CN116738580A - Speed-thrust matching system and method for high subsonic unmanned aerial vehicle - Google Patents
Speed-thrust matching system and method for high subsonic unmanned aerial vehicle Download PDFInfo
- Publication number
- CN116738580A CN116738580A CN202311008917.1A CN202311008917A CN116738580A CN 116738580 A CN116738580 A CN 116738580A CN 202311008917 A CN202311008917 A CN 202311008917A CN 116738580 A CN116738580 A CN 116738580A
- Authority
- CN
- China
- Prior art keywords
- speed
- time sequence
- training
- aircraft
- input vector
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 42
- 239000013598 vector Substances 0.000 claims abstract description 156
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 52
- 238000012300 Sequence Analysis Methods 0.000 claims abstract description 12
- 238000012549 training Methods 0.000 claims description 76
- 239000011159 matrix material Substances 0.000 claims description 22
- 238000009826 distribution Methods 0.000 claims description 18
- 238000012545 processing Methods 0.000 claims description 18
- 230000014509 gene expression Effects 0.000 claims description 13
- 238000005457 optimization Methods 0.000 claims description 13
- 238000000605 extraction Methods 0.000 claims description 11
- 238000012546 transfer Methods 0.000 claims description 10
- 238000004458 analytical method Methods 0.000 claims description 3
- 239000002356 single layer Substances 0.000 claims description 2
- 238000012731 temporal analysis Methods 0.000 claims 1
- 238000000700 time series analysis Methods 0.000 claims 1
- 238000013135 deep learning Methods 0.000 abstract description 7
- 239000010410 layer Substances 0.000 description 16
- 238000010586 diagram Methods 0.000 description 10
- 230000008859 change Effects 0.000 description 7
- 230000006870 function Effects 0.000 description 7
- 230000035939 shock Effects 0.000 description 7
- 230000008569 process Effects 0.000 description 6
- 230000015572 biosynthetic process Effects 0.000 description 4
- 239000000284 extract Substances 0.000 description 4
- 238000011176 pooling Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000013507 mapping Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 239000012530 fluid Substances 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000000844 transformation Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/15—Vehicle, aircraft or watercraft design
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Computer Hardware Design (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Aviation & Aerospace Engineering (AREA)
- Automation & Control Theory (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Traffic Control Systems (AREA)
- Feedback Control In General (AREA)
Abstract
A speed-thrust matching system of a high subsonic unmanned aerial vehicle and a method thereof are disclosed. Firstly, acquiring flight speed values of a plurality of preset time points in a preset time period, then, carrying out time sequence analysis on the flight speed values of the preset time points to obtain a multi-scale aircraft speed time sequence feature vector, and then, determining a thrust output value of an aircraft engine based on the multi-scale aircraft speed time sequence feature vector. In this way, the convolutional neural network model based on deep learning can be utilized to extract the characteristics from the time sequence data of the flight speed of the aircraft, and generate the thrust output value of the proper aircraft engine, thereby completing the speed-thrust matching.
Description
Technical Field
The present disclosure relates to the field of unmanned aerial vehicles, and more particularly, to a speed-thrust matching system of a high subsonic unmanned aerial vehicle and a method thereof.
Background
The high subsonic unmanned aerial vehicle is an unmanned aerial vehicle capable of flying in a speed range below hypersonic speed, and has the advantages of high maneuverability, gao Yinshen performance, high outburst prevention capability and the like.
High subsonic unmanned aerial vehicles need to have sufficient thrust to overcome drag during flight. The aircraft has different flight resistances at different flight speeds and the required thrust is different. In order to realize effective control and optimization of flight performance of a high subsonic unmanned aerial vehicle, a speed-thrust matching system capable of dynamically adjusting a thrust output value of an aircraft engine according to a change condition of the flight speed needs to be designed.
However, since the time sequence data of the flight speed of the high subsonic unmanned aerial vehicle has complex nonlinear relation and non-stationarity characteristics, the conventional speed-thrust matching method is difficult to accurately perform speed-thrust matching. Thus, an optimized speed-thrust matching scheme for high subsonic unmanned aerial vehicles is desired.
Disclosure of Invention
In view of this, the present disclosure proposes a speed-thrust matching system of a high subsonic unmanned aerial vehicle and a method thereof, which can perform feature extraction from flight speed time series data of an aircraft using a convolutional neural network model based on deep learning, and generate a thrust output value of an appropriate aircraft engine therefrom, thereby completing speed-thrust matching.
According to an aspect of the present disclosure, there is provided a speed-thrust matching method of a high subsonic unmanned aerial vehicle, including: acquiring flying speed values of a plurality of preset time points in a preset time period; performing time sequence analysis on the flying speed values of the plurality of preset time points to obtain a multi-scale airplane speed time sequence feature vector; and determining a thrust output value of an aircraft engine based on the multi-scale aircraft speed timing feature vector.
According to another aspect of the present disclosure, there is provided a speed-thrust matching system of a high subsonic unmanned aerial vehicle, comprising: the flight speed value acquisition module is used for acquiring flight speed values of a plurality of preset time points in a preset time period; the time sequence analysis module is used for performing time sequence analysis on the flying speed values of the plurality of preset time points to obtain a multi-scale aircraft speed time sequence feature vector; and the thrust output value control module is used for determining the thrust output value of the aircraft engine based on the multi-scale aircraft speed time sequence feature vector.
According to the embodiment of the disclosure, firstly, the flying speed values of a plurality of preset time points in a preset time period are acquired, then, time sequence analysis is carried out on the flying speed values of the preset time points to obtain a multi-scale airplane speed time sequence characteristic vector, and then, the thrust output value of an airplane engine is determined based on the multi-scale airplane speed time sequence characteristic vector. In this way, the convolutional neural network model based on deep learning can be utilized to extract the characteristics from the time sequence data of the flight speed of the aircraft, and generate the thrust output value of the proper aircraft engine, thereby completing the speed-thrust matching.
Other features and aspects of the present disclosure will become apparent from the following detailed description of exemplary embodiments, which proceeds with reference to the accompanying drawings.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate exemplary embodiments, features and aspects of the present disclosure and together with the description, serve to explain the principles of the disclosure.
Fig. 1 shows a flow chart of a speed-thrust matching method of a high subsonic unmanned aerial vehicle according to an embodiment of the present disclosure.
Fig. 2 shows an architectural diagram of a speed-thrust matching method of a high subsonic unmanned aerial vehicle according to an embodiment of the present disclosure.
Fig. 3 shows a flowchart of sub-step S120 of a speed-thrust matching method of a high subsonic unmanned aerial vehicle according to an embodiment of the present disclosure.
Fig. 4 shows a flowchart of sub-step S121 of a speed-thrust matching method of a high subsonic unmanned aerial vehicle according to an embodiment of the present disclosure.
Fig. 5 shows a flowchart of sub-step S122 of a speed-thrust matching method of a high subsonic unmanned aerial vehicle according to an embodiment of the present disclosure.
Fig. 6 shows a flowchart of training steps further included in a speed-thrust matching method of a high subsonic unmanned aerial vehicle according to an embodiment of the present disclosure.
Fig. 7 shows a block diagram of a speed-thrust matching system of a high subsonic unmanned aerial vehicle according to an embodiment of the present disclosure.
Fig. 8 illustrates an application scenario diagram of a speed-thrust matching method of a high subsonic unmanned aerial vehicle according to an embodiment of the present disclosure.
Detailed Description
The following description of the embodiments of the present disclosure will be made clearly and fully with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some, but not all embodiments of the disclosure. All other embodiments, which can be made by one of ordinary skill in the art without undue burden based on the embodiments of the present disclosure, are also within the scope of the present disclosure.
As used in this disclosure and in the claims, the terms "a," "an," "the," and/or "the" are not specific to a singular, but may include a plurality, unless the context clearly dictates otherwise. In general, the terms "comprises" and "comprising" merely indicate that the steps and elements are explicitly identified, and they do not constitute an exclusive list, as other steps or elements may be included in a method or apparatus.
Various exemplary embodiments, features and aspects of the disclosure will be described in detail below with reference to the drawings. In the drawings, like reference numbers indicate identical or functionally similar elements. Although various aspects of the embodiments are illustrated in the accompanying drawings, the drawings are not necessarily drawn to scale unless specifically indicated.
In addition, numerous specific details are set forth in the following detailed description in order to provide a better understanding of the present disclosure. It will be understood by those skilled in the art that the present disclosure may be practiced without some of these specific details. In some instances, methods, means, elements, and circuits well known to those skilled in the art have not been described in detail in order not to obscure the present disclosure.
It should be appreciated that conventional speed-thrust matching methods typically employ a linear model or a simple nonlinear model for speed-thrust matching. These models are typically based on a simplified physical model or empirical formula and assume that there is a linear or simplified nonlinear relationship between the flight speed and thrust.
However, the flight speed timing data of a high subsonic unmanned aerial vehicle has a complex nonlinear relationship that is difficult to accurately capture by a conventional linear model or a simplified nonlinear model. This is because the physical phenomena involved in high subsonic flight are very complex, such as shock wave formation, turbulence drag, aerodynamic effects, etc., resulting in a highly nonlinear relationship between the flight speed and the thrust. Here, it is worth mentioning that shock wave formation and turbulence resistance are two important physical phenomena involved in high subsonic flight. The shock wave formation means that when the high subsonic aircraft exceeds the sound velocity, shock waves are formed around the high subsonic aircraft, the shock waves are pressure waves generated in the air when the moving speed of the aircraft exceeds the sound velocity, the shock waves can cause severe compression and acceleration of airflow, so that aerodynamic phenomena around the aircraft are changed, the formation and the propagation of the shock waves have important influence on the aerodynamic performance of the aircraft, and the increase of resistance, the reduction of lift force and the like can be caused; turbulence is a fluid motion state in which the velocity and direction of the fluid are irregularly varied, and in high subsonic flight, turbulence phenomenon is formed due to the influence of drastic variation of the air flow around the aircraft, and turbulence causes energy loss and resistance increase of the air flow, thereby affecting the relationship between the velocity and thrust of the aircraft. The complexity of these physical phenomena results in a highly nonlinear relationship between high subsonic flight speed and thrust. Thus, conventional linear models or simplified nonlinear models tend to have difficulty accurately capturing this complexity, requiring the use of more complex models and methods for modeling and analysis.
In addition, the time sequence data of the flying speed of the high subsonic unmanned aerial vehicle has the characteristic of non-stationarity, namely the flying speed can change along with the change of time. However, conventional linear models or simplified nonlinear models are generally not adaptable to changes in actual flight speed, resulting in inaccurate matching results.
The convolutional neural network (Convolutional Neural Network, CNN) model based on deep learning can capture and model complex nonlinear relations, and has great advantages in processing speed-thrust matching. In particular, convolutional neural network models automatically learn and extract implicit features in the data through a multi-layer convolutional structure without requiring manual design of the features. For the speed-thrust matching problem of the high subsonic unmanned plane, the convolutional neural network model can learn the association of the flight speed and the flight from the time sequence data of the flight speed of the plane, so that accurate matching is realized. In summary, the convolutional neural network model based on deep learning can effectively capture and model complex nonlinear relations which are difficult to capture by the traditional method, so that the convolutional neural network model has great potential and advantages in the speed-thrust matching problem of the high subsonic unmanned aerial vehicle.
Based on the technical problems, the technical concept of the present disclosure is to perform feature extraction from time series data of the flight speed of an aircraft by using a convolutional neural network model based on deep learning, and generate a suitable thrust output value of the aircraft engine, thereby completing speed-thrust matching.
Fig. 1 shows a flow chart of a speed-thrust matching method of a high subsonic unmanned aerial vehicle according to an embodiment of the present disclosure. Fig. 2 shows an architectural diagram of a speed-thrust matching method of a high subsonic unmanned aerial vehicle according to an embodiment of the present disclosure. As shown in fig. 1 and 2, a speed-thrust matching method of a high subsonic unmanned aerial vehicle according to an embodiment of the present disclosure includes the steps of: s110, acquiring flight speed values of a plurality of preset time points in a preset time period; s120, carrying out time sequence analysis on the flying speed values of the plurality of preset time points to obtain a multi-scale airplane speed time sequence feature vector; and S130, determining a thrust output value of the aircraft engine based on the multi-scale aircraft speed time sequence feature vector.
Specifically, in the technical scheme of the present disclosure, first, flight speed values at a plurality of predetermined time points within a predetermined period of time are acquired. And then, carrying out data structuring processing and information expression multidimensional processing on the flying speed values at a plurality of preset time points to obtain an airplane speed multidimensional time sequence input vector.
In a specific example of the disclosure, the implementation process of performing data structuring processing and information expression multidimensional processing on the flight speed values of the plurality of predetermined time points to obtain the multi-dimensional time sequence input vector of the aircraft speed is as follows: firstly, arranging the flying speed values of the plurality of preset time points into flying speed time sequence input vectors according to a time dimension; then, calculating the difference value between the flight speed values of every two adjacent preset time points in the flight speed time sequence input vector to obtain a flight speed neighborhood fluctuation time sequence input vector; calculating the difference between the flying speed value of other positions in the flying speed time sequence input vector and the flying speed value of the first position to obtain a flying speed cross-domain fluctuation time sequence input vector; and then, merging the flying speed time sequence input vector, the flying speed neighborhood fluctuation time sequence input vector and the flying speed cross-domain fluctuation time sequence input vector to obtain the multi-dimensional time sequence input vector of the flying speed.
The method comprises the steps of calculating the difference value between the flying speed values of every two adjacent preset time points in the flying speed time sequence input vector and calculating the difference value between the flying speed values of other positions in the flying speed time sequence input vector and the flying speed value of the first position, wherein the difference value is obtained by converting absolute change data into relative change data, and fluctuation of the speed values can be reflected better. It is worth mentioning that, calculating the difference between the flying speed value of other positions in the flying speed time sequence input vector and the flying speed value of the first position can represent the fluctuation condition of other various time points relative to the initial speed, and can reflect the non-stationarity of the aircraft speed.
And then, the multi-dimensional time sequence input vector of the aircraft speed passes through a multi-scale aircraft speed time sequence feature extractor comprising a first convolutional neural network model and a second convolutional neural network model to obtain a multi-scale aircraft speed time sequence feature vector.
Accordingly, as shown in fig. 3, performing time sequence analysis on the flying speed values at the plurality of preset time points to obtain a multi-scale aircraft speed time sequence feature vector, which includes: s121, carrying out data structuring processing and information expression multidimensional processing on the flying speed values of the plurality of preset time points to obtain an airplane speed multidimensional time sequence input vector; and S122, carrying out feature extraction on the multi-dimensional time sequence input vector of the airplane speed to obtain the multi-scale time sequence feature vector of the airplane speed.
More specifically, in step S121, as shown in fig. 4, the data structuring process and the information expression multidimensional process are performed on the flying speed values at the plurality of predetermined time points to obtain an aircraft speed multidimensional time sequence input vector, including: s1211, arranging the flying speed values of the plurality of preset time points into flying speed time sequence input vectors according to a time dimension; s1212, calculating the difference value between the flight speed values of every two adjacent preset time points in the flight speed time sequence input vector to obtain a flight speed neighborhood fluctuation time sequence input vector; s1213, calculating the difference between the flying speed value of other positions in the flying speed time sequence input vector and the flying speed value of the first position to obtain a flying speed cross-domain fluctuation time sequence input vector; and S1214, merging the flying speed time sequence input vector, the flying speed neighborhood fluctuation time sequence input vector and the flying speed cross-domain fluctuation time sequence input vector to obtain the multi-dimensional time sequence input vector of the flying speed.
More specifically, in step S122, feature extraction is performed on the multi-dimensional time sequence input vector of the aircraft speed to obtain the multi-scale time sequence feature vector of the aircraft speed, including: and the multi-dimensional time sequence input vector of the aircraft speed passes through a multi-scale aircraft speed time sequence feature extractor comprising a first convolutional neural network model and a second convolutional neural network model to obtain a multi-scale aircraft speed time sequence feature vector. More specifically, in step S122, as shown in fig. 5, passing the aircraft speed multi-dimensional time series input vector through a multi-scale aircraft speed time series feature extractor including a first convolutional neural network model and a second convolutional neural network model to obtain a multi-scale aircraft speed time series feature vector, including: s1221, enabling the aircraft speed multidimensional time sequence input vector to pass through the first convolutional neural network model to obtain a first scale aircraft speed time sequence feature vector; s1222, passing the aircraft speed multi-dimensional time sequence input vector through the second convolutional neural network model to obtain a second scale aircraft speed time sequence feature vector; and S1223, fusing the first-scale aircraft speed time sequence feature vector and the second-scale aircraft speed time sequence feature vector to obtain the multi-scale aircraft speed time sequence feature vector.
It should be noted that the convolutional neural network (Convolutional Neural Network, CNN) is a deep learning model, which is widely applied to image processing and computer vision tasks, and the core idea of the convolutional neural network is to extract the characteristics of input data through a convolutional layer and a pooling layer, and classify or regress through a fully connected layer. The convolution layer convolves the input data using a series of learnable filters (also referred to as convolution kernels) to capture local features of the input data. The pooling layer is used to reduce the size of the feature map and preserve the most important features. The full connection layer converts the aggregated feature map into a final output result. The filter of the convolution layer of the convolution neural network only focuses on the local area of the input data, and local characteristics are extracted in a weight sharing mode, so that the convolution neural network can effectively process translation, rotation, scaling and other transformations in the image. The filters in the convolution layer share weights across the entire input image, reducing the number of parameters that need to be learned, which not only reduces the complexity of the model, but also improves the generalization ability of the model. Convolutional neural networks CNNs are able to progressively extract high-level abstract features of input data by stacking multiple convolutional layers and pooling layers. Low-level convolution layers may capture low-level features such as edges and textures, while high-level convolution layers may capture more abstract and semantic features.
In step S122, the first convolutional neural network model and the second convolutional neural network model are used as a multi-scale aircraft speed timing feature extractor for extracting features of different scales from the aircraft speed multi-dimensional timing input vector. Specifically, the first convolutional neural network model takes an aircraft speed multi-dimensional time sequence input vector as input, and extracts an aircraft speed time sequence feature vector of a first scale through operations of a convolutional layer, a pooling layer and the like. This model may be concerned with speed changes over a short time frame, capturing some of the rapidly changing features. Similarly, the second convolutional neural network model also takes as input an aircraft speed multi-dimensional time sequence input vector, but may focus on speed changes over a longer time range to extract a second scale of aircraft speed time sequence feature vectors. Finally, the multi-scale aircraft speed time sequence feature vector is obtained by fusing the aircraft speed time sequence feature vectors of the first scale and the second scale. The purpose of this is to comprehensively consider speed change information on different time scales to more fully describe the speed characteristics of the aircraft. In summary, the first convolutional neural network model and the second convolutional neural network model extract features of the aircraft speed timing on multiple scales so that the models can better understand and represent the complexity of the aircraft speed, thereby improving the performance of subsequent tasks (such as classification or regression).
Further, the multi-scale aircraft speed timing feature vector is passed through a decoder to obtain a decoded value representing a recommended thrust output value of the aircraft engine. Accordingly, determining a thrust output value of an aircraft engine based on the multi-scale aircraft speed timing feature vector comprises: the multi-scale aircraft speed time sequence feature vector is passed through a decoder to obtain a decoded value, wherein the decoded value is used for representing a recommended thrust output value of an aircraft engine. It is worth mentioning that the decoder is a neural network model for converting the multi-scale aircraft speed timing feature vector into a decoded value representing the recommended thrust output value of the aircraft engine. The decoder is used for converting the abstract feature vector into an actual thrust output value, and the decoder decodes the input feature vector into a corresponding thrust output value by learning the mapping relation between the feature vector and the thrust output value. The decoder typically consists of a number of fully connected layers that linearly and non-linearly transform the input through a weight matrix and an activation function, resulting in decoded values. The weight matrices are learned during training to maximize accuracy in predicting thrust output values. The decoder can be used for converting the multi-scale airplane speed time sequence feature vector into a recommended airplane engine thrust output value, so that a decision basis is provided for airplane thrust control.
Further, in the technical scheme of the present disclosure, the speed-thrust matching method of the high subsonic unmanned aerial vehicle further includes a training step: the multi-scale aircraft speed timing feature extractor and the decoder comprising a first convolutional neural network model and a second convolutional neural network model are trained. It should be appreciated that in the technical solution of the present disclosure, the purpose of the training step is to enable the multi-scale aircraft speed timing feature extractor and decoder to accurately learn and capture the relationship between the aircraft speed and the engine thrust by training them, thereby achieving accurate prediction of the thrust output value of the aircraft engine. Specifically, the training step works as follows: 1. feature extractor training: by training a multi-scale aircraft speed time sequence feature extractor, the model can learn the capability of extracting useful features from the original aircraft speed time sequence data, and the features can better represent the complexity and change of the aircraft speed and provide more accurate input for the subsequent thrust output value prediction; 2. decoder training: the model can learn the capability of mapping the multi-scale aircraft speed time sequence feature vector into the thrust output value by training the decoder, and the decoder can accurately predict the thrust output value in the thrust control process by learning the mapping relation between the feature vector and the thrust output value, so that the accurate control of the aircraft engine is realized. Through the training step, the model can gradually optimize and adjust own parameters, and the understanding and expression capability of the complex relationship between the airplane speed and the engine thrust are improved. In practical application, the model can recommend proper engine thrust output value according to the speed time sequence data of the airplane more accurately so as to realize speed-thrust matching of the high subsonic unmanned plane.
In one specific example, as shown in fig. 6, the training step includes: s210, acquiring training data, wherein the training data comprises training flight speed values at a plurality of preset time points in a preset time period and a true value of a thrust output value of an aircraft engine; s220, arranging the training flying speed values of the plurality of preset time points into training flying speed time sequence input vectors according to a time dimension; s230, calculating the difference value between training flight speed values of every two adjacent preset time points in the training flight speed time sequence input vector to obtain a training flight speed neighborhood fluctuation time sequence input vector; s240, calculating the difference between the training flight speed value of other positions in the training flight speed time sequence input vector and the training flight speed value of the first position to obtain a training flight speed cross-domain fluctuation time sequence input vector; s250, fusing the training flying speed time sequence input vector, the training flying speed neighborhood fluctuation time sequence input vector and the training flying speed cross-domain fluctuation time sequence input vector to obtain a training airplane speed multi-dimensional time sequence input vector; s260, passing the training aircraft speed multi-dimensional time sequence input vector through the multi-scale aircraft speed time sequence feature extractor comprising the first convolutional neural network model and the second convolutional neural network model to obtain a training multi-scale aircraft speed time sequence feature vector; s270, passing the training multi-scale aircraft speed time sequence feature vector through a decoder to obtain a decoding loss function value; and S280, training the multi-scale aircraft speed time sequence feature extractor comprising the first convolutional neural network model and the second convolutional neural network model and the decoder by using the decoding loss function value, wherein in each round of iteration of the training, cross-domain attention transfer optimization iteration of feature distribution is performed on a weight matrix of the decoder.
In the technical scheme of the disclosure, after the training flight speed time sequence input vector, the training flight speed neighborhood fluctuation time sequence input vector and the training flight speed cross-domain fluctuation time sequence input vector are fused, the training aircraft speed multi-dimensional time sequence input vector comprises diversified flight speed time sequence distribution characteristics, so that the training multi-scale aircraft speed time sequence characteristic vector has diversified characteristic distribution degree expression through a multi-scale aircraft speed time sequence characteristic extractor comprising a first convolutional neural network model and a second convolutional neural network model.
Thus, when the training multi-scale aircraft speed timing feature vector is decoded by a decoder, it may have better distribution transferability than the speed variation feature expression, and vice versa, when the weight matrix of the decoder is adapted relative to the absolute speed feature expression, taking into account the distribution transferability differences of the diversified feature distribution expressions in the decoded domain transfer process. Therefore, the weight matrix of the decoder needs to adaptively optimize the training multi-scale aircraft speed time sequence feature vector so as to improve the training effect of decoding training of the training multi-scale aircraft speed time sequence feature vector through the decoder, namely, improve the decoding speed and the accuracy of the obtained decoding result. Thus, the applicant of the present disclosure, in each iteration of the weighting matrix of the decoder, for said weighting matrixAnd performing cross-domain attention transfer optimization of feature distribution.
Accordingly, in one specific example, in each iteration of the training, performing a cross-domain attention transfer optimization iteration of feature distribution on a weight matrix of the decoder, including: the weight matrix of the decoder is optimized by the following formulaCross-domain attention transfer optimization iteration of feature distribution; wherein, the optimization formula is:wherein (1)>Is a weight matrix of the decoder, +.>Is of the scale +.>，/>To->Is the weight matrix->Is->Individual row vectors>Representing the two norms of the feature vector, +.>Representing a transpose operation->Is to the weight matrix +.>The sum value of each row vector of (a) is arranged to obtain a row vector, and +.>And->All represent a single layer convolution operation, ">Representing matrix multiplication +.>Representing the weight matrix of the decoder after iteration.
Here, the feature distribution-based cross-domain attention-diversion optimization is based on a weight matrix of the decoder for different representations of the feature distribution of the training multi-scale aircraft speed timing feature vector present in a feature space domain and a decoding target domainCross-domain diversity feature representation of the training multi-scale aircraft speed timing feature vector relative to the training multi-scale aircraft speed timing feature vector to be decoded by +.>Is focused by convolution operations to enhance the transferability of cross-domain gaps of good transferred feature distributions in a diversified feature distribution while suppressing negative transfer (negative transfer) of bad transferred feature distributions to be based on the weight matrix ∈ ->Realizing a weight matrix by itself relative to the distribution structure of the training multi-scale aircraft speed time sequence feature vector to be decoded>The unsupervised domain transfer adaptive optimization of the system is achieved, so that the training effect of decoding training of the training multi-scale airplane speed time sequence feature vector through a decoder is improved.
In summary, according to the speed-thrust matching method of the high subsonic unmanned aerial vehicle according to the embodiment of the disclosure, a convolutional neural network model based on deep learning can be utilized to perform feature extraction from time series data of the flight speed of the aircraft, and a thrust output value of a proper aircraft engine is generated according to the feature extraction, so that speed-thrust matching is completed.
Fig. 7 shows a block diagram of a speed-thrust matching system 100 of a high subsonic unmanned aerial vehicle according to an embodiment of the present disclosure. As shown in fig. 7, a speed-thrust matching system 100 of a high subsonic unmanned aerial vehicle according to an embodiment of the present disclosure includes: a flying speed value obtaining module 110, configured to obtain flying speed values at a plurality of predetermined time points within a predetermined period; the time sequence analysis module 120 is configured to perform time sequence analysis on the flying speed values at the multiple predetermined time points to obtain a multi-scale aircraft speed time sequence feature vector; and a thrust output value control module 130 for determining a thrust output value of the aircraft engine based on the multi-scale aircraft speed timing feature vector.
In one possible implementation, the timing analysis module 120 includes: the data processing unit is used for carrying out data structuring processing and information expression multidimensional processing on the flying speed values of the plurality of preset time points so as to obtain an airplane speed multidimensional time sequence input vector; and the characteristic extraction unit is used for carrying out characteristic extraction on the multi-dimensional time sequence input vector of the airplane speed so as to obtain the multi-scale time sequence characteristic vector of the airplane speed.
Here, it will be understood by those skilled in the art that the specific functions and operations of the respective units and modules in the above-described speed-thrust matching system 100 of the high subsonic unmanned aerial vehicle have been described in detail in the above description of the speed-thrust matching method of the high subsonic unmanned aerial vehicle with reference to fig. 1 to 6, and thus, repetitive descriptions thereof will be omitted.
As described above, the speed-thrust matching system 100 of the high subsonic unmanned aerial vehicle according to the embodiment of the present disclosure may be implemented in various wireless terminals, such as a server or the like having a speed-thrust matching algorithm of the high subsonic unmanned aerial vehicle. In one possible implementation, the speed-thrust matching system 100 of a high subsonic unmanned aerial vehicle according to embodiments of the present disclosure may be integrated into a wireless terminal as one software module and/or hardware module. For example, the speed-thrust matching system 100 of the high subsonic unmanned aerial vehicle may be a software module in the operating system of the wireless terminal, or may be an application developed for the wireless terminal; of course, the speed-thrust matching system 100 of the high subsonic unmanned aerial vehicle may also be one of a number of hardware modules of the wireless terminal.
Alternatively, in another example, the speed-thrust matching system 100 of the high subsonic unmanned aerial vehicle and the wireless terminal may be separate devices, and the speed-thrust matching system 100 of the high subsonic unmanned aerial vehicle may be connected to the wireless terminal through a wired and/or wireless network and transmit the interactive information in a agreed data format.
Fig. 8 illustrates an application scenario diagram of a speed-thrust matching method of a high subsonic unmanned aerial vehicle according to an embodiment of the present disclosure. As shown in fig. 8, in this application scenario, first, flight speed values at a plurality of predetermined time points within a predetermined period of time (for example, D illustrated in fig. 8) are acquired, and then, the flight speed values at the plurality of predetermined time points are input to a server (for example, S illustrated in fig. 8) in which a speed-thrust matching algorithm of a high-subsonic unmanned aerial vehicle is deployed, wherein the server is capable of processing the flight speed values at the plurality of predetermined time points using the speed-thrust matching algorithm of the high-subsonic unmanned aerial vehicle to obtain a decoded value representing a recommended thrust output value of an aircraft engine.
The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more executable instructions for implementing the specified logical function(s). In some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
The foregoing description of the embodiments of the present disclosure has been presented for purposes of illustration and description, and is not intended to be exhaustive or limited to the embodiments disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the various embodiments described. The terminology used herein was chosen in order to best explain the principles of the embodiments, the practical application, or the improvement of technology in the marketplace, or to enable others of ordinary skill in the art to understand the embodiments disclosed herein.
Claims (10)
1. The speed-thrust matching method of the high subsonic unmanned aerial vehicle is characterized by comprising the following steps of: acquiring flying speed values of a plurality of preset time points in a preset time period; performing time sequence analysis on the flying speed values of the plurality of preset time points to obtain a multi-scale airplane speed time sequence feature vector; and determining a thrust output value of an aircraft engine based on the multi-scale aircraft speed timing feature vector.
2. The method of speed-thrust matching for a high subsonic unmanned aircraft according to claim 1, wherein performing a time series analysis of the flight speed values at the plurality of predetermined time points to obtain a multi-scale aircraft speed time series feature vector comprises: carrying out data structuring processing and information expression multidimensional processing on the flying speed values of the plurality of preset time points to obtain an airplane speed multidimensional time sequence input vector; and extracting features of the multi-dimensional time sequence input vector of the aircraft speed to obtain the multi-scale time sequence feature vector of the aircraft speed.
3. The method for matching speed-thrust of a high subsonic unmanned aerial vehicle according to claim 2, wherein the data structuring and information expression multidimensional processing are performed on the flying speed values of the plurality of predetermined time points to obtain a multi-dimensional time sequence input vector of the aircraft speed, comprising: arranging the flying speed values of the plurality of preset time points into flying speed time sequence input vectors according to a time dimension; calculating the difference value between the flight speed values of every two adjacent preset time points in the flight speed time sequence input vector to obtain a flight speed neighborhood fluctuation time sequence input vector; calculating the difference between the flying speed value of other positions in the flying speed time sequence input vector and the flying speed value of the first position to obtain a flying speed cross-domain fluctuation time sequence input vector; and fusing the flying speed time sequence input vector, the flying speed neighborhood fluctuation time sequence input vector and the flying speed cross-domain fluctuation time sequence input vector to obtain the airplane speed multi-dimensional time sequence input vector.
4. A speed-thrust matching method of a high subsonic unmanned aerial vehicle according to claim 3, wherein feature extraction of the aircraft speed multi-dimensional time series input vector to obtain the multi-scale aircraft speed time series feature vector comprises: and the multi-dimensional time sequence input vector of the aircraft speed passes through a multi-scale aircraft speed time sequence feature extractor comprising a first convolutional neural network model and a second convolutional neural network model to obtain a multi-scale aircraft speed time sequence feature vector.
5. The method of speed-thrust matching for a high subsonic unmanned aircraft of claim 4, wherein passing the aircraft speed multi-dimensional time series input vector through a multi-scale aircraft speed time series feature extractor comprising a first convolutional neural network model and a second convolutional neural network model to obtain a multi-scale aircraft speed time series feature vector, comprising: passing the aircraft speed multidimensional time sequence input vector through the first convolutional neural network model to obtain a first scale aircraft speed time sequence feature vector; passing the aircraft speed multidimensional time sequence input vector through the second convolutional neural network model to obtain a second scale aircraft speed time sequence feature vector; and fusing the first-scale aircraft speed timing feature vector and the second-scale aircraft speed timing feature vector to obtain the multi-scale aircraft speed timing feature vector.
6. The method of speed-thrust matching for a high subsonic unmanned aerial vehicle of claim 5, wherein determining a thrust output value of an aircraft engine based on the multi-scale aircraft speed timing feature vector comprises: the multi-scale aircraft speed time sequence feature vector is passed through a decoder to obtain a decoded value, wherein the decoded value is used for representing a recommended thrust output value of an aircraft engine.
7. The speed-thrust matching method of a high subsonic unmanned aerial vehicle of claim 6, further comprising the training step of: training the multi-scale aircraft speed timing feature extractor and the decoder comprising a first convolutional neural network model and a second convolutional neural network model; wherein the training step comprises: acquiring training data, wherein the training data comprises training flight speed values at a plurality of preset time points in a preset time period and true values of thrust output values of an aircraft engine; arranging the training flying speed values of the plurality of preset time points into training flying speed time sequence input vectors according to the time dimension; calculating the difference value between training flight speed values of every two adjacent preset time points in the training flight speed time sequence input vector to obtain a training flight speed neighborhood fluctuation time sequence input vector; calculating the difference between the training flight speed value of other positions in the training flight speed time sequence input vector and the training flight speed value of the first position to obtain a training flight speed cross-domain fluctuation time sequence input vector; fusing the training flying speed time sequence input vector, the training flying speed neighborhood fluctuation time sequence input vector and the training flying speed cross-domain fluctuation time sequence input vector to obtain a training airplane speed multidimensional time sequence input vector; passing the training aircraft speed multi-dimensional time sequence input vector through the multi-scale aircraft speed time sequence feature extractor comprising a first convolutional neural network model and a second convolutional neural network model to obtain a training multi-scale aircraft speed time sequence feature vector; passing the training multi-scale aircraft speed time sequence feature vector through a decoder to obtain a decoding loss function value; and training the multi-scale aircraft speed timing feature extractor comprising a first convolutional neural network model and a second convolutional neural network model and the decoder with the decoding loss function values, wherein in each iteration of the training, a cross-domain attention transfer optimization iteration of feature distribution is performed on a weight matrix of the decoder.
8. The method of speed-thrust matching for a high subsonic unmanned aerial vehicle of claim 7, wherein in each iteration of the training, performing a cross-domain distraction optimization iteration of feature distribution on the weight matrix of the decoder comprises: performing cross-domain attention transfer optimization iteration of feature distribution on the weight matrix of the decoder according to the following optimization formula; wherein, the optimization formula is:wherein (1)>Is a weight matrix of the decoder, +.>Is of the scale +.>，/>To->Is the weight matrix->A kind of electronic deviceIndividual row vectors>Representing the two norms of the feature vector, +.>Representing a transpose operation->Is to the weight matrix +.>The sum value of each row vector of (a) is arranged to obtain a row vector, and +.>And->All represent a single layer convolution operation, ">Representing matrix multiplication +.>Representing the weight matrix of the decoder after iteration.
9. A speed-thrust matching system for a high subsonic unmanned aerial vehicle, comprising: the flight speed value acquisition module is used for acquiring flight speed values of a plurality of preset time points in a preset time period; the time sequence analysis module is used for performing time sequence analysis on the flying speed values of the plurality of preset time points to obtain a multi-scale aircraft speed time sequence feature vector; and the thrust output value control module is used for determining the thrust output value of the aircraft engine based on the multi-scale aircraft speed time sequence feature vector.
10. The speed-thrust matching system of a high subsonic unmanned aerial vehicle of claim 9, wherein the timing analysis module comprises: the data processing unit is used for carrying out data structuring processing and information expression multidimensional processing on the flying speed values of the plurality of preset time points so as to obtain an airplane speed multidimensional time sequence input vector; and the characteristic extraction unit is used for carrying out characteristic extraction on the multi-dimensional time sequence input vector of the airplane speed so as to obtain the multi-scale time sequence characteristic vector of the airplane speed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311008917.1A CN116738580B (en) | 2023-08-11 | 2023-08-11 | Speed-thrust matching system and method for high subsonic unmanned aerial vehicle |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311008917.1A CN116738580B (en) | 2023-08-11 | 2023-08-11 | Speed-thrust matching system and method for high subsonic unmanned aerial vehicle |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116738580A true CN116738580A (en) | 2023-09-12 |
| CN116738580B CN116738580B (en) | 2023-11-14 |
Family
ID=87909941
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311008917.1A Active CN116738580B (en) | 2023-08-11 | 2023-08-11 | Speed-thrust matching system and method for high subsonic unmanned aerial vehicle |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116738580B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116907282A (en) * | 2023-09-14 | 2023-10-20 | 杭州牧星科技有限公司 | Unmanned target aircraft ultra-low altitude flight control method based on artificial intelligence algorithm |
| CN117382435A (en) * | 2023-10-17 | 2024-01-12 | 浙江加力仓储设备股份有限公司 | Vehicle speed control method and system based on dip angle monitoring |
| CN117555355A (en) * | 2023-10-07 | 2024-02-13 | 杭州牧星科技有限公司 | Unmanned aerial vehicle cluster control method and system based on artificial intelligence |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105974822A (en) * | 2016-06-13 | 2016-09-28 | 北京航空航天大学 | Spacecraft autonomous fly-around intersection control system verification device and spacecraft autonomous fly-around intersection control method |
| US20190291889A1 (en) * | 2018-03-26 | 2019-09-26 | Honeywell International Inc. | Method and system for generating an alert for an aircraft potentially exceeding speed limits in restricted airspace |
| CN113128047A (en) * | 2021-04-16 | 2021-07-16 | 长安大学 | Aero-engine thrust calculation method, storage medium and calculation device |
| CN114545771A (en) * | 2022-01-25 | 2022-05-27 | 南京航空航天大学 | Multi-mode self-adaptive switching control method and system for composite wing unmanned aerial vehicle |
| CN114954841A (en) * | 2022-06-24 | 2022-08-30 | 江苏科技大学 | Ship-machine-paddle power real-time matching method and equipment for ship power matching |
-
2023
- 2023-08-11 CN CN202311008917.1A patent/CN116738580B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105974822A (en) * | 2016-06-13 | 2016-09-28 | 北京航空航天大学 | Spacecraft autonomous fly-around intersection control system verification device and spacecraft autonomous fly-around intersection control method |
| US20190291889A1 (en) * | 2018-03-26 | 2019-09-26 | Honeywell International Inc. | Method and system for generating an alert for an aircraft potentially exceeding speed limits in restricted airspace |
| CN113128047A (en) * | 2021-04-16 | 2021-07-16 | 长安大学 | Aero-engine thrust calculation method, storage medium and calculation device |
| CN114545771A (en) * | 2022-01-25 | 2022-05-27 | 南京航空航天大学 | Multi-mode self-adaptive switching control method and system for composite wing unmanned aerial vehicle |
| CN114954841A (en) * | 2022-06-24 | 2022-08-30 | 江苏科技大学 | Ship-machine-paddle power real-time matching method and equipment for ship power matching |
Non-Patent Citations (2)
| Title |
|---|
| ANDREAS BRANDSÆTER 等: "ship speed prediction based on full scale sensor measurements of shaft thrust and environmental conditions", 《OCEAN ENGINEERING》 * |
| 尤顺 等: "分布式电推进飞机动力偏航非线性动态逆控制", 《航空动力学报》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116907282A (en) * | 2023-09-14 | 2023-10-20 | 杭州牧星科技有限公司 | Unmanned target aircraft ultra-low altitude flight control method based on artificial intelligence algorithm |
| CN116907282B (en) * | 2023-09-14 | 2024-01-05 | 杭州牧星科技有限公司 | Unmanned target aircraft ultra-low altitude flight control method based on artificial intelligence algorithm |
| CN117555355A (en) * | 2023-10-07 | 2024-02-13 | 杭州牧星科技有限公司 | Unmanned aerial vehicle cluster control method and system based on artificial intelligence |
| CN117555355B (en) * | 2023-10-07 | 2024-06-07 | 杭州牧星科技有限公司 | Unmanned aerial vehicle cluster control method and system based on artificial intelligence |
| CN117382435A (en) * | 2023-10-17 | 2024-01-12 | 浙江加力仓储设备股份有限公司 | Vehicle speed control method and system based on dip angle monitoring |
| CN117382435B (en) * | 2023-10-17 | 2024-05-03 | 浙江加力仓储设备股份有限公司 | Vehicle speed control method and system based on dip angle monitoring |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116738580B (en) | 2023-11-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116738580B (en) | Speed-thrust matching system and method for high subsonic unmanned aerial vehicle | |
| CN109543502B (en) | Semantic segmentation method based on deep multi-scale neural network | |
| CN107292333B (en) | A kind of rapid image categorization method based on deep learning | |
| CN108764063B (en) | Remote sensing image time-sensitive target identification system and method based on characteristic pyramid | |
| US10296804B2 (en) | Image recognizing apparatus, computer-readable recording medium, image recognizing method, and recognition apparatus | |
| CN108764317B (en) | Residual convolutional neural network image classification method based on multipath feature weighting | |
| CN111401516A (en) | Neural network channel parameter searching method and related equipment | |
| CN108304920B (en) | Method for optimizing multi-scale learning network based on MobileNet | |
| CN109063719B (en) | Image classification method combining structure similarity and class information | |
| CN107886169B (en) | Multi-scale convolution kernel method for generating confrontation network model based on text-image | |
| US11551076B2 (en) | Event-driven temporal convolution for asynchronous pulse-modulated sampled signals | |
| CN113326930A (en) | Data processing method, neural network training method, related device and equipment | |
| CN113128558A (en) | Target detection method based on shallow space feature fusion and adaptive channel screening | |
| WO2016036664A1 (en) | Event-driven spatio-temporal short-time fourier transform processing for asynchronous pulse-modulated sampled signals | |
| CN109559315B (en) | Water surface segmentation method based on multipath deep neural network | |
| CN112491442B (en) | Self-interference elimination method and device | |
| JP6891626B2 (en) | Information processing equipment, information processing system, information processing program and information processing method | |
| CN111738403A (en) | Neural network optimization method and related equipment | |
| CN115830596A (en) | Remote sensing image semantic segmentation method based on fusion pyramid attention | |
| Ye et al. | Recognition algorithm of emitter signals based on PCA+ CNN | |
| CN111027472A (en) | Video identification method based on fusion of video optical flow and image space feature weight | |
| CN116449881A (en) | Intelligent cylinder and control method thereof | |
| CN116704267A (en) | Deep learning 3D printing defect detection method based on improved YOLOX algorithm | |
| CN116662929A (en) | Training method of radar signal recognition model and radar signal recognition method | |
| CN115935257A (en) | Classification recognition method, computer device, and storage medium |
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 |