CN114296343B - Deep reinforcement learning-based aeroengine compression part characteristic correction method - Google Patents
Deep reinforcement learning-based aeroengine compression part characteristic correction method Download PDFInfo
- Publication number
- CN114296343B CN114296343B CN202111351956.2A CN202111351956A CN114296343B CN 114296343 B CN114296343 B CN 114296343B CN 202111351956 A CN202111351956 A CN 202111351956A CN 114296343 B CN114296343 B CN 114296343B
- Authority
- CN
- China
- Prior art keywords
- network
- model
- deep
- engine
- aeroengine
- 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
- 238000012937 correction Methods 0.000 title claims abstract description 49
- 238000000034 method Methods 0.000 title claims abstract description 28
- 230000006835 compression Effects 0.000 title claims abstract description 27
- 238000007906 compression Methods 0.000 title claims abstract description 27
- 230000002787 reinforcement Effects 0.000 title claims abstract description 24
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 29
- 238000013178 mathematical model Methods 0.000 claims abstract description 23
- 238000005259 measurement Methods 0.000 claims abstract description 22
- 238000010586 diagram Methods 0.000 claims abstract description 16
- 230000009471 action Effects 0.000 claims description 30
- 238000004088 simulation Methods 0.000 claims description 28
- 238000013528 artificial neural network Methods 0.000 claims description 13
- 238000012360 testing method Methods 0.000 claims description 11
- 238000012549 training Methods 0.000 claims description 11
- 239000000446 fuel Substances 0.000 claims description 9
- 238000013461 design Methods 0.000 claims description 7
- 230000006870 function Effects 0.000 claims description 6
- 230000015556 catabolic process Effects 0.000 abstract description 2
- 238000006731 degradation reaction Methods 0.000 abstract description 2
- 230000000694 effects Effects 0.000 abstract description 2
- 238000003745 diagnosis Methods 0.000 abstract 1
- 238000004364 calculation method Methods 0.000 description 6
- 230000008569 process Effects 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000006978 adaptation Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000003628 erosive effect Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000012821 model calculation Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
Abstract
The invention discloses an aeroengine compression part characteristic correction method based on deep reinforcement learning, which comprises the steps of firstly establishing an aeroengine part level mathematical model, calculating a measurement error between the mathematical model and an actual aeroengine, defining a correction coefficient based on the compression part characteristic of an engine nonlinear part model, correcting a compression part characteristic diagram through autonomous learning by adopting a value-based deep reinforcement learning algorithm DQN, and calculating a compression part characteristic correction coefficient; finally, updating the characteristic curve of the aircraft engine component level mathematical model to reduce the error of output data; the invention solves the problems of low precision and weak generalization capability of the existing component-level model, is suitable for the correction of the model after the performance degradation of the engine, and has positive promotion effects on the health management of the engine, the self-adaptive correction of the model, the fault diagnosis of the sensor and the like.
Description
Technical Field
The invention relates to the technical field of aeroengines, in particular to a characteristic correction method for compression parts of an aeroengine based on deep reinforcement learning.
Background
In the field of overall performance simulation and control algorithm design of aeroengines, researchers need to know the accurate performance state of typical components of the current engine, namely the actual component characteristics of the engine, so as to calculate or diagnose the overall performance of the engine. When the new engine is shipped, an engine manufacturer establishes an engine reference performance calculation model according to the part characteristics obtained by part characteristic tests or theoretical calculations before the new engine is shipped. However, for reasons such as blade fouling or erosion caused by long-term service, the performance of engine components will naturally deteriorate with the increase of the number of times of use, and as a result, the original component characteristics in the rated state deviate from the actual performance of the components in the degraded state. At this time, if the engine performance calculation is performed using the component characteristics before the degradation in the rated state, a large modeling error is definitely brought, so that a large error occurs between the engine model calculation result and the test data. In view of the above, it is a very important task for aircraft engine researchers to explore component property modification techniques.
Since the 80 s of the 20 th century, research in the field of artificial intelligence has been gradually rising. As a representative branch of the artificial intelligence field, deep reinforcement learning related theory and technology has been rapidly developed. By utilizing the strong generalization capability of the neural network in the deep learning and combining with the reinforcement learning theory, people can solve the complex decision problem in a high-dimensional state space by reinforcement learning. In the calculation process, the neural network plays a role of human brain, helps the algorithm to make efficient decisions, and continuously learns and evolves in the decision process, so that more correct decisions are made. In addition, in the process of training the neural network, the deep reinforcement learning can continuously generate new training data, so that a user is not required to prepare additional training data, the training efficiency is high, and a large number of data processing flows are saved.
Disclosure of Invention
The invention aims to: aiming at the problems in the background art, the invention provides a characteristic correction method for an aeroengine compression part based on deep reinforcement learning, which can lead a neural network to autonomously learn how to correct a characteristic diagram in a simulation environment according to an initial simulation error of an engine model and select a final correction coefficient.
The technical scheme is as follows: in order to achieve the above purpose, the invention adopts the following technical scheme:
the characteristic correction method for the compression part of the aeroengine based on deep reinforcement learning is characterized by comprising the following steps of:
step S1, establishing an aeroengine component level mathematical model, calculating a measurement error between the mathematical model and an actual aeroengine, and defining a correction coefficient based on the characteristics of a compression component of a nonlinear component model of the engine;
s2, correcting the characteristic diagram of the compression part by adopting a value-based deep reinforcement learning algorithm DQN through autonomous learning, and calculating a characteristic correction coefficient of the compression part; and finally, updating the characteristic curve of the aircraft engine component level mathematical model to reduce the error of output data.
Further, the measurement error E between the mathematical model and the actual aeroengine in the step S1 is represented as follows:
wherein k represents the number of test states, m represents the number of parameters obtained by the test, y represents the value of the engine test parameter, y p Calculated values of parameters representing mathematical models, w i Weights representing corresponding parameters;
based on the characteristics of each compression element at different rotational speeds, a correction coefficient is defined as follows:
wherein a, b and c are undetermined coefficients; n is n i Represents the rotation speed of a non-design point, n D Representing the rotation speed of a design point, and a subscript i represents the position of a rotation speed line in a component characteristic diagram; c (C) π,i ，C W,i And C η,i The correction coefficients of the pressure ratio, the flow rate, and the efficiency are sequentially shown when the rotation speed is i.
Further, a value-based deep reinforcement learning algorithm DQN is adopted to correct the characteristic diagram, namely, the DQN is used for searching undetermined coefficients a, b and c, so that the measurement error between a simulation result of the digital model and an actual aeroengine is the smallest; in particular, the method comprises the steps of,
step S2.1, depth Q-based network and notationThe structure of the memory library network defines a DQN algorithm, input and output parameters and initializes the network; the depth Q network and the memory bank network are all full-connection layer networks with hidden layers larger than 2 layers, and the input end comprises flight conditions, fuel flow, nozzle areas, guide vane angle parameters and current model errors of all measurement state points of the engine; the input of each network in the DQN algorithm is a state vector, denoted as phi(s), the output is the value of all actions in the action set A, denoted as Q A The method comprises the steps of carrying out a first treatment on the surface of the The action set A in the DQN algorithm comprises the range of all a, b and c allowed variations; the DQN algorithm also comprises input iteration times T, a batch gradient descent sample number m and an attenuation factor gamma;
s2.2, after defining each input and output item and initializing of the DQN algorithm, finishing the weight updating of the deep Q network through self-learning, and correcting the model;
s2.2.1, taking the measurement error E, the fuel flow, the nozzle area, the guide vane angle and the flight condition of each measurement state point of the engine between the digital model and the actual aeroengine in the step S1 as phi (S), inputting the phi (S) into a depth Q network, and obtaining the Q values of all actions in an action set A calculated by the Q network; selecting a corresponding set of actions A among all current outputs according to an E greedy algorithm j So that the set of correction coefficients most likely reduces the simulation error of the mathematical model;
step S2.2.2 based on the action set A j And after the acquired correction coefficient corrects the mathematical model, further calculating a simulation error, namely E ', forming a new state vector phi (s') with the fuel flow, the nozzle area, the guide vane angle and the flight condition of each measurement state point of the engine, and calculating a reward function R, wherein the specific formula is as follows:
s2.2.3, selecting action A of the current Q network by using phi(s) used in the deep Q network and phi (s') output by the Q network j The action gets a prize R j Storing into a memory bank network; and use after updatingThe state Φ (s') of (c) as input to the depth network starts a new round of computation. After m samples are stored, the Q value y of the Q network selected action in the latest state is calculated as follows:
wherein S is t Represents the termination state, w net Represents weights in a deep Q network, Q (w net A, s represents the Q value calculated by the depth Q network;
step S2.2.4, based on the following mean square error loss function:
updating weight parameters w in all deep Q networks by neural network gradient back propagation net ；
And S2.3, after the deep Q network training is completed, finding out a proper characteristic correction coefficient according to the simulation error and the given range of the allowed variation of a, b and c.
The beneficial effects are that:
the invention provides a characteristic correction method for an aeroengine compression part based on deep reinforcement learning, which can lead a neural network to autonomously learn how to correct a characteristic diagram in a simulation environment according to an initial simulation error of an engine model and select a final correction coefficient. The self-learning capability of deep reinforcement learning is utilized, the problem that the characteristic curve cannot be corrected is effectively solved, and the precision of the characteristic of the component is improved. The method is suitable for correction of various gas turbine engine models.
Drawings
FIG. 1 is a flow chart of a method for correcting characteristics of compression parts of an aeroengine;
FIG. 2 is a graph comparing flow-efficiency curves before and after correction in a simulated embodiment of the invention;
FIG. 3 is a graph comparing flow-to-pressure ratio curves before and after correction in a simulated embodiment of the invention.
Detailed Description
The invention will be further described with reference to the accompanying drawings. It will be apparent that the described embodiments are some, but not all, embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The invention provides an aeroengine compression part characteristic correction method based on deep reinforcement learning, which is specifically shown in fig. 1.
And S1, establishing an aeroengine component level mathematical model, calculating a measurement error between the mathematical model and an actual aeroengine, and defining a correction coefficient based on the compression component characteristics of the engine nonlinear component model.
The measurement error E between the mathematical model established first and the actual aeroengine is expressed as follows:
wherein k represents the number of test states, m represents the number of parameters obtained by the test, y represents the value of the engine test parameter, y p Calculated values of parameters representing mathematical models, w i Weights representing corresponding parameters;
then, based on the characteristics of each compression element at different rotational speeds, correction coefficients are defined as follows:
wherein a, b and c are undetermined coefficients; n is n i Represents the rotation speed of a non-design point, n D Representing the rotation speed of a design point, and a subscript i represents the position of a rotation speed line in a component characteristic diagram; c (C) π,i ，C W,i And C η,i The correction coefficients of the pressure ratio, the flow rate, and the efficiency are sequentially shown when the rotation speed is i.
The process of correcting the characteristic diagram of the compression part is a process of searching undetermined coefficients a, b and c through the DQN so that errors between a model simulation result and a real engine measurement result are minimized.
And S2.1, defining the structures of two neural networks, a deep Q network and a memory bank network required by the DQN algorithm calculation, inputting and outputting parameters, and initializing the networks. The depth Q network and the memory bank network are all full-connection layer networks with hidden layers larger than 2 layers, and the input end comprises flight conditions, fuel flow, nozzle areas, guide vane angle parameters and current model errors of all measurement state points of the engine; the input of each network in the DQN algorithm is a state vector, denoted as phi(s), the output is the value of all actions in the action set A, denoted as Q A The method comprises the steps of carrying out a first treatment on the surface of the The action set A in the DQN algorithm comprises the range of all a, b and c allowed variations; the DQN algorithm also comprises input iteration times T, a batch gradient descent sample number m and an attenuation factor gamma;
and step S2.2, after the initialization is completed, the DQN algorithm needs to complete the weight updating of the neural network through self-learning, and the model is corrected. Notably, the neural network training data of the DQN algorithm need not be given prior to training, and the DQN algorithm will generate training data during self-learning.
S2.2.1, taking the measurement error E, the fuel flow, the nozzle area, the guide vane angle and the flight condition of each measurement state point of the engine between the digital model and the actual aeroengine in the step S1 as phi (S), inputting the phi (S) into a depth Q network, and obtaining the Q values of all actions in an action set A calculated by the Q network; selecting a corresponding set of actions A among all current outputs according to an E greedy algorithm j So that the set of correction coefficients most likely reduces the simulation error of the mathematical model;
step S2.2.2 based on the action set A j And after the acquired correction coefficient corrects the mathematical model, further calculating a simulation error, namely E ', forming a new state vector phi (s') with the fuel flow, the nozzle area, the guide vane angle and the flight condition of each measurement state point of the engine, and calculating a reward function R, wherein the specific formula is as follows:
the meaning of the formula is that if the correction coefficient selected by the Q network increases the model simulation error, the correction effect of the correction coefficient is poor, a penalty of-100 is obtained, and if the correction coefficient reduces the model simulation error, a penalty of 100e is obtained (E-E') Is a reward for (a).
Step S2.2.3 after the calculated state Φ (S'), the DQN algorithm also needs to determine whether to stop the network learning, i.e. whether to terminate the state S t . Action A of selecting this Q network by outputting phi(s) used in the deep Q network and phi (s') outputted by the Q network j The action gets a prize R j Storing into a memory bank network; and starts a new round of computation using the updated state Φ (s') as input to the deep Q network. After m samples are stored, the Q value y of the Q network selected action in the latest state is calculated as follows:
wherein S is t Represents the termination state, w net Represents weights in a deep Q network, Q (w net A, s) represents the Q value calculated by the deep Q network;
step S2.2.4, based on the following mean square error loss function:
updating weight parameters w in all deep Q networks by neural network gradient back propagation net ；
And S2.3, after the deep Q network training is completed, finding out a proper characteristic correction coefficient according to the simulation error and the given range of the allowed variation of a, b and c.
In order to ensure that the compression part characteristic correction method based on deep reinforcement learning designed by the invention is effective, a specific embodiment is provided below, and digital simulation is carried out on the correction of a characteristic curve of a certain engine.
Firstly, a characteristic curve is multiplied by a constant value to change the characteristic curve so as to achieve the purpose of modifying the model. By the method, the model output parameters are obviously changed after the engine component level model is iterated, and the characteristic curve change has a great influence on the calculation result of the engine model. The current error results are shown in table 1 below:
table 1 partial operating point error after changing the compression element characteristic
The model simulation result before the change of the characteristic diagram is used as a reference value, and the model after the change of the characteristic diagram is used as a decision environment of the DQN. And calculating the error between the model simulation result and the reference value after the characteristic diagram is changed, inputting the DQN, training a neural network in the DQN to obtain a decision network which can correctly select the characteristic correction coefficient, and selecting the characteristic correction coefficient by using the network. The corrected model simulation errors are shown in table 2 below:
TABLE 2 simulation errors after correction of characteristic curves
As can be seen from Table 2, the method for correcting the characteristic curve by applying the deep reinforcement learning can remarkably reduce the simulation error of the engine component level model, and the corrected model simulation error is within 2%.
As can be seen from fig. 2 and 3, when the curve with the rotation speed of 0.95 in the compressor characteristic diagram is subjected to the deviation pulling treatment, a larger deviation exists between the unbiased curve and the deviation pulling curve, and the corrected curve is basically overlapped with the characteristic curve after the deviation pulling after the adjustment of the deep reinforcement learning algorithm.
In summary, the method for correcting the characteristics of the compression part of the aeroengine based on the deep reinforcement learning can lead the neural network to autonomously learn how to correct the characteristic diagram in a simulation environment according to the initial simulation error of the engine model, and select the final correction coefficient. The self-learning capability of deep reinforcement learning is utilized, the problem that the characteristic curve cannot be corrected is effectively solved, and the precision of the characteristic of the component is improved. The method is suitable for correction of various gas turbine engine models.
The foregoing is only a preferred embodiment of the invention, it being noted that: it will be apparent to those skilled in the art that various modifications and adaptations can be made without departing from the principles of the present invention, and such modifications and adaptations are intended to be comprehended within the scope of the invention.
Claims (1)
1. The characteristic correction method for the compression part of the aeroengine based on deep reinforcement learning is characterized by comprising the following steps of:
step S1, establishing an aeroengine component level mathematical model, calculating a measurement error between the mathematical model and an actual aeroengine, and defining a correction coefficient based on the characteristics of a compression component of a nonlinear component model of the engine;
s2, correcting the characteristic diagram of the compression part by adopting a value-based deep reinforcement learning algorithm DQN through autonomous learning, and calculating a characteristic correction coefficient of the compression part; finally, updating the characteristic curve of the aircraft engine component level mathematical model to reduce the error of output data;
the measurement error E between the mathematical model and the actual aeroengine in the step S1 is represented as follows:
wherein k represents the number of test states, m represents the number of parameters obtained by the test, y represents the value of the engine test parameter, y p Calculated values of parameters representing mathematical models, w i Weights representing corresponding parameters;
based on the characteristics of each compression element at different rotational speeds, a correction coefficient is defined as follows:
wherein a, b and c are undetermined coefficients; n is n i Represents the rotation speed of a non-design point, n D Representing the rotation speed of a design point, and a subscript i represents the position of a rotation speed line in a component characteristic diagram; c (C) π,i ，C W,i And C η,i Sequentially representing correction coefficients of the pressure ratio, the flow rate and the efficiency when the rotating speed is i;
correcting the characteristic diagram by adopting a value-based deep reinforcement learning algorithm DQN, namely searching undetermined coefficients a, b and c through the DQN, so that the measurement error between a simulation result of the digital model and an actual aeroengine is the smallest; in particular, the method comprises the steps of,
s2.1, defining a DQN algorithm and input and output parameters based on the structures of a deep Q network and a memory bank network, and initializing the network; the deep Q network and the memory bank network are all full-connection layer networks with hidden layers larger than 2 layers, and input end packetsThe method comprises the steps of measuring flight conditions, fuel flow, nozzle areas, guide vane angle parameters and current model errors of all measurement state points of an engine; the input of each network in the DQN algorithm is a state vector, denoted as phi(s), the output is the value of all actions in the action set A, denoted as Q A The method comprises the steps of carrying out a first treatment on the surface of the The action set A in the DQN algorithm comprises the range of all a, b and c allowed variations; the DQN algorithm is additionally input with iteration times T, the number of samples is reduced by a batch gradient, and the attenuation factor gamma;
s2.2, after defining each input and output item and initializing of the DQN algorithm, finishing the weight updating of the deep Q network through self-learning, and correcting the model;
s2.2.1, taking the measurement error E, the fuel flow, the nozzle area, the guide vane angle and the flight condition of each measurement state point of the engine between the digital model and the actual aeroengine in the step S1 as phi (S), inputting the phi (S) into a depth Q network, and obtaining the Q values of all actions in an action set A calculated by the Q network; selecting a corresponding set of actions A among all current outputs according to an E greedy algorithm j So that the set of correction coefficients most likely reduces the simulation error of the mathematical model;
step S2.2.2 based on the action set A j And after the acquired correction coefficient corrects the mathematical model, further calculating a simulation error, namely E ', forming a new state vector phi (s') with the fuel flow, the nozzle area, the guide vane angle and the flight condition of each measurement state point of the engine, and calculating a reward function R, wherein the specific formula is as follows:
s2.2.3, selecting action A of the current Q network by using phi(s) used in the deep Q network and phi (s') output by the Q network j The action gets a prize R j Storing into a memory bank network; and using the updated state Φ (s') as an input to the deep Q network to begin a new round of computation; after m samples are stored, the Q value y of the Q network selected action in the latest state is calculated as follows:
wherein S is t Represents the termination state, w net Representing weights in a deep Q network, Q (W net A, s) represents the Q value calculated by the deep Q network;
step S2.2.4, based on the following mean square error loss function:
updating weight parameters w in all deep Q networks by neural network gradient back propagation net ；
And S2.3, after the deep Q network training is completed, finding out a proper characteristic correction coefficient according to the simulation error and the given range of the allowed variation of a, b and c.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111351956.2A CN114296343B (en) | 2021-11-16 | 2021-11-16 | Deep reinforcement learning-based aeroengine compression part characteristic correction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111351956.2A CN114296343B (en) | 2021-11-16 | 2021-11-16 | Deep reinforcement learning-based aeroengine compression part characteristic correction method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114296343A CN114296343A (en) | 2022-04-08 |
| CN114296343B true CN114296343B (en) | 2023-10-27 |
Family
ID=80964209
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111351956.2A Active CN114296343B (en) | 2021-11-16 | 2021-11-16 | Deep reinforcement learning-based aeroengine compression part characteristic correction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114296343B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116702380A (en) * | 2023-08-03 | 2023-09-05 | 南京航空航天大学 | Aeroengine performance degradation monitoring and model correction method based on digital twin |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108416086A (en) * | 2018-01-25 | 2018-08-17 | 大连理工大学 | A kind of aero-engine whole envelope model adaptation modification method based on deep learning algorithm |
| CN109800449A (en) * | 2018-12-10 | 2019-05-24 | 南京航空航天大学 | A kind of aero-engine compression member characteristic revision method neural network based |
| CN111914362A (en) * | 2020-07-22 | 2020-11-10 | 中国航发沈阳发动机研究所 | Self-adaptive method for turbofan engine model in research and development stage |
| CN113569319A (en) * | 2021-07-29 | 2021-10-29 | 大连理工大学人工智能大连研究院 | PSO _ DE intelligent algorithm-based design method for adaptive corrector of aircraft engine model |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7058556B2 (en) * | 2001-09-26 | 2006-06-06 | Goodrich Pump & Engine Control Systems, Inc. | Adaptive aero-thermodynamic engine model |
-
2021
- 2021-11-16 CN CN202111351956.2A patent/CN114296343B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108416086A (en) * | 2018-01-25 | 2018-08-17 | 大连理工大学 | A kind of aero-engine whole envelope model adaptation modification method based on deep learning algorithm |
| CN109800449A (en) * | 2018-12-10 | 2019-05-24 | 南京航空航天大学 | A kind of aero-engine compression member characteristic revision method neural network based |
| CN111914362A (en) * | 2020-07-22 | 2020-11-10 | 中国航发沈阳发动机研究所 | Self-adaptive method for turbofan engine model in research and development stage |
| CN113569319A (en) * | 2021-07-29 | 2021-10-29 | 大连理工大学人工智能大连研究院 | PSO _ DE intelligent algorithm-based design method for adaptive corrector of aircraft engine model |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114296343A (en) | 2022-04-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108416086B (en) | Aero-engine model self-adaptive correction method based on deep learning | |
| CN109117951B (en) | BP neural network-based probability load flow online calculation method | |
| WO2020000248A1 (en) | Space reconstruction based method for predicting key performance parameters of transition state acceleration process of aircraft engine | |
| CN111222677A (en) | Wind speed prediction method and system based on long-short term memory time neural network | |
| CN108829928B (en) | Turboshaft engine adaptive component-level simulation model construction method | |
| Demeulenaere et al. | Application of multipoint optimization to the design of turbomachinery blades | |
| CN109162813B (en) | One kind being based on the modified Aeroengine Smart method for controlling number of revolution of iterative learning | |
| CN109800449B (en) | Neural network-based aeroengine compression component characteristic correction method | |
| CN109472062A (en) | A kind of variable cycle engine self-adaptive component grade simulation model construction method | |
| CN114296343B (en) | Deep reinforcement learning-based aeroengine compression part characteristic correction method | |
| Loboda et al. | Gas turbine fault diagnosis using probabilistic neural networks | |
| CN111042928A (en) | Variable cycle engine intelligent control method based on dynamic neural network | |
| CN108804850B (en) | Method for predicting parameters of aircraft engine in transient acceleration process based on spatial reconstruction | |
| CN108594793A (en) | A kind of improved RBF flight control systems fault diagnosis network training method | |
| CN114781435A (en) | Power electronic circuit fault diagnosis method based on improved Harris eagle optimization algorithm optimized variation modal decomposition | |
| CN107016239B (en) | steam turbine valve flow characteristic analysis method | |
| CN114065598B (en) | Fault early warning method for generator of wind turbine generator | |
| CN115146538A (en) | Power system state estimation method based on message passing graph neural network | |
| CN116910485A (en) | Rotary machine fault diagnosis method based on IMPA-VMD energy entropy and improved BiGRU | |
| CN116702380A (en) | Aeroengine performance degradation monitoring and model correction method based on digital twin | |
| CN114154234A (en) | Modeling method, system and storage medium for aircraft engine | |
| CN112364446B (en) | Engine whole-engine performance attenuation prediction method based on EC-RBELM algorithm | |
| CN110991073B (en) | Simulation method, device and equipment for liquid rocket engine | |
| CN116292057A (en) | Control method, device, equipment and medium for variable-speed pumped storage unit | |
| CN113032934B (en) | Intelligent checking method for dynamic parameters of wind power plant based on ME-TD3 algorithm |
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 |