CN109866931B - Airplane throttle control method based on self-encoder - Google Patents
Airplane throttle control method based on self-encoder Download PDFInfo
- Publication number
- CN109866931B CN109866931B CN201910196261.8A CN201910196261A CN109866931B CN 109866931 B CN109866931 B CN 109866931B CN 201910196261 A CN201910196261 A CN 201910196261A CN 109866931 B CN109866931 B CN 109866931B
- Authority
- CN
- China
- Prior art keywords
- airplane
- throttle control
- encoder
- self
- network model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Landscapes
- Feedback Control In General (AREA)
Abstract
The invention provides an aircraft throttle control method based on a self-encoder, which is characterized in that aircraft parameters are normalized, then data after normalization are input into an aircraft throttle control network model, low-dimensional data features are extracted from high-dimensional data through the operation of the self-encoder, the high-dimensional data can be subjected to feature expression by using the low-dimensional data, and then the extracted low-dimensional data is used as the input of a full connection layer, so that an aircraft throttle coefficient is obtained. And then increasing or decreasing the aircraft accelerator according to the obtained aircraft accelerator coefficient. The invention effectively reduces the data volume and the parameters of the whole network, greatly improves the operation and convergence speed of the whole network, can obtain results more quickly, establishes the network for extracting the connection among all data of the airplane in an intelligent and scientific mode, greatly reduces the calculation steps in the prior art, ensures that the results are quickly obtained in a very short time, and greatly improves the accuracy.
Description
Technical Field
The invention relates to the field of deep learning and flight control, in particular to a control method based on an autoencoder in deep learning.
Background
Modern combat aircrafts are developing in highly automatic, informationized, integrated and intelligent directions, information provided for pilots is also increased explosively, and rapid and accurate operation is difficult to be performed in a short time only by the pilots. Therefore, an airplane throttle control method is sought and used for assisting decisions of pilots to achieve the purposes of improving the effectiveness of system task completion, improving operational survivability and reducing the workload of the pilots, which is particularly important in battlefield environments.
At present, most methods for controlling the aircraft accelerator utilize a certain formula to calculate attitude parameters, and then corresponding operations are performed on the accelerator by combining judgment of a pilot on real-time space conditions. For example, the proportional, derivative and integral parameters of a fuzzy PID controller are optimized by adopting a generalized adaptive genetic algorithm (GSAGA), and the airplane throttle control method for the civil airplane automatic throttle control system, such as the fuzzy PID controller, the parallel dual-redundancy automatic throttle system, the automatic throttle control based on constant speed, the automatic throttle control based on constant attack angle and the like, is designed. The method adopts a fixed calculation method, only optimizes the proportional, differential and integral parameters of the controller, improves the control effect of the system, but cannot quickly and accurately estimate the accelerator control quantity in a variable-condition combat environment. At present, the amount of information required to be received by a pilot is huge, the space combat condition changes constantly, and the control of the airplane throttle by only depending on the traditional throttle control method is not enough.
In the prior art, the airplane accelerator coefficient is calculated according to the classification of the airplane-carrying parameters and a fixed calculation method, so that the method has the defects of more required steps, long consumed time, low accuracy and incapability of intelligently controlling the accelerator of the airplane.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides an aircraft throttle control method based on a self-encoder. Compared with the existing airplane throttle control method, the method adopted by the invention is that the aircraft parameters are normalized, then the normalized data are input into the constructed airplane throttle control network model, the data are firstly calculated by the self-encoder to extract the low-dimensional data characteristics from the high-dimensional data, the low-dimensional data can be used for carrying out characteristic expression on the high-dimensional data, and then the extracted low-dimensional data are used as the input of the full link layer, so that the airplane throttle coefficient is obtained. And then increasing or decreasing the aircraft accelerator according to the obtained aircraft accelerator coefficient. The invention uses the deep learning network based on the self-encoder, effectively reduces the data volume and the parameters of the whole network, greatly improves the operation and convergence speed of the whole network, can obtain results more quickly in the process of high-speed flight of the airplane, establishes the network for extracting the connection among all data of the airplane in an intelligent and scientific mode, greatly reduces the calculation steps in the prior art, ensures that the results can be quickly obtained in a very short time, and greatly improves the accuracy.
The technical scheme adopted by the invention for solving the technical problem comprises the following steps:
step 1, acquiring flight state data of an airplane in real time, and carrying out normalization processing on the flight state data;
step 2, inputting the data after the normalization processing into an airplane throttle control network to obtain an airplane throttle coefficient; the detailed steps are as follows:
step 2.1; dividing the flight state data subjected to normalization processing in the step 1 into a training set and a test set, wherein Q% group data serve as the training set, and (100-Q)% group data serve as the test set, and 0< Q < 100;
step 2.2: designing an airplane throttle control network model based on a self-encoder, inputting training set data into the airplane throttle control network model, and training the airplane throttle control network model by using a gradient descent algorithm to obtain a trained airplane throttle control network model;
the structure of the airplane throttle control network model is as follows:
an input layer, a hidden layer, 3 full-connection layers and an output layer of the self-encoder are sequentially connected; the number of neurons in an input layer of the self-encoder is the dimension of flight state data, the number of neurons in a hidden layer of the self-encoder is n, the number of neurons in an output layer of the self-encoder is the same as that of neurons in the input layer, and the number of neurons in three full-connection layers is m, l and r respectively; the number of neurons in the output layer is 1;
step 2.3: inputting the aircraft state data of the test set into the aircraft throttle control network model trained in the step 2.2, and obtaining the output of the aircraft throttle control network model, namely the mean square error of the aircraft throttle coefficient;
step 2.4: comparing the hyperparameters of the mean square error adjusting network model of the aircraft accelerator coefficient obtained in the step 2.3, and operating the hyperparameters of the adjusting network model to obtain a network model with the minimum relative mean square error; the adjusting of the hyper-parameters of the network model specifically comprises increasing or decreasing the number of layers of the network, reducing or improving the learning rate and changing the number of learning rounds;
and step 3: and (3) after the airplane throttle coefficient is obtained in the step (2), carrying out normalization reduction on the obtained airplane throttle coefficient to obtain the throttle control quantity after the normalization reduction.
The step 2.2 of training the airplane throttle control network model by using the gradient descent algorithm comprises the following specific training steps:
(1) randomly initializing a self-encoder, inputting the airplane state data of the training set into an airplane throttle control network model, and obtaining an actual airplane throttle coefficient;
(2) comparing and calculating the actual airplane throttle coefficient with the airplane throttle coefficient in the training set to obtain the mean square error of the airplane throttle coefficient, namely the mean value of the square sum of the difference value of the theoretical airplane throttle coefficient and the actual airplane throttle coefficient;
(3) calculating an error term of each neuron according to a back propagation algorithm;
(4) calculating the gradient of the weight and the bias according to a back propagation algorithm;
(5) updating the weight and the bias according to a back propagation algorithm;
(6) and (5) repeating the steps (1) to (5) until the precision requirement is met or the set iteration number is reached.
The invention has the beneficial effects that:
1. the aircraft throttle control method based on the self-encoder provided by the invention can directly input the aircraft parameters into the constructed aircraft throttle control network, so that a large number of calculation steps in the prior art are omitted, and the result can be quickly and accurately obtained.
2. Deep learning can search potential rules from a large amount of data, and extract and summarize the rules. An airplane throttle control network based on a self-encoder is constructed, and the relation among data can be extracted through layer-by-layer feature extraction. An airplane throttle control network is scientifically established, and the airplane throttle can be controlled more efficiently and rapidly through a throttle network model.
3. The self-encoder is a representative network in deep learning, can perform dimensionality reduction on data, perform feature extraction and learning from high-dimensional data, remove redundant parameters in initial input data, and achieve the effect of expressing high-dimensional data characteristics by using fewer dimensions. The method greatly reduces the input dimensionality of the network, so that the whole network structure has fewer parameters and stronger feature extraction capability, the convergence speed of the whole network model is increased, the operation burden of the airborne computer is effectively reduced, and the resource occupancy rate is reduced. The invention constructs the airplane throttle control network based on the self-encoder, achieves the effects of high speed and high accuracy, and simultaneously consumes less resources.
4. The invention links the control and the intellectualization of the airplane throttle, can provide an auxiliary decision for the pilot under the increasingly complicated flying environment, greatly lightens the operating pressure of the pilot, provides an accurate operating suggestion, saves the resources of an onboard computer and has good development prospect.
Drawings
Fig. 1 is a flowchart illustrating the steps for implementing aircraft throttle control in accordance with the present invention.
Fig. 2 is a model of the aircraft throttle control network based on the self-encoder of the present invention.
Fig. 3 is a step of constructing an aircraft throttle control network of the present invention.
Detailed Description
The invention is further illustrated with reference to the following figures and examples.
The technical scheme of the invention is as follows: an aircraft throttle control method based on self-encoder comprises the following steps:
step 1, acquiring flight state data of an airplane in real time, and carrying out normalization processing on the flight state data;
step 2, inputting the data after the normalization processing into an airplane throttle control network to obtain an airplane throttle coefficient; the detailed steps are as follows:
step 2.1; dividing the flight state data subjected to normalization processing in the step 1 into a training set and a test set, wherein Q% group data serve as the training set, and (100-Q)% group data serve as the test set, and 0< Q < 100;
step 2.2: designing an airplane throttle control network model based on a self-encoder, inputting training set data into the airplane throttle control network model, and training the airplane throttle control network model by using a gradient descent algorithm to obtain a trained airplane throttle control network model;
the structure of the airplane throttle control network model is as follows:
an input layer, a hidden layer, 3 full-connection layers and an output layer of the self-encoder are sequentially connected; the number of neurons in an input layer of the self-encoder is the dimension of flight state data, the number of neurons in a hidden layer of the self-encoder is n, the number of neurons in an output layer of the self-encoder is the same as that of neurons in the input layer, and the number of neurons in three full-connection layers is m, l and r respectively; the number of neurons in the output layer is 1; the number of n, m, l and r is increased or decreased according to the actual situation, so that the mean square error of the accelerator control quantity is reduced.
Step 2.3: inputting the aircraft state data of the test set into the aircraft throttle control network model trained in the step 2.2, and obtaining the output of the aircraft throttle control network model, namely the mean square error of the aircraft throttle coefficient, wherein the smaller the obtained mean square error is, the closer the mean square error is to 0, the better the effect of the network model is, otherwise, the worse the effect of the network model is;
step 2.4: comparing the hyperparameters of the mean square error adjusting network model of the aircraft accelerator coefficient obtained in the step 2.3, and operating the hyperparameters of the adjusting network model to obtain a network model with the minimum relative mean square error; the adjusting of the hyper-parameters of the network model specifically comprises increasing or decreasing the number of layers of the network, reducing or improving the learning rate and changing the number of learning rounds;
and step 3: and (3) after the airplane throttle coefficient is obtained in the step (2), carrying out normalization reduction on the obtained airplane throttle coefficient to obtain the throttle control quantity after the normalization reduction.
The input in the airplane throttle control network, namely Flight state data, is acquired through a Flight simulation system (such as open source Flight simulation software Flight Gear and the like), and consists of a plurality of groups of data, wherein each group of data contains a plurality of airplane state data of each simulation step length; and the aircraft throttle control network model takes the aircraft state data after the normalization processing as input, and outputs the aircraft throttle coefficient after calculation through the aircraft throttle control network model.
The step 2.2 of training the airplane throttle control network model by using the gradient descent algorithm comprises the following specific training steps:
(1) randomly initializing a self-encoder, inputting the airplane state data of the training set into an airplane throttle control network model, and obtaining an actual airplane throttle coefficient;
(2) comparing and calculating the actual airplane throttle coefficient with the airplane throttle coefficient in the training set to obtain the mean square error of the airplane throttle coefficient, namely the mean value of the square sum of the difference value of the theoretical airplane throttle coefficient and the actual airplane throttle coefficient;
(3) calculating an error term of each neuron according to a back propagation algorithm;
(4) calculating the gradient of the weight and the bias according to a back propagation algorithm;
(5) updating the weight and the bias according to a back propagation algorithm;
(6) and (5) repeating the steps (1) to (5) until the precision requirement is met or the set iteration number is reached, wherein the iteration number is 1500.
The embodiment provides an aircraft throttle control method based on deep learning, a flow chart of which is shown in fig. 1, and the specific steps are as follows:
step 1, acquiring state data of the airplane in real time, and normalizing the data.
And 2, inputting the data after the normalization processing into an aircraft throttle control network (shown in figure 2) based on a self-encoder to obtain an aircraft throttle coefficient.
Step 3, after obtaining the accelerator coefficient of the airplane, restoring the normalized data to obtain the accelerator control quantity required by the airplane;
in this example, the airplane throttle control network is a network model constructed based on a self-encoder.
Data were collected from the flight simulation system, in this example 16628 sets of data were collected, each set containing 47 flight status data and 1 aircraft throttle coefficient for each simulation step. 10000 groups of data are adopted in the training set, and 2500 groups of data are adopted in the testing set.
The 47 flight status data in this example are: the simulation time length, the survival probability, the position components of the local machine and the enemy machine on X, Y and Z axes, the speed components of the local machine and the enemy machine on X, Y and Z axes, the course angle, the roll angle and the pitch angle of the local machine, the attack angle and the coefficient of a speed reduction plate of the local machine and 24 Boolean variables.
The 24 Boolean variables include whether the radar has a target, whether the pilot sees a target, whether the radar sensing system has a target, whether the radar sensing system can predict a target, whether the velocity is greater than the steady flight velocity, whether the tangential tank door is greater than 0, whether the target is in front of the target, whether the target is being tailed, whether the radar is tracking a target, whether there is a possibility of collision with a target, whether there are missiles that are not launched, whether the target distance is greater than 300 meters, whether the target distance is greater than 250 meters, whether there are missiles that capture a target, whether there are missiles that are ready to be launched, whether the target aircraft is within the firing range of the cannon, whether the approach velocity is greater than-1 meter/second, whether the target distance is greater than 500 meters, whether the velocity is less than the most-recently flown velocity, whether the target distance is greater than 4000 meters, whether the product of the bor, Whether the target distance is greater than 1500 meters and the azimuth angle of the target is less than 100 degrees.
In this example, the input data required for the throttle control network is 47 flight status data. The output data required by the throttle control network is the aircraft throttle coefficient.
The architecture of the self-encoder based aircraft throttle control network in this example is shown in fig. 2.
The network model based on the self-encoder is sequentially connected according to the sequence of an input layer, an encoding layer, a decoding layer and 3 full-connection layers.
The parameter data of the airplane is normalized and then set as input, and 47 neurons are set in an input layer of a self-encoder.
After the airplane parameter data is input into the self-encoder, 20-dimensional feature data capable of expressing all characteristics of the data is extracted from 47-dimensional airplane data through continuous training from encoding to decoding, and the 20-dimensional data is used as the input of a full connection layer.
The number of 3 full-connection layer neurons is 10, 8 and 4 respectively. And transmitting the 20-dimensional characteristic data obtained by the self-encoder into a network, and sequentially performing operation through a full connection layer.
And inputting the data output by the full connection layer to an output layer, and finally outputting the airplane throttle coefficient.
The method for constructing the airplane throttle control network is shown in figure 3 and comprises the following construction steps:
step 2.1: performing multiple flight simulation, collecting flight state data of the airplane and normalizing the data;
step 2.2: taking 80% group data of the normalized data as a training set, and taking 20% group data as a test set;
step 2.3: designing an airplane throttle control network model based on a self-encoder, and training by using a training set to obtain a throttle control network;
step 2.4: inputting the flight state data of the test set into the trained network, and comparing the aircraft throttle coefficient output by the network with the aircraft throttle coefficient in the test set to obtain an evaluation index;
step 2.5: and adjusting the over-parameters of the throttle control network according to the evaluation indexes, and further obtaining the network parameters which enable the mean square error function of the throttle control quantity to be minimum.
In this example, the step of training the throttle control network in step 2.3 may be divided into six steps:
the first step is as follows: an airplane throttle control network structure and parameters based on self-encoders are designed. The structure of the airplane throttle control network model based on the self-encoder is shown in figure 2.
The second step is that: and inputting the airplane state data of the training set into a self-encoder to obtain the airplane state data extracted by training of the self-encoder, wherein 20 neurons are set in a hidden layer of the self-encoder.
And transmitting the data into an input layer of a self-encoder, firstly passing through an encoding layer, and obtaining hidden layer data through data encoding operation. The operation of the coding layer is:
Y＝f(X*W+b) (1)
y is hidden layer data; x is an input function; w is a weight; b is an offset; f is the activation function.
And then, decoding the data of the hidden layer to obtain the data with the same input number. The operations of the decoding layer are as follows:
X′＝f(Y*W′+b′) (2)
y is hidden layer data; x' is an output function; w' is a weight; b' is an offset; f is the activation function.
The self-encoder enables the final output data to be as close to the input data as possible in the process of encoding and decoding. The error function set for the self-encoder network is therefore:
and updating the weight and the offset of the self-encoder by repeated operation until the error function reaches the set precision or the iteration number reaches the set maximum value. And when the training is finished, finally extracting the 20-dimensional data in the hidden layer of the encoder, and using the extracted 20-dimensional data as the input of the fully-connected layer.
The third step: and inputting the 20-dimensional airplane state data extracted from the encoder training to a full connection layer to obtain an actually output throttle coefficient.
And inputting the 20-dimensional airplane state data obtained by the self-encoder in the second step into 3 full-connection layers, and calculating in sequence. The calculation formula of the full connection layer is as follows:
Y＝∫(X*W+b) (4)
y is an output matrix; x is an input matrix; w is a weight matrix; b is a bias matrix; f is the activation function.
The fourth step: will actually outputThe accelerator coefficient and the theoretical accelerator coefficient are compared and calculated to obtain an error function Ed. The error function is mean square error, namely the expected value of the square of the difference between the theoretical throttle coefficient and the actually output throttle coefficient.
Y is a theoretical output value of the output layer; y is the actual output value of the output layer; n is the total number of training data.
The fifth step: the error term for each neuron is calculated in reverse.
For the output layer, the error term is:
f′(x)*(Y-y) (6)
x is input data of an output layer neuron; f' is the derivative of the activation function, and Y is the theoretical output value of the output layer; y is the actual output value of the output layer;
for the hidden layer, the error term is:
And a sixth step: the gradient of the weights and the bias is calculated.
wi，jThe gradient of (d) is:
Offset term WbGradient of (2)
The seventh step: and updating the weight and the bias of the network according to the error term and the gradient.
In this example, the mean square error is used as an evaluation index.
Mean square error refers to Mse, which is an expectation of calculating the square of the predicted and true difference values, and is often used to fully evaluate the quality of the network. The smaller the Mse, the better the network performance is demonstrated.
In this example, the evaluation index of the test set is obtained, and the network graph performance is evaluated according to the evaluation index, so as to adjust the network parameters. Finally, the performance of the constructed throttle control network is optimal when the hidden layer of the self-encoder is 20 neurons and the full connection layer is 3 layers.
The evaluation indexes of the test set of the throttle control network in this example are shown in table 1.
TABLE 1 Accelerator coefficient test set evaluation index
| Number of self-encoders | Number of full connection layers | Mean square error of self-encoder | Mean square error of network as a whole |
| 1 | 3 | 6.868612e-05 | 0.007874242 |
From the above table, the airplane throttle control network constructed based on the self-encoder has high precision in calculating the throttle coefficient, and can meet the requirement of the needed throttle control.
In summary, the aircraft throttle control method based on deep learning provided by the invention has the following technical advantages:
1. the aircraft throttle control method based on the self-encoder provided by the invention can directly input the aircraft parameters into the constructed aircraft throttle control network, so that a large number of calculation steps in the prior art are omitted, and the result can be quickly and accurately obtained.
2. Deep learning can search potential rules from a large amount of data, and extract and summarize the rules. An airplane throttle control network based on a self-encoder is constructed, and the relation among data can be extracted through layer-by-layer feature extraction. The self-encoder, which is a representative network in deep learning, can perform dimensionality reduction on data, perform feature extraction and learning from high-dimensional data, and express the characteristics of the high-dimensional data with fewer dimensions. The input dimensionality of the network is greatly reduced, so that the whole network structure has fewer parameters and stronger feature extraction capability, and the convergence speed of the whole network model is accelerated. The invention constructs the airplane throttle control network based on the self-encoder, and achieves the effects of high efficiency, rapidness and high accuracy.
3. The invention links the control and the intellectualization of the airplane throttle, can provide an auxiliary decision for the pilot under the increasingly complicated flying environment, greatly lightens the operating pressure of the pilot, provides an accurate operating suggestion and has good development prospect.
Claims (2)
1. An aircraft throttle control method based on a self-encoder is characterized by comprising the following steps:
step 1, acquiring flight state data of an airplane in real time, and carrying out normalization processing on the flight state data;
step 2, inputting the data after the normalization processing into an airplane throttle control network to obtain an airplane throttle coefficient; the detailed steps are as follows:
step 2.1; dividing the flight state data subjected to normalization processing in the step 1 into a training set and a test set, wherein Q% group data serve as the training set, and (100-Q)% group data serve as the test set, and 0< Q < 100;
step 2.2: designing an airplane throttle control network model based on a self-encoder, inputting training set data into the airplane throttle control network model, and training the airplane throttle control network model by using a gradient descent algorithm to obtain a trained airplane throttle control network model;
the structure of the airplane throttle control network model is as follows:
an input layer, a hidden layer, 3 full-connection layers and an output layer of the self-encoder are sequentially connected; the number of neurons in an input layer of the self-encoder is the dimension of flight state data, the number of neurons in a hidden layer of the self-encoder is n, the number of neurons in an output layer of the self-encoder is the same as that of neurons in the input layer, and the number of neurons in three full-connection layers is m, l and r respectively; the number of neurons in the output layer is 1;
step 2.3: inputting the aircraft state data of the test set into the aircraft throttle control network model trained in the step 2.2, and obtaining the output of the aircraft throttle control network model, namely the mean square error of the aircraft throttle coefficient;
step 2.4: comparing the hyperparameters of the mean square error adjusting network model of the aircraft accelerator coefficient obtained in the step 2.3, and operating the hyperparameters of the adjusting network model to obtain a network model with the minimum relative mean square error; the adjusting of the hyper-parameters of the network model specifically comprises increasing or decreasing the number of layers of the network, reducing or improving the learning rate and changing the number of learning rounds;
and step 3: and (3) after the airplane throttle coefficient is obtained in the step (2), carrying out normalization reduction on the obtained airplane throttle coefficient to obtain the throttle control quantity after the normalization reduction.
2. The aircraft throttle control method based on the self-encoder as claimed in claim 1, wherein:
the step 2.2 of training the airplane throttle control network model by using the gradient descent algorithm comprises the following specific training steps:
(1) randomly initializing a self-encoder, inputting the airplane state data of the training set into an airplane throttle control network model, and obtaining an actual airplane throttle coefficient;
(2) comparing and calculating the actual airplane throttle coefficient with the airplane throttle coefficient in the training set to obtain the mean square error of the airplane throttle coefficient, namely the mean value of the square sum of the difference value of the theoretical airplane throttle coefficient and the actual airplane throttle coefficient;
(3) calculating an error term of each neuron according to a back propagation algorithm;
(4) calculating the gradient of the weight and the bias according to a back propagation algorithm;
(5) updating the weight and the bias according to a back propagation algorithm;
(6) and (5) repeating the steps (1) to (5) until the precision requirement is met or the set iteration number is reached.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910196261.8A CN109866931B (en) | 2019-03-15 | 2019-03-15 | Airplane throttle control method based on self-encoder |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910196261.8A CN109866931B (en) | 2019-03-15 | 2019-03-15 | Airplane throttle control method based on self-encoder |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109866931A CN109866931A (en) | 2019-06-11 |
| CN109866931B true CN109866931B (en) | 2020-10-27 |
Family
ID=66920505
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910196261.8A Active CN109866931B (en) | 2019-03-15 | 2019-03-15 | Airplane throttle control method based on self-encoder |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109866931B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106707759A (en) * | 2017-02-17 | 2017-05-24 | 中国空气动力研究与发展中心计算空气动力研究所 | Airplane Herbst maneuvering control method |
| CN108182452A (en) * | 2017-12-29 | 2018-06-19 | 哈尔滨工业大学（威海） | Aero-engine fault detection method and system based on grouping convolution self-encoding encoder |
| WO2018117872A1 (en) * | 2016-12-25 | 2018-06-28 | Baomar Haitham | The intelligent autopilot system |
| CN108791910A (en) * | 2018-05-03 | 2018-11-13 | 深圳市道通智能航空技术有限公司 | A kind of method, apparatus and unmanned plane of Throttle Opening Control |
| CN108983800A (en) * | 2018-07-27 | 2018-12-11 | 西北工业大学 | A kind of aspect control method based on deep learning |
| CN109188502A (en) * | 2018-07-05 | 2019-01-11 | 中国科学技术大学 | A kind of beam transport network method for detecting abnormality and device based on self-encoding encoder |
-
2019
- 2019-03-15 CN CN201910196261.8A patent/CN109866931B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018117872A1 (en) * | 2016-12-25 | 2018-06-28 | Baomar Haitham | The intelligent autopilot system |
| CN106707759A (en) * | 2017-02-17 | 2017-05-24 | 中国空气动力研究与发展中心计算空气动力研究所 | Airplane Herbst maneuvering control method |
| CN108182452A (en) * | 2017-12-29 | 2018-06-19 | 哈尔滨工业大学（威海） | Aero-engine fault detection method and system based on grouping convolution self-encoding encoder |
| CN108791910A (en) * | 2018-05-03 | 2018-11-13 | 深圳市道通智能航空技术有限公司 | A kind of method, apparatus and unmanned plane of Throttle Opening Control |
| CN109188502A (en) * | 2018-07-05 | 2019-01-11 | 中国科学技术大学 | A kind of beam transport network method for detecting abnormality and device based on self-encoding encoder |
| CN108983800A (en) * | 2018-07-27 | 2018-12-11 | 西北工业大学 | A kind of aspect control method based on deep learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109866931A (en) | 2019-06-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102831269B (en) | Method for determining technological parameters in flow industrial process | |
| CN113158445B (en) | Prediction algorithm for residual service life of aero-engine with convolution memory residual error self-attention mechanism | |
| CN114997051A (en) | Aero-engine service life prediction and health assessment method based on transfer learning | |
| CN114048889A (en) | Aircraft trajectory prediction method based on long-term and short-term memory network | |
| CN115618733B (en) | Multi-scale hybrid attention mechanism modeling method for predicting remaining service life of aircraft engine | |
| CN107766668B (en) | Complex simulation model verification method based on neural network | |
| CN112859898B (en) | Aircraft trajectory prediction method based on two-channel bidirectional neural network | |
| CN108983800B (en) | Airplane attitude control method based on deep learning | |
| CN110082738B (en) | Radar target identification method based on Gaussian mixture and tensor recurrent neural network | |
| CN116050515B (en) | XGBoost-based parallel deduction multi-branch situation prediction method | |
| Secco et al. | Artificial neural networks applied to airplane design | |
| CN113076686A (en) | Aircraft trajectory prediction method based on social long-term and short-term memory network | |
| CN106980262A (en) | Self-adaptive flight device robust control method based on Kernel recursive least square algorithm | |
| CN110046590B (en) | One-dimensional image identification method based on particle swarm optimization deep learning feature selection | |
| CN108009320B (en) | Control-oriented multi-system association modeling method for hypersonic aircraft | |
| CN109866931B (en) | Airplane throttle control method based on self-encoder | |
| Xie et al. | Estimating a civil aircraft’s development cost with a GM (1, N) model and an MLP neural network | |
| CN117332327A (en) | Aircraft track prediction system based on semi-supervised learning and neural network | |
| CN115186378A (en) | Real-time solution method for tactical control distance in air combat simulation environment | |
| CN116757545A (en) | Multi-stage manufacturing system quality prediction method based on multi-task deep learning | |
| CN113742860B (en) | Scroll engine power estimation method based on DBN-Bayes algorithm | |
| CN114358558A (en) | Method for evaluating threat of air attack target based on experience learning of commander | |
| CN112698666B (en) | Aircraft route optimization method based on meteorological grid | |
| CN112416913B (en) | GWO-BP algorithm-based aircraft fuel system state missing value supplementing method | |
| CN105955029B (en) | A kind of pid control parameter optimization method for protecting robustness |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |