment degradation feature information and equipment fault diagnosis results. Health status management of eVTOL is carried out based on the maintenance plan and target fault isolation decisions.

[0005] Optionally, the equipment operating parameters for acquiring the electric vertical takeoff and landing (eVTOL) aircraft include: Flight data and electric drive data of eVTOL are collected based on a combination of sensors; The battery status parameters of the eVTOL are collected by the battery management system, and the remaining battery power is analyzed based on the battery status parameters to obtain the remaining battery power information. Based on the battery state parameters, perform battery health status analysis to obtain battery health status information; Based on the flight data, electric drive data, battery status parameters, remaining battery power information, and battery health status information, the device operating parameters for eVTOL are generated.

[0006] Optionally, the step of extracting features from the small-scale dataset to obtain statistical feature information and engineering feature information, and then fusing the statistical feature information and engineering feature information to obtain fused feature information, includes: The small-scale dataset is normalized to obtain a normalized small-scale dataset, and the normalized small-scale dataset is standardized to obtain a standardized small-scale dataset. Statistical features are extracted from a small-scale dataset after standardization to obtain statistical feature information; Engineering feature extraction is performed on a small-scale dataset after standardization to obtain engineering feature information; Determine the target weight, and then fuse the statistical feature information and engineering feature information based on the target weight to obtain fused feature information.

[0007] Optionally, the step of extracting engineering features from the standardized small-scale dataset to obtain engineering feature information includes: Signal rate of change analysis is performed on a small-scale dataset after standardization to obtain signal rate of change information. Signal kurtosis analysis was performed on a small-scale dataset after standardization to obtain signal kurtosis information; Signal volatility analysis is performed on a small dataset after standardization to obtain signal volatility information, and engineering feature information is generated based on the signal rate of change information, signal kurtosis information, and signal volatility information.

[0008] Optionally, the step of performing rule-based reasoning based on the fused feature information to obtain an initial fault isolation decision includes: Acquire formal expert knowledge and construct a risk prediction reasoning framework based on the formal expert knowledge; Based on the risk prediction and reasoning framework, the fused feature information is used to perform rule-based reasoning to obtain an initial fault isolation decision.

[0009] Optionally, the process of using the large-scale dataset based on the artificial intelligence model to perform equipment fault diagnosis and obtain equipment fault diagnosis results includes: The large-scale dataset is subjected to state labeling processing to obtain a state-labeled large-scale dataset; Feature extraction is performed on the large-scale dataset after state labeling to obtain target feature information; A large-scale artificial intelligence model is trained based on a sample dataset to obtain a well-trained large-scale artificial intelligence model. The target feature information is input into a trained artificial intelligence model to diagnose equipment faults and obtain equipment fault diagnosis results.

[0010] Optionally, the training process of the large artificial intelligence model includes: The sample dataset is input into a large artificial intelligence model for forward propagation calculation to obtain the predicted value; The Huber loss function is calculated between the predicted and actual values. Backpropagation of the large AI model is performed based on the Huber loss function, and the parameters of the large AI model are updated using the Gradient Descent optimizer.

[0011] Optionally, the step of extracting device degradation features from the large-scale dataset to obtain device degradation feature information includes: Vibration kurtosis analysis was performed on the aforementioned large-scale dataset to obtain vibration kurtosis information. Temperature change rate information is obtained by analyzing the temperature change rate based on the large-scale dataset. Vibration root mean square analysis is performed based on the large-scale dataset to obtain vibration root mean square information, and equipment degradation characteristics are determined based on the vibration kurtosis information, temperature change rate information, and vibration root mean square information.

[0012] Optionally, determining the maintenance plan based on the equipment degradation characteristic information and equipment fault diagnosis results includes: Based on the aforementioned equipment degradation characteristic information, a slow-change anomaly analysis is performed to obtain slow-change anomaly information; The equipment degradation characteristic information is compared with the preset warning value to obtain the comparison result, and a maintenance plan is determined based on the slow-change anomaly information, the comparison result and the equipment fault diagnosis result.

[0013] In addition, the present invention also provides an eVTOL health management system, the system comprising: Dataset partitioning module: used to acquire the equipment operating parameters of the electric vertical takeoff and landing (eVTOL) aircraft and partition the equipment operating parameters into large-scale datasets and small-scale datasets; Feature fusion module: used to extract features from the small-scale dataset to obtain statistical feature information and engineering feature information, and to fuse the statistical feature information and engineering feature information to obtain fused feature information; Fault Decision Analysis Module: Used to perform rule-based reasoning based on the fused feature information to obtain an initial fault isolation decision, and to adjust the initial fault isolation decision based on a lightweight artificial intelligence model to obtain a target fault isolation decision; Fault diagnosis module: used to perform equipment fault diagnosis based on the large-scale dataset using an artificial intelligence model, and obtain equipment fault diagnosis results; Maintenance plan determination module: used to extract equipment degradation features from the large-scale dataset, obtain equipment degradation feature information, and determine a maintenance plan based on the equipment degradation feature information and equipment fault diagnosis results; Health management module: used to manage the health status of eVTOL based on the maintenance plan and target fault isolation decisions.

[0014] In this embodiment of the invention, the operating parameters of the eVTOL equipment are acquired and divided into large-scale and small-scale datasets. Feature extraction is performed on the small-scale datasets to obtain statistical and engineering feature information. This statistical and engineering feature information is then fused to obtain fused feature information, which captures more comprehensive feature information. Rule-based reasoning is performed based on the fused feature information to obtain initial fault isolation decisions. These initial fault isolation decisions are adjusted using a lightweight artificial intelligence model, improving the response speed of fault isolation actions and accurately identifying risk factors. A large-scale artificial intelligence model is used for equipment fault diagnosis using the large-scale dataset. Equipment degradation features are extracted from the large-scale dataset to obtain equipment degradation feature information. Based on this degradation feature information and the equipment fault diagnosis results, a maintenance plan is determined, resulting in a more accurate maintenance plan and reducing false alarm rates. Based on the maintenance plan and the target fault isolation decision, the health status of the eVTOL is managed, improving the safety and economy of eVTOL operation. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating the health management method for eVTOL in an embodiment of the present invention; Figure 2 This is a flowchart illustrating the health management method for eVTOL in another embodiment of the present invention; Figure 3 This is a schematic diagram of the structural composition of the eVTOL health management system in an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1 Please see Figure 1 , Figure 1 This is a flowchart illustrating the health management method for eVTOL according to an embodiment of the present invention. The method includes: S11: Obtain the equipment operating parameters of the electric vertical takeoff and landing (eVTOL) aircraft, and divide the equipment operating parameters into a large-scale dataset and a small-scale dataset; In the specific implementation of this invention, flight data and electric drive data of the electric vertical take-off and landing (eVTOL) are collected based on a sensor combination; battery status parameters of the eVTOL are collected based on the battery management system, and the remaining battery power is analyzed based on the battery status parameters to obtain the remaining battery power information; battery health status is analyzed based on the battery status parameters to obtain battery health status information; and equipment operating parameters of the eVTOL are generated based on the flight data, electric drive data, battery status parameters, remaining battery power information, and battery health status information. The equipment operating parameters are divided into large-scale datasets and small-scale datasets to obtain more comprehensive data.

[0019] S12: Extract features from the small dataset to obtain statistical feature information and engineering feature information, and fuse the statistical feature information and engineering feature information to obtain fused feature information; In the specific implementation of this invention, the small-scale dataset is normalized to obtain a normalized small-scale dataset; the normalized small-scale dataset is then standardized to obtain a standardized small-scale dataset; statistical features are extracted from the standardized small-scale dataset to obtain statistical feature information; engineering features are extracted from the standardized small-scale dataset to obtain engineering feature information; a target weight is determined, and the statistical feature information and engineering feature information are fused based on the target weight to obtain fused feature information. This process captures more comprehensive feature information and provides more reliable data support for subsequent fault isolation decision analysis.

[0020] S13: Based on the fused feature information, perform rule reasoning to obtain an initial fault isolation decision, and adjust the initial fault isolation decision based on a lightweight artificial intelligence model to obtain a target fault isolation decision; In the specific implementation of this invention, formalized expert knowledge is acquired, and a risk prediction reasoning framework is constructed based on the formalized expert knowledge. Based on the risk prediction reasoning framework, rule reasoning is performed using the fused feature information to obtain an initial fault isolation decision. The initial fault isolation decision is adjusted based on a lightweight artificial intelligence model, which is beneficial to improving decision accuracy and reducing response time.

[0021] S14: Based on the large-scale artificial intelligence model, use the large-scale dataset to perform equipment fault diagnosis and obtain equipment fault diagnosis results; In the specific implementation of this invention, the large-scale dataset is subjected to state labeling processing to obtain a large-scale dataset after state labeling processing; feature extraction is performed on the large-scale dataset after state labeling processing to obtain target feature information; a large artificial intelligence model is trained based on the sample dataset to obtain a trained large artificial intelligence model; the target feature information is input into the trained large artificial intelligence model to perform equipment fault diagnosis to obtain equipment fault diagnosis results, which can quickly locate fault information.

[0022] S15: Extract equipment degradation features from the large-scale dataset to obtain equipment degradation feature information, and determine a maintenance plan based on the equipment degradation feature information and equipment fault diagnosis results; In the specific implementation of this invention, vibration kurtosis analysis is performed based on the large-scale dataset to obtain vibration kurtosis information; temperature change rate is performed based on the large-scale dataset to obtain temperature change rate information; root mean square (RMS) analysis is performed based on the large-scale dataset to obtain vibration RMS information; and equipment degradation characteristic information is determined based on the vibration kurtosis information, temperature change rate information, and vibration RMS information; gradual change anomaly analysis is performed based on the equipment degradation characteristic information to obtain gradual change anomaly information; the equipment degradation characteristic information is compared with preset warning values ​​to obtain comparison results; and a maintenance plan is determined based on the gradual change anomaly information, comparison results, and equipment fault diagnosis results, making the generated maintenance plan more reliable and comprehensive.

[0023] S16: Perform health status management of eVTOL based on the maintenance plan and target fault isolation decision.

[0024] In the specific implementation of this invention, health status management of eVTOL based on the maintenance plan and target fault isolation decision can improve the safety and economy of route operation.

[0025] In this embodiment of the invention, the operating parameters of the eVTOL equipment are acquired and divided into large-scale and small-scale datasets. Feature extraction is performed on the small-scale datasets to obtain statistical and engineering feature information. This statistical and engineering feature information is then fused to obtain fused feature information, which captures more comprehensive feature information. Rule-based reasoning is performed based on the fused feature information to obtain initial fault isolation decisions. These initial fault isolation decisions are adjusted using a lightweight artificial intelligence model, improving the response speed of fault isolation actions and accurately identifying risk factors. A large-scale artificial intelligence model is used for equipment fault diagnosis using the large-scale dataset. Equipment degradation features are extracted from the large-scale dataset to obtain equipment degradation feature information. Based on this degradation feature information and the equipment fault diagnosis results, a maintenance plan is determined, resulting in a more accurate maintenance plan and reducing false alarm rates. Based on the maintenance plan and the target fault isolation decision, the health status of the eVTOL is managed, improving the safety and economy of eVTOL operation.

[0026] Example 2 Please see Figure 2 , Figure 2 This is a flowchart illustrating a health management method for eVTOL according to another embodiment of the present invention, the method comprising: S201: Obtain the equipment operating parameters of the electric vertical takeoff and landing (eVTOL) aircraft, and divide the equipment operating parameters into a large-scale dataset and a small-scale dataset; In the specific implementation of this invention, the acquisition of the equipment operating parameters of the electric vertical takeoff and landing (eVTOL) aircraft includes: collecting flight data and electric drive data of the eVTOL based on a combination of sensors; collecting battery status parameters of the eVTOL based on a battery management system; and analyzing the remaining battery power based on the battery status parameters to obtain the remaining battery power information. Battery health status analysis is performed based on the battery status parameters to obtain battery health status information; eVTOL equipment operating parameters are generated based on the flight data, electric drive data, battery status parameters, remaining battery power information, and battery health status information.

[0027] Specifically, the eVTOL collects flight data and electric drive data based on a sensor suite, which includes attitude sensors, current sensors, voltage sensors, and temperature sensors. Flight data includes flight trajectory, airspeed, attitude (pitch angle, roll angle, yaw angle), overload, angle of attack, vibration signal, sideslip angle, etc. Electric drive data includes voltage, current, temperature, motor speed, etc.

[0028] The eVTOL's battery state parameters are collected by the battery management system. The eVTOL's power battery has its own battery management system, and the battery state parameters include battery voltage, current, and temperature. Based on these battery state parameters, the remaining battery capacity is analyzed to obtain the remaining battery capacity information, which can be calculated using the ampere-hour integration method.

[0029] Battery health status analysis is performed based on the battery state parameters to obtain battery health status information. Battery health status analysis can be performed using the capacity comparison method and the internal resistance estimation method. The capacity comparison method determines the battery health status by calculating the percentage of the current capacity to the rated capacity, while the internal resistance estimation method determines the battery health status by the change in internal resistance.

[0030] The eVTOL equipment operating parameters are generated based on the flight data, electric drive data, battery status parameters, remaining battery power information, and battery health status information. In other words, the eVTOL equipment operating parameters are composed of these data. Considering the limitations of onboard computing and storage resources, the lightweight onboard AI model only performs intelligent analysis on key information affecting flight safety. Other operating parameters during flight are sent to the ground data storage module via the data link. Therefore, the equipment operating parameters are divided into large-scale and small-scale datasets. The datasets are divided according to data type or size. Small-scale datasets contain data affecting flight safety, such as temperature and power. Large-scale datasets contain other operating parameters during eVTOL flight besides the small-scale datasets, such as electric drive temperature and electric drive speed at various time series.

[0031] S202: Normalize the small-scale dataset to obtain a normalized small-scale dataset, and standardize the normalized small-scale dataset to obtain a standardized small-scale dataset. In the specific implementation of this invention, the small-scale dataset is normalized by scaling it proportionally to the [0,1] interval to obtain a normalized small-scale dataset. Then, the normalized small-scale dataset is standardized by converting it into a distribution with a mean of 0 and a variance of 1 to obtain a standardized small-scale dataset. S203: Extract statistical features from a standardized small-scale dataset to obtain statistical feature information; In the specific implementation of this invention, statistical features are extracted from the standardized small-scale dataset to obtain statistical feature information. The statistical features are extracted through corresponding calculation formulas. The statistical feature information includes mean, variance, maximum value, minimum value, etc.

[0032] S204: Extract engineering features from a standardized small-scale dataset to obtain engineering feature information; In the specific implementation of this invention, the step of extracting engineering features from the standardized small-scale dataset to obtain engineering feature information includes: performing signal rate of change analysis based on the standardized small-scale dataset to obtain signal rate of change information; performing signal kurtosis analysis based on the standardized small-scale dataset to obtain signal kurtosis information; performing signal volatility analysis based on the standardized small-scale dataset to obtain signal volatility information, and generating engineering feature information based on the signal rate of change information, signal kurtosis information, and signal volatility information.

[0033] Specifically, signal rate of change analysis is performed on a small-scale dataset after standardization to obtain signal rate of change information. This information is the rate of change of the temperature signal. The expression for the signal rate of change is obtained by dividing the temperature difference between adjacent time points by the time interval: v = ΔT / Δt, Where v is the rate of change of the signal, ΔT is the temperature difference, and Δt is the time interval.

[0034] Signal kurtosis analysis is performed on a small-scale dataset after standardization to obtain signal kurtosis information, which is the kurtosis of the vibration signal and reflects the impact failure of the eVTOL. The mean and standard deviation of the vibration signal are calculated based on the small-scale dataset after standardization, and the fourth central moment is calculated based on the mean and standard deviation of the vibration signal. This fourth central moment is the signal kurtosis.

[0035] Signal volatility analysis is performed on a standardized small-scale dataset to obtain signal volatility information, which can be the volatility of the rotational speed signal. The highest and lowest rotational speeds are extracted from the standardized small-scale dataset, and the average rotational speed is calculated. The signal volatility is obtained by dividing the difference between the highest and lowest rotational speeds by the average rotational speed. Engineering feature information is generated based on this signal rate of change, signal kurtosis, and signal volatility information; that is, engineering feature information is composed of these three factors, making the obtained engineering feature information more comprehensive.

[0036] S205: Determine the target weight, and based on the target weight, fuse the statistical feature information and the engineering feature information to obtain fused feature information; In the specific implementation of this invention, the target weight is determined. The target weight can be assigned according to the importance of the sensor. Based on the target weight, the statistical feature information and the engineering feature information are fused to obtain the fused feature information. That is, the statistical feature information and the engineering feature information are weighted and summed according to the target weight to obtain the fused feature information.

[0037] S206: Based on the fused feature information, perform rule reasoning to obtain an initial fault isolation decision, and adjust the initial fault isolation decision based on a lightweight artificial intelligence model to obtain a target fault isolation decision; In a specific implementation of the present invention, the step of performing rule-based reasoning based on the fused feature information to obtain an initial fault isolation decision includes: acquiring formal expert knowledge and constructing a risk prediction reasoning framework based on the formal expert knowledge; and performing rule-based reasoning based on the risk prediction reasoning framework using the fused feature information to obtain an initial fault isolation decision.

[0038] Specifically, formal expert knowledge is acquired, which transforms the experience, rules, and principles of domain experts into a formal language that computers can understand and reason about. A risk prediction reasoning framework is then constructed based on this formal expert knowledge. This framework aims to utilize a formal knowledge base combined with formal expert knowledge. The risk prediction reasoning framework can be divided into a data layer, a knowledge layer, a reasoning engine layer, and an application layer. The data layer contains real-time sensor data (voltage, current, temperature, vibration, attitude), maintenance records, historical fault data, etc. The knowledge layer stores the formal expert knowledge base constructed in the previous part (such as BN networks, logic rule bases, and state machine models). The reasoning engine layer carries different reasoning algorithms, using the knowledge layer to interpret and infer information from the data layer. The application layer provides visualization of risk prediction results, alarms, diagnostic reports, and decision suggestions.

[0039] Based on the risk prediction and reasoning framework, rule-based reasoning is performed using the fused feature information. That is, the fused feature information is input into the risk prediction and reasoning framework to perform rule-based reasoning and obtain an initial fault isolation decision. For example, when the rate of change of the temperature signal is higher than the set value, based on expert knowledge, it can be determined that the battery module has failed and there is a high probability of thermal runaway. The battery module should be shut down immediately and the aircraft should execute an emergency landing procedure.

[0040] The initial fault isolation decision is adjusted based on a lightweight AI model to obtain the target fault isolation decision. This lightweight AI model, mounted on an airborne platform, employs a lightweight recurrent neural network. By simplifying the gating structure, it maintains the sequential modeling capabilities of the recurrent neural network while achieving faster parallel computing and a smaller model size. For example, when the motor power gradually increases during vertical takeoff and landing, the rate of change of the battery module temperature signal may peak within a short period. This model can capture this correlation and correct the judgment on whether the battery is experiencing thermal runaway. Through this lightweight AI model, indication-alarm-decision management is integrated, reducing the response time for fault isolation execution and enabling rapid, autonomous decision-making for fault isolation. It can promptly disconnect the fault source when a fault occurs, preventing the fault from spreading to the entire system.

[0041] S207: Based on the large-scale artificial intelligence model, use the aforementioned large-scale dataset to perform equipment fault diagnosis and obtain equipment fault diagnosis results; In the specific implementation of this invention, the step of using the large-scale dataset to perform equipment fault diagnosis based on the large-scale artificial intelligence model and obtain equipment fault diagnosis results includes: performing state labeling processing on the large-scale dataset to obtain a state-labeled large-scale dataset; performing feature extraction on the state-labeled large-scale dataset to obtain target feature information; training the large-scale artificial intelligence model based on the sample dataset to obtain a trained large-scale artificial intelligence model; and inputting the target feature information into the trained large-scale artificial intelligence model to perform equipment fault diagnosis and obtain equipment fault diagnosis results.

[0042] Specifically, the large-scale dataset undergoes state labeling processing, that is, the data in the large-scale dataset is labeled as normal / abnormal states to obtain a state-labeled large-scale dataset. Feature extraction is then performed on the state-labeled large-scale dataset to obtain target feature information, which may include statistical features and engineering features corresponding to the state-labeled large-scale dataset.

[0043] The AI ​​model is trained using a sample dataset. The dataset is divided into training, testing, and validation sets for training. The AI ​​model is deployed on the ground and can employ the GPT model. The target feature information is then input into the trained AI model for equipment fault diagnosis, yielding diagnostic results such as whether the electric drive radiator is blocked or whether the electric drive coolant is leaking.

[0044] Furthermore, the training process of the large AI model includes: inputting the sample dataset into the large AI model for forward propagation calculation to obtain predicted values; calculating the Huber loss function between the predicted values ​​and the true values, and performing backpropagation of the large AI model based on the Huber loss function; and using the Gradient Descent optimizer to update the parameters of the large AI model.

[0045] Specifically, the sample dataset is input into the large-scale AI model for forward propagation calculation to obtain predicted values. The Huber loss function is calculated between the predicted and true values. The Huber loss function is a computational function used to measure the difference between the model's predicted and true values. It is a loss function used for regression problems, combining the advantages of mean squared error and mean absolute error, aiming to be more robust to outliers. Backpropagation of the large-scale AI model is performed based on the Huber loss function, and the parameters of the large-scale AI model are updated using the Gradient Descent optimizer. The Gradient Descent optimizer is an optimizer that uses the gradient descent algorithm, an iterative optimization algorithm for finding local minima of differentiable functions. Its core idea is to use the gradient information of the objective function to determine the search direction. Simultaneously, after each round of training, the performance is evaluated using a validation set, and the optimal model is saved. Real flight data not used in training is used for testing to evaluate the model's generalization ability.

[0046] S208: Extract equipment degradation features from the large-scale dataset to obtain equipment degradation feature information, and determine a maintenance plan based on the equipment degradation feature information and equipment fault diagnosis results; In a specific implementation of the present invention, the step of extracting equipment degradation features from the large-scale dataset to obtain equipment degradation feature information includes: performing vibration kurtosis analysis based on the large-scale dataset to obtain vibration kurtosis information; performing temperature change rate analysis based on the large-scale dataset to obtain temperature change rate information; performing vibration root mean square analysis based on the large-scale dataset to obtain vibration root mean square information; and determining equipment degradation feature information based on the vibration kurtosis information, temperature change rate information, and vibration root mean square information.

[0047] Specifically, vibration kurtosis analysis is performed based on the large-scale dataset to obtain vibration kurtosis information, and temperature change rate is performed based on the large-scale dataset to obtain temperature change rate information. The steps for vibration kurtosis analysis and temperature change rate analysis can be referenced from those for signal change rate analysis, and will not be repeated here.

[0048] Vibration root mean square analysis is performed based on the large-scale dataset to obtain vibration root mean square information. That is, the root mean square of the vibration signal is calculated based on the large-scale dataset, and the equipment degradation characteristic information is determined based on the vibration kurtosis information, temperature change rate information and vibration root mean square information. The equipment degradation characteristic information is composed of vibration kurtosis information, temperature change rate information and vibration root mean square information.

[0049] Furthermore, the step of determining the maintenance plan based on the equipment degradation feature information and the equipment fault diagnosis result includes: performing a gradual change anomaly analysis based on the equipment degradation feature information to obtain gradual change anomaly information; comparing the equipment degradation feature information with a preset warning value to obtain a comparison result; and determining the maintenance plan based on the gradual change anomaly information, the comparison result, and the equipment fault diagnosis result.

[0050] Specifically, based on the equipment degradation characteristic information, a gradual change anomaly analysis is performed to obtain gradual change anomaly information. Gradual change anomaly phenomena include the wear degree and fatigue degree of components. For example, the wear / fatigue degradation of bearings is a gradual process. Therefore, it is necessary to perform gradual change anomaly analysis based on the equipment degradation characteristic information to determine the abnormal faults of these equipment components. Gradual change anomaly analysis can be performed using equipment degradation characteristic information through machine learning algorithms.

[0051] The equipment degradation characteristic information is compared with a preset warning value to obtain a comparison result, i.e., whether the equipment degradation characteristic information exceeds the preset warning value. Based on the gradual change anomaly information, the comparison result, and the equipment fault diagnosis result, a maintenance plan is determined. The maintenance methods corresponding to the gradual change anomaly information, the comparison result, and the equipment fault diagnosis result can be matched with the database. For example, if the critical degradation characteristic of the bearing is higher than the warning value, the health management system will prompt maintenance personnel to perform maintenance on the bearing.

[0052] S209: Perform health status management of eVTOL based on the maintenance plan and target fault isolation decision.

[0053] In the specific implementation of this invention, the health status management of the eVTOL is carried out based on the maintenance plan and the target fault isolation decision. That is, the maintenance plan and the target fault isolation decision are sent to the corresponding maintenance personnel and automatic maintenance equipment to perform maintenance on the eVTOL, thereby improving the safety of eVTOL operation and the reliability of dispatch.

[0054] In this embodiment of the invention, the operating parameters of the eVTOL equipment are acquired and divided into large-scale and small-scale datasets. Feature extraction is performed on the small-scale datasets to obtain statistical and engineering feature information. This statistical and engineering feature information is then fused to obtain fused feature information, which captures more comprehensive feature information. Rule-based reasoning is performed based on the fused feature information to obtain initial fault isolation decisions. These initial fault isolation decisions are adjusted using a lightweight artificial intelligence model, improving the response speed of fault isolation actions and accurately identifying risk factors. A large-scale artificial intelligence model is used for equipment fault diagnosis using the large-scale dataset. Equipment degradation features are extracted from the large-scale dataset to obtain equipment degradation feature information. Based on this degradation feature information and the equipment fault diagnosis results, a maintenance plan is determined, resulting in a more accurate maintenance plan and reducing false alarm rates. Based on the maintenance plan and the target fault isolation decision, the health status of the eVTOL is managed, improving the safety and economy of eVTOL operation.

[0055] Example 3 Please see Figure 3 , Figure 3 This is a schematic diagram of the structural composition of the eVTOL health management system in an embodiment of the present invention. The system includes: Dataset partitioning module 31: used to acquire the equipment operating parameters of the electric vertical take-off and landing aircraft (eVTOL) and partition the equipment operating parameters into a large-scale dataset and a small-scale dataset; Feature fusion module 32: used to extract features from the small-scale dataset, obtain statistical feature information and engineering feature information, and fuse the statistical feature information and engineering feature information to obtain fused feature information; Fault decision analysis module 33: used to perform rule reasoning based on the fused feature information to obtain an initial fault isolation decision, and to adjust the initial fault isolation decision based on a lightweight artificial intelligence model to obtain a target fault isolation decision; Fault diagnosis module 34: used to perform equipment fault diagnosis based on the large-scale dataset using an artificial intelligence model, and obtain equipment fault diagnosis results; Maintenance plan determination module 35: used to extract equipment degradation features from the large-scale dataset, obtain equipment degradation feature information, and determine a maintenance plan based on the equipment degradation feature information and equipment fault diagnosis results; Health management module 36: Used to manage the health status of eVTOL based on the maintenance plan and target fault isolation decision.

[0056] In the specific implementation of this invention, the specific implementation methods of the system items can be referred to the implementation methods of the above-mentioned method items, and will not be repeated here.

[0057] In this embodiment of the invention, the operating parameters of the eVTOL equipment are acquired and divided into large-scale and small-scale datasets. Feature extraction is performed on the small-scale datasets to obtain statistical and engineering feature information. This statistical and engineering feature information is then fused to obtain fused feature information, which captures more comprehensive feature information. Rule-based reasoning is performed based on the fused feature information to obtain initial fault isolation decisions. These initial fault isolation decisions are adjusted using a lightweight artificial intelligence model, improving the response speed of fault isolation actions and accurately identifying risk factors. A large-scale artificial intelligence model is used for equipment fault diagnosis using the large-scale dataset. Equipment degradation features are extracted from the large-scale dataset to obtain equipment degradation feature information. Based on this degradation feature information and the equipment fault diagnosis results, a maintenance plan is determined, resulting in a more accurate maintenance plan and reducing false alarm rates. Based on the maintenance plan and the target fault isolation decision, the health status of the eVTOL is managed, improving the safety and economy of eVTOL operation.

[0058] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, which may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0059] Furthermore, the above provides a detailed description of the eVTOL health management method and system provided by the embodiments of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A health management method for eVTOL, characterized in that, The method includes: Obtain the equipment operating parameters of the electric vertical takeoff and landing (eVTOL) aircraft, and divide the equipment operating parameters into large-scale datasets and small-scale datasets; Feature extraction is performed on the small-scale dataset to obtain statistical feature information and engineering feature information, and the statistical feature information and engineering feature information are fused to obtain fused feature information; Based on the fused feature information, rule reasoning is performed to obtain an initial fault isolation decision, and the initial fault isolation decision is adjusted based on a lightweight artificial intelligence model to obtain a target fault isolation decision. Based on the large-scale artificial intelligence model, the equipment fault diagnosis is performed using the large-scale dataset to obtain equipment fault diagnosis results; Equipment degradation features are extracted from the large-scale dataset to obtain equipment degradation feature information, and a maintenance plan is determined based on the equipment degradation feature information and equipment fault diagnosis results. Health status management of eVTOL is carried out based on the maintenance plan and target fault isolation decisions.

2. The health management method for eVTOL according to claim 1, characterized in that, The equipment operating parameters for acquiring electric vertical takeoff and landing (eVTOL) aircraft include: Flight data and electric drive data of eVTOL are collected based on a combination of sensors; The battery status parameters of the eVTOL are collected by the battery management system, and the remaining battery power is analyzed based on the battery status parameters to obtain the remaining battery power information. Based on the battery state parameters, perform battery health status analysis to obtain battery health status information; Based on the flight data, electric drive data, battery status parameters, remaining battery power information, and battery health status information, the device operating parameters for eVTOL are generated.

3. The health management method for eVTOL according to claim 1, characterized in that, The step of extracting features from the small-scale dataset to obtain statistical and engineering feature information, and then fusing the statistical and engineering feature information to obtain fused feature information, includes: The small-scale dataset is normalized to obtain a normalized small-scale dataset, and the normalized small-scale dataset is standardized to obtain a standardized small-scale dataset. Statistical features are extracted from a small-scale dataset after standardization to obtain statistical feature information; Engineering feature extraction is performed on a small-scale dataset after standardization to obtain engineering feature information; Determine the target weight, and then fuse the statistical feature information and engineering feature information based on the target weight to obtain fused feature information.

4. The eVTOL health management method according to claim 3, characterized in that, The process of extracting engineering features from a standardized small-scale dataset to obtain engineering feature information includes: Signal rate of change analysis is performed on a small-scale dataset after standardization to obtain signal rate of change information. Signal kurtosis analysis was performed on a small-scale dataset after standardization to obtain signal kurtosis information; Signal volatility analysis is performed on a small dataset after standardization to obtain signal volatility information, and engineering feature information is generated based on the signal rate of change information, signal kurtosis information, and signal volatility information.

5. The health management method for eVTOL according to claim 1, characterized in that, The step of performing rule-based reasoning based on the fused feature information to obtain an initial fault isolation decision includes: Acquire formal expert knowledge and construct a risk prediction reasoning framework based on the formal expert knowledge; Based on the risk prediction and reasoning framework, the fused feature information is used to perform rule-based reasoning to obtain an initial fault isolation decision.

6. The eVTOL health management method according to claim 1, characterized in that, The method of using the large-scale dataset based on the artificial intelligence model to perform equipment fault diagnosis and obtain equipment fault diagnosis results includes: The large-scale dataset is subjected to state labeling processing to obtain a state-labeled large-scale dataset; Feature extraction is performed on the large-scale dataset after state labeling to obtain target feature information; A large-scale artificial intelligence model is trained based on a sample dataset to obtain a well-trained large-scale artificial intelligence model. The target feature information is input into a trained artificial intelligence model to diagnose equipment faults and obtain equipment fault diagnosis results.

7. The eVTOL health management method according to claim 6, characterized in that, The training process of the aforementioned large-scale artificial intelligence model includes: The sample dataset is input into a large artificial intelligence model for forward propagation calculation to obtain the predicted value; The Huber loss function is calculated between the predicted and actual values. Backpropagation of the large AI model is performed based on the Huber loss function, and the parameters of the large AI model are updated using the Gradient Descent optimizer.

8. The eVTOL health management method according to claim 1, characterized in that, The step of extracting device degradation features from the large-scale dataset to obtain device degradation feature information includes: Vibration kurtosis analysis was performed on the aforementioned large-scale dataset to obtain vibration kurtosis information. Temperature change rate information is obtained by analyzing the temperature change rate based on the large-scale dataset. Vibration root mean square analysis is performed based on the large-scale dataset to obtain vibration root mean square information, and equipment degradation characteristics are determined based on the vibration kurtosis information, temperature change rate information, and vibration root mean square information.

9. The health management method for eVTOL according to claim 1, characterized in that, The process of determining a maintenance plan based on the equipment degradation characteristic information and equipment fault diagnosis results includes: Based on the aforementioned equipment degradation characteristic information, a slow-change anomaly analysis is performed to obtain slow-change anomaly information; The equipment degradation characteristic information is compared with the preset warning value to obtain the comparison result, and a maintenance plan is determined based on the slow-change anomaly information, the comparison result and the equipment fault diagnosis result.

10. A health management system for eVTOL, characterized in that, The system includes: Dataset partitioning module: used to acquire the equipment operating parameters of the electric vertical takeoff and landing (eVTOL) aircraft and partition the equipment operating parameters into large-scale datasets and small-scale datasets; Feature fusion module: used to extract features from the small-scale dataset to obtain statistical feature information and engineering feature information, and to fuse the statistical feature information and engineering feature information to obtain fused feature information; Fault Decision Analysis Module: Used to perform rule-based reasoning based on the fused feature information to obtain an initial fault isolation decision, and to adjust the initial fault isolation decision based on a lightweight artificial intelligence model to obtain a target fault isolation decision; Fault diagnosis module: used to perform equipment fault diagnosis based on the large-scale dataset using an artificial intelligence model, and obtain equipment fault diagnosis results; Maintenance plan determination module: used to extract equipment degradation features from the large-scale dataset, obtain equipment degradation feature information, and determine a maintenance plan based on the equipment degradation feature information and equipment fault diagnosis results; Health management module: used to manage the health status of eVTOL based on the maintenance plan and target fault isolation decisions.

Images