Helicopter entering autorotation intelligent judgment method and device
By establishing a helicopter power failure database and using neural networks to analyze engine and flight status parameters, the problem of rapid and accurate judgment of helicopter power failure was solved, and the safety of autorotation control was improved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA HELICOPTER RES & DEV INST
- Filing Date
- 2023-11-13
- Publication Date
- 2026-07-24
AI Technical Summary
In existing technologies, helicopters cannot quickly and accurately determine whether they have entered autorotation when power fails, resulting in control delays and affecting safety.
By establishing a helicopter power failure flight database, constructing a neural network using the LM algorithm, and integrating engine and flight status parameters, an intelligent judgment method is designed to quickly and accurately determine whether the helicopter has entered a rotational state.
It enables rapid and accurate judgment in the event of power failure, reduces control delay, improves the safety of the rotation process, and meets the safety requirements of the U.S. Federal Aviation Administration.
Smart Images
Figure CN117592185B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of helicopter flight control, specifically relating to a method and device for intelligent judgment of helicopter entering autorotation. Background Technology
[0002] When a helicopter experiences power failure during normal flight, engine parameters and flight attitude change drastically, seriously threatening the safety of the helicopter and the pilot. In such cases, timely autorotation is necessary to control the helicopter's descent and ensure a safe landing. The prerequisite for initiating autorotation is the accurate and rapid identification of a power failure, allowing the pilot to execute autorotation controls more promptly, reducing control delays, and improving the safety of the autorotation process.
[0003] When a helicopter experiences power failure, sensors cannot directly issue a failure command. The pilot must make a comprehensive judgment based on the engine status, helicopter status, and early warning information, which involves a certain delay. Currently, there are no research reports on methods for quickly and accurately determining when a helicopter enters autorotation. Summary of the Invention
[0004] This invention proposes an intelligent method for judging whether a helicopter is entering autorotation. It can analyze engine and flight status information to quickly and accurately judge whether it is entering autorotation, while avoiding misjudging state changes caused by special flight maneuvers and transmission mechanism failures as power failure.
[0005] Firstly, this application provides a method for intelligently determining when a helicopter enters autorotation, the method comprising:
[0006] Step 1: Establish a helicopter power failure flight database;
[0007] Step 2: Select engine parameters and flight status parameters based on the power failure characteristics, and process the data accordingly;
[0008] Step 3: Construct a training and learning method for the neural network using the LM (Levenberg-Marquardt) algorithm;
[0009] Step 4: Based on the neural network training and learning method, the helicopter power failure flight database, engine parameters and flight state parameters, determine the network structure and train the network to obtain the trained neural network structure;
[0010] Step 5: Use the trained neural network structure to obtain the self-rotation judgment result.
[0011] Specifically, step 1 includes:
[0012] Based on the rotational flight dynamics model, different power failure flight scenarios were set up using a flight simulator, and a large number of flight simulations were carried out to obtain a flight database corresponding to different flight scenarios and failure responses.
[0013] Specifically, dynamic failure characteristics include:
[0014] The torque decreases to near zero at a relatively high rate of decay.
[0015] The speed of the gas turbine decreases continuously, and the rate of decrease is inversely proportional to the moment of inertia, so it will not decrease sharply to 0.
[0016] The rotational speed of the power turbine gradually decreases, and the rate of decrease is inversely proportional to the moment of inertia, so it will not decrease sharply to 0.
[0017] As the rotor speed gradually decreases, the decay rate is directly proportional to the initial torque and inversely proportional to the moment of inertia.
[0018] Yaw rate occurs, and yaw angle increases. When the flight speed is high, the yaw rate is suppressed by yaw damping.
[0019] Specifically, engine parameters include engine torque Q and gas turbine speed N during helicopter flight. g Power turbine speed N P Rotor speed N r ;
[0020] Flight status parameters include yaw angular rate r and vertical acceleration a. z Vertical velocity v z The parameters are: heading angle ψ, pitch rate q, and roll rate p.
[0021] Specifically, step 5 includes:
[0022] Step 51: Input the helicopter's current flight parameters into the trained neural network structure;
[0023] Step 52: If the helicopter is in an engine failure state, it will enter autorotation; if the helicopter is in normal flight, it will not enter autorotation.
[0024] Secondly, this application provides a helicopter entering autorotation intelligent judgment device, the device comprising an establishment unit, a parameter selection unit, a construction unit, and a judgment unit, wherein:
[0025] Establish a unit for building a helicopter power failure flight database;
[0026] The parameter selection unit is used to select engine parameters and flight status parameters based on power failure characteristics, and to process the data.
[0027] The building unit is used to construct a training and learning method for a neural network using the LM (Levenberg-Marquardt) algorithm. Based on the training and learning method of the neural network, the helicopter power failure flight database, engine parameters and flight state parameters, the network structure is determined and the network is trained to obtain a trained neural network structure.
[0028] The judgment unit is used to obtain the self-rotation judgment result using the trained neural network structure.
[0029] Specifically, dynamic failure characteristics include:
[0030] The torque decreases to near zero at a relatively high rate of decay.
[0031] The speed of the gas turbine decreases continuously, and the rate of decrease is inversely proportional to the moment of inertia, so it will not decrease sharply to 0.
[0032] The rotational speed of the power turbine gradually decreases, and the rate of decrease is inversely proportional to the moment of inertia, so it will not decrease sharply to 0.
[0033] As the rotor speed gradually decreases, the decay rate is directly proportional to the initial torque and inversely proportional to the moment of inertia.
[0034] Yaw rate occurs, and yaw angle increases. When the flight speed is high, the yaw rate is suppressed by yaw damping.
[0035] Specifically, engine parameters include engine torque Q and gas turbine speed N during helicopter flight. g Power turbine speed N P Rotor speed N r ;
[0036] Flight status parameters include yaw angular rate r and vertical acceleration a. z Vertical velocity v z The parameters are: heading angle ψ, pitch rate q, and roll rate p.
[0037] In summary, the technical problem solved by this invention is to provide a method for intelligently determining whether a helicopter needs to enter autorotation flight when a helicopter may experience power failure, based on changes in engine parameters and flight status. Attached Figure Description
[0038] Figure 1 This application provides a technical framework diagram for intelligent judgment of entry rotation.
[0039] Figure 2 The torque change curve after power failure is provided for this application;
[0040] Figure 3The turbine speed change curve provided in this application after power failure;
[0041] Figure 4 The rotor speed change curve provided in this application after power failure;
[0042] Figure 5 A curve showing the change in heading angular rate after power failure provided for this application;
[0043] Figure 6 The diagram of the neural network structure for entering the self-rotation judgment provided in this application;
[0044] Figure 7 A diagram illustrating the neural network intelligent judgment method provided in this application;
[0045] Figure 8 The error function graph provided for this application;
[0046] Figure 9 Training accuracy graph provided for this application;
[0047] Figure 10 The simulation diagram for intelligent judgment after power failure provided in this application. Detailed Implementation
[0048] Example 1
[0049] like Figure 1 As shown, this application provides a method for intelligently determining when a helicopter enters autorotation, the method comprising:
[0050] Step 1: Establish a helicopter power failure flight database;
[0051] Specifically, step 1 includes: based on the autorotation flight dynamics model, using a flight simulator to set different power failure flight scenarios, mainly including: power failure when flight control functions are not engaged, power failure during attitude maintenance, power failure during airspeed maintenance, power failure during altitude maintenance, power failure during coordinated turns, and power failure during stick control. Extensive flight simulations are conducted to ultimately obtain a flight database corresponding to different flight scenarios and failure responses.
[0052] Step 2: Select engine parameters and flight status parameters based on the power failure characteristics, and process the data accordingly;
[0053] Specifically, dynamic failure characteristics include:
[0054] The torque decreases to near zero at a relatively high rate of decay.
[0055] The speed of the gas turbine decreases continuously, and the rate of decrease is inversely proportional to the moment of inertia, so it will not decrease sharply to 0.
[0056] The rotational speed of the power turbine gradually decreases, and the rate of decrease is inversely proportional to the moment of inertia, so it will not decrease sharply to 0.
[0057] As the rotor speed gradually decreases, the decay rate is directly proportional to the initial torque and inversely proportional to the moment of inertia.
[0058] Yaw rate occurs, and yaw angle increases. When the flight speed is high, the yaw rate is suppressed by yaw damping.
[0059] The engine parameters include the engine torque Q and the gas turbine speed N during helicopter flight. g Power turbine speed N P Rotor speed N r Flight state parameters include yaw angular rate r and vertical acceleration a. z Vertical velocity v z Parameters such as heading angle ψ, pitch rate q, and roll rate p.
[0060] The difference in magnitude between different numerical features is large. In order to facilitate calculation and eliminate the influence of the data units of different features on the network convergence efficiency and accuracy, the feature data is normalized.
[0061] Step 3: Use the LM (Levenberg-Marquardt) algorithm to construct a training and learning method for the neural network structure.
[0062] Among them, the LM algorithm is an optimized training algorithm with advantages such as strong robustness, adaptability and fast convergence. It has a good effect on learning large-scale networks. The training process uses backpropagation of errors to correct and update weights.
[0063] Activation function. The tan Sigmoid function, which performs well, is chosen as the activation function between different layers of the learning network.
[0064] Loss function. In this invention, the mean squared error is used as the evaluation metric for learning.
[0065] Step 4: Based on the training and learning method of the neural network structure, the helicopter power failure flight database, engine parameters and flight state parameters, determine the network structure and train the network to obtain the trained neural network structure;
[0066] It should be noted that the network structure in this application selects 12 features, including 10 inputs and 2 outputs. The output features represent the helicopter's power failure states, specifically engine failure and normal flight. The structure of the neural network affects the training effect; selecting fewer hidden layers results in relatively lower model accuracy, while selecting too many increases computational cost, making the training process more difficult and time-consuming. Optimizing the number of hidden layers and nodes in each layer, setting it to 10, yields better training results. Figure 8 and Figure 9 As shown.
[0067] Step 5: Use the trained neural network structure to obtain the self-rotation judgment result.
[0068] Specifically, step 5 includes:
[0069] Step 51: Input the helicopter's current flight parameters into the trained neural network structure;
[0070] The helicopter's current flight parameters include engine parameters and flight status parameters.
[0071] Step 52: If the helicopter is in an engine failure state, it will enter autorotation;
[0072] If the helicopter is in normal flight mode, it will not enter autorotation.
[0073] Example 2
[0074] This invention utilizes an intelligent judgment algorithm to comprehensively analyze changes in helicopter instrument data, engine and flight status, and other information to determine whether timely autorotation is necessary. This improves the speed and accuracy of the judgment, avoiding misinterpretations of instrument data and flight status changes caused by special flight maneuvers or mechanical malfunctions. Specifically, it consists of three stages: power failure characteristic analysis, key criterion selection, and neural network intelligent judgment. Figure 1 As shown.
[0075] (1) Dynamic failure characteristics analysis:
[0076] Based on a rotational dynamics model, a comprehensive analysis of the flight characteristics after engine failure is conducted to determine the flight parameters related to power failure, thus avoiding the influence of special flight operations or other fault scenarios on subsequent judgment logic. Helicopter power failure is mainly manifested in drastic changes in engine parameters and flight status parameters.
[0077] Changes in key parameters of engine A
[0078] Helicopter engines mainly involve components such as compressors, combustion chambers, gas turbines, and power turbines. The output torque of the power turbine is:
[0079]
[0080] Where N P Where W45 is the turbine rotational speed, W45 is the atmospheric mass flow rate entering the turbine, and ΔH is the turbine rotational speed. PT K represents the enthalpy change at the power turbine. P N is the coefficient. Pd The required rotational speed.
[0081] Turbine speed can be calculated based on the conservation of angular momentum, where the gas turbine speed is:
[0082]
[0083] Q GT Q is the output torque of the gas generator. C The required torque for the compressor.
[0084] The power turbine speed is:
[0085]
[0086] Q req This is the required power.
[0087] Based on the above formula, the changing trends of different parameters after engine failure can be analyzed:
[0088] Torque: After power failure, the mass flow rate of the atmosphere entering the power turbine decreases sharply, causing the torque to drop to near zero at a large decay rate. Figure 2 As shown.
[0089] Gas turbine speed: As the output torque of the gas generator continuously decreases, the speed will continuously decline. The rate of decline is inversely proportional to the moment of inertia, and it will not rapidly decline to 0. Figure 3 As shown.
[0090] Power turbine speed: As the output torque of the power turbine continuously decreases, the speed will continuously decrease. The rate of decrease is inversely proportional to the moment of inertia and will not decrease sharply to 0.
[0091] b. Changes in key flight status parameters
[0092] When a helicopter experiences power failure and its engine output power rapidly decreases to zero, the rotor speed will gradually decrease under the action of counter-torque. The mathematical expression for this rotor speed can be described as follows:
[0093]
[0094] In the formula: I is the sum of the rotational inertia of moving parts such as the rotor, main reducer, and tail rotor; Ω is the rotor speed; and Q is the rotor counter-torque.
[0095] The decay law of rotor speed can be obtained as follows:
[0096]
[0097] Where Ω0 is the initial rotor speed and Q0 is the initial rotor counter-torque.
[0098] It can be seen that the decay rate of the rotor speed is mainly related to the initial torque and the moment of inertia, being directly proportional to the initial torque and inversely proportional to the moment of inertia, such as... Figure 4 As shown.
[0099] Furthermore, after engine failure, the helicopter's yaw path will change rapidly and significantly. When the helicopter is stable, the yaw moment at the center of gravity is:
[0100] M Z =Q MR -M Y -M T -M V
[0101] Q MR M is the rotor torque; Y M is the yaw moment generated by the lateral force of the rotor; T M is the yaw moment generated by the tail rotor. V The yaw moment generated by the fuselage and vertical tail.
[0102] Consider the yaw dynamics equation as follows:
[0103]
[0104] In the formula: r is the yaw angular velocity; I x I y I z ρ is the moment of inertia of each axis of the helicopter fuselage; p is the roll rate and q is the pitch rate.
[0105] The unbalanced yaw moment following power failure will cause the helicopter to produce the following yaw response:
[0106]
[0107] When the engine fails, the engine torque rapidly decreases to zero, resulting in a large yaw acceleration. The yaw damping generated by this yaw acceleration then helps to suppress the yaw rate. Figure 5 As shown.
[0108] (2) Selection of key criteria:
[0109] Based on the analysis of power failure characteristics, when a manned helicopter experiences power failure during actual flight, it can be judged mainly based on the following two aspects:
[0110] a. Changes in helicopter instruments.
[0111] After power failure, instrument data will change significantly. Engine torque will drop to zero (in actual flight it is close to zero); gas turbine speed and power turbine speed will drop sharply, leading to a decrease in rotor speed. Simultaneously, warnings such as oil pressure warning and speed warning will appear.
[0112] b. Changes in helicopter flight status.
[0113] After power failure, the helicopter's counter-torque disappears, at which point the tail rotor thrust causes the helicopter to yaw in the same direction as the main rotor rotation; the reduced rotor speed leads to decreased lift, which in turn reduces flight altitude; and the altered force and torque balance causes the helicopter to roll and pitch...
[0114] However, some special flight maneuvers and malfunctions during flight can also cause changes in flight parameters. For example, during autorotation simulation training or when the engine output shaft breaks, the rotor and main reducer disengage, which can lead to a decrease in torque and speed, but the speed of the gas turbine and power turbine remains relatively normal. Strong disturbances such as gusts and wind shear can also cause sudden changes in heading, attitude and altitude, but the relevant engine parameters remain normal.
[0115] By using a simulator to conduct flight simulations of power failure response under different flight scenarios, the response characteristics of parameters that are only related to power failure are analyzed, providing target parameters for formulating power failure criteria.
[0116] Based on the above analysis, helicopter power failure, special flight operations, and other malfunctions can all cause changes in engine parameters and flight status. This invention selects parameters strongly correlated with power failure response as key parameters, and combines them with various flight status parameters as comprehensive criteria. These mainly include: engine torque Q and gas turbine speed N during helicopter flight. g Power turbine speed N P Rotor speed N r Angular rate of heading r, vertical acceleration a z Vertical velocity v z Parameters such as heading angle ψ, pitch rate q, and roll rate p.
[0117] (3) Neural Network Intelligent Judgment Method:
[0118] To address the challenges of complex flight characteristics and intricate combinations of key parameters in judgment indicators caused by different flight scenarios and special flight maneuvers, this invention utilizes neural networks to design an intelligent judgment method that improves judgment speed and accuracy.
[0119] Given the powerful nonlinear fitting capability of neural networks, the neural network designed in this invention takes rotor speed, engine parameters, flight attitude, and other information as input, and outputs whether the helicopter's power has failed. Figure 6 As shown, the gradient descent method is used to minimize the loss between the predicted and actual values, and a deep network model is learned to characterize the mapping relationship between flight state and power failure.
[0120] A neural network-based intelligent judgment technology framework for entering the self-rotation judgment, such as... Figure 7 As shown.
[0121] The specific steps of this method are as follows:
[0122] a. Establish a helicopter power failure flight database
[0123] Based on a rotational flight dynamics model, different power failure flight scenarios were set up using a flight simulator, mainly including: power failure when flight control functions are not engaged, power failure during attitude maintenance, power failure during airspeed maintenance, power failure during altitude maintenance, power failure during coordinated turns, and power failure during stick control. Extensive flight simulations were conducted, with data collected for each scenario for 30 seconds of normal flight and 10 seconds after power failure. Finally, a flight database corresponding to different flight scenarios and failure responses was obtained.
[0124] b. Feature parameter selection and data processing
[0125] Based on the key parameters obtained from the previous section, these are used as input features, specifically the engine torque Q and gas turbine speed N during helicopter flight. g Power turbine speed N P Rotor speed N r Angular rate of heading r, vertical acceleration a z Vertical velocity v z Parameters such as heading angle ψ, pitch rate q, and roll rate p are included. The output characteristics represent the helicopter's power failure state, specifically engine failure and normal flight conditions.
[0126] The difference in magnitude between different numerical features is large. In order to facilitate calculation and eliminate the influence of the data units of different features on the network convergence efficiency and accuracy, the feature data is normalized.
[0127] c. Constructing learning algorithms
[0128] To address the issues of slow convergence, long training time, and susceptibility to local optima in gradient descent, the Levenberg-Marquardt (LM) algorithm is proposed. The LM algorithm is an optimized training algorithm with advantages such as strong robustness, adaptability, and fast convergence, demonstrating excellent performance for learning large-scale networks. The training process employs backpropagation to correct and update weights.
[0129] Activation function. The tan Sigmoid function, which performs well, is chosen as the activation function between different layers of the learning network.
[0130] Loss function. In this invention, the mean squared error (MSE) is used as the evaluation metric for learning.
[0131] d. Determine the network structure and train the network
[0132] This paper selects 12 features, including 10 inputs and 2 outputs. The structure of the neural network affects the training effect. Selecting fewer hidden layers results in relatively lower model accuracy, while selecting too many increases the computational load, making the training process more difficult and time-consuming. Optimizing the number of hidden layers and the number of nodes in each layer, setting it to 10, yields better training results.
[0133] e. Enter the rotation judgment result
[0134] By inputting the helicopter's current flight parameters into a trained neural network structure, the system can quickly determine whether the helicopter is experiencing power failure or normal flight, thereby deciding whether to initiate rotation.
[0135] Key points of this invention:
[0136] a) Based on the autorotation flight dynamics model, the significant changes in typical parameters of the engine and flight status after helicopter power failure were analyzed. The main characteristics are: torque rapidly decreases to zero; gas turbine and power turbine speeds continuously decrease; rotor speed continuously decreases, with the decrease rate being proportional to the initial torque and inversely proportional to the moment of inertia; yaw acceleration suddenly increases, and yaw response is large; descent rate increases, and altitude gradually and rapidly decreases.
[0137] b) By analyzing the parameter change patterns after engine failure, key state parameters are selected as the criteria for entering self-rotation. This invention selects engine torque Q and gas turbine speed N. g Power turbine speed N P Rotor speed N r Angular rate of heading r, vertical acceleration a z Vertical velocity v z Parameters such as heading angle ψ, pitch rate q, and roll rate p are used as criteria.
[0138] c) This invention utilizes simulator flight simulation to establish a flight database for power failure under different scenarios, thereby improving the accuracy of the judgment results. This mainly includes: power failure when flight control functions are not engaged, power failure during attitude maintenance, power failure during airspeed maintenance, power failure during altitude maintenance, power failure during coordinated turns, and power failure during stick control.
[0139] d) A multi-layer neural network structure was designed, and the correspondence between the entry spin criterion and dynamic failure was established. To improve the network training effect, on the one hand, the higher-performance LM training algorithm was selected, and on the other hand, the number of hidden layers and nodes was appropriately increased to improve training accuracy.
[0140] Example 3
[0141] Taking a certain type of manned helicopter as an example, flight data under different scenarios of power failure were accumulated through desktop simulation and simulator flight simulation to form a training database. A total of 2932 sets of data were selected, of which 2502 sets were used as the training set, 440 sets as the validation set, and 440 sets as the test set. A neural network with 10 inputs and 2 outputs was trained, with 18 hidden layers. The training results are shown below. Figure 8 and Figure 9 As shown, the model trained under these conditions achieves an accuracy of 99.9%.
[0142] Flight data from a helicopter that experienced a power failure during a normal flight was selected, and the above training results were used to make a judgment. The results are as follows. Figure 10 As shown in the diagram, the state is 0 during normal flight and 1 during power failure. In the sample data, power failure occurred at 43.8 seconds, and the intelligent judgment method detected it at 44 seconds, meaning it issued an autorotation signal 0.2 seconds after power failure. The US Federal Aviation Administration stipulates that the pilot's autorotation operation should not lag by more than 1 second when the engine fails. The judgment method of this invention takes less time than this requirement, allowing for timely autorotation and notification of subsequent operational safety.
[0143] This invention analyzes and studies the response characteristics, power failure criteria, and rapid judgment methods for power failure during normal flight of a manned helicopter, and designs an intelligent judgment method by integrating multiple criteria. Simulation results show that the intelligent judgment method for helicopter entering autorotation designed in this invention can quickly and accurately determine power failure, thereby providing an autorotation signal.
Claims
1. A method for intelligently determining when a helicopter enters a rotational state, characterized in that, The method includes: Step 1: Establish a helicopter power failure flight database; Step 2: Select engine parameters and flight status parameters based on the power failure characteristics, and process the data accordingly; Step 3: Construct a training and learning method for the neural network using the LM (Levenberg-Marquardt) algorithm; Step 4: Based on the neural network training and learning method, the helicopter power failure flight database, engine parameters and flight state parameters, determine the network structure and train the network to obtain the trained neural network structure; Step 5: Use the trained neural network structure to obtain the self-rotation judgment result.
2. The method according to claim 1, characterized in that, Step 1 includes: Based on the rotational flight dynamics model, different power failure flight scenarios were set up using a flight simulator, and a large number of flight simulations were carried out to obtain a flight database corresponding to different flight scenarios and failure responses.
3. The method according to claim 1, characterized in that, Dynamic failure characteristics include: The torque decreases to near zero at a relatively high rate of decay. The speed of the gas turbine decreases continuously, and the rate of decrease is inversely proportional to the moment of inertia, so it will not decrease sharply to 0. The rotational speed of the power turbine gradually decreases, and the rate of decrease is inversely proportional to the moment of inertia, so it will not decrease sharply to 0. As the rotor speed gradually decreases, the decay rate is directly proportional to the initial torque and inversely proportional to the moment of inertia. Yaw rate occurs, and yaw angle increases. When the flight speed is high, the yaw rate is suppressed by yaw damping.
4. The method according to claim 1, characterized in that, Engine parameters include engine torque Q and gas turbine speed N during helicopter flight. g Power turbine speed N P Rotor speed N r ; Flight status parameters include yaw angular rate r and vertical acceleration a. z Vertical velocity v z The parameters are: heading angle ψ, pitch rate q, and roll rate p.
5. The method according to claim 1, characterized in that, Step 5 includes: Step 51: Input the helicopter's current flight parameters into the trained neural network structure; Step 52: If the helicopter is in an engine failure state, it will enter autorotation; if the helicopter is in normal flight, it will not enter autorotation.
6. A helicopter entering an autorotation intelligent judgment device, characterized in that, The device includes an establishment unit, a parameter selection unit, a construction unit, and a judgment unit, wherein: Establish a unit for building a helicopter power failure flight database; The parameter selection unit is used to select engine parameters and flight status parameters based on power failure characteristics, and to process the data. The building unit is used to construct a training and learning method for a neural network using the LM (Levenberg-Marquardt) algorithm. Based on the training and learning method of the neural network structure, the helicopter power failure flight database, engine parameters and flight state parameters, the network structure is determined and the network is trained to obtain a trained neural network structure. The judgment unit is used to obtain the self-rotation judgment result using the trained neural network structure.
7. The apparatus according to claim 6, characterized in that, Dynamic failure characteristics include: The torque decreases to near zero at a relatively high rate of decay. The speed of the gas turbine decreases continuously, and the rate of decrease is inversely proportional to the moment of inertia, so it will not decrease sharply to 0. The rotational speed of the power turbine gradually decreases, and the rate of decrease is inversely proportional to the moment of inertia, so it will not decrease sharply to 0. As the rotor speed gradually decreases, the decay rate is directly proportional to the initial torque and inversely proportional to the moment of inertia. Yaw rate occurs, and yaw angle increases. When the flight speed is high, the yaw rate is suppressed by yaw damping.
8. The apparatus according to claim 6, characterized in that, Engine parameters include engine torque Q and gas turbine speed N during helicopter flight. g Power turbine speed N P Rotor speed N r ; Flight status parameters include yaw angular rate r and vertical acceleration a. z Vertical velocity v z The parameters are: heading angle ψ, pitch rate q, and roll rate p.