CN117288418B - Hypersonic wind tunnel aerodynamic force intelligent identification method based on wavelet decomposition - Google Patents
Hypersonic wind tunnel aerodynamic force intelligent identification method based on wavelet decomposition Download PDFInfo
- Publication number
- CN117288418B CN117288418B CN202311162585.2A CN202311162585A CN117288418B CN 117288418 B CN117288418 B CN 117288418B CN 202311162585 A CN202311162585 A CN 202311162585A CN 117288418 B CN117288418 B CN 117288418B
- Authority
- CN
- China
- Prior art keywords
- signal
- wavelet
- balance signal
- balance
- decomposition
- 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
- 238000000354 decomposition reaction Methods 0.000 title claims abstract description 67
- 238000000034 method Methods 0.000 title claims abstract description 59
- 238000012549 training Methods 0.000 claims abstract description 48
- 238000012360 testing method Methods 0.000 claims abstract description 30
- 238000013136 deep learning model Methods 0.000 claims abstract description 15
- 230000002457 bidirectional effect Effects 0.000 claims description 37
- 230000006870 function Effects 0.000 claims description 10
- 230000006798 recombination Effects 0.000 claims description 10
- 238000005215 recombination Methods 0.000 claims description 10
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 5
- 230000010355 oscillation Effects 0.000 claims description 4
- 238000010586 diagram Methods 0.000 description 17
- 238000001228 spectrum Methods 0.000 description 10
- 238000005259 measurement Methods 0.000 description 9
- 230000035939 shock Effects 0.000 description 4
- 230000001052 transient effect Effects 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000002238 attenuated effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 230000008054 signal transmission Effects 0.000 description 1
- 230000000087 stabilizing effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M9/00—Aerodynamic testing; Arrangements in or on wind tunnels
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M9/00—Aerodynamic testing; Arrangements in or on wind tunnels
- G01M9/02—Wind tunnels
- G01M9/04—Details
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M9/00—Aerodynamic testing; Arrangements in or on wind tunnels
- G01M9/06—Measuring arrangements specially adapted for aerodynamic testing
- G01M9/062—Wind tunnel balances; Holding devices combined with measuring arrangements
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Physics & Mathematics (AREA)
- Fluid Mechanics (AREA)
- General Physics & Mathematics (AREA)
- Aerodynamic Tests, Hydrodynamic Tests, Wind Tunnels, And Water Tanks (AREA)
Abstract
The invention discloses a hypersonic wind tunnel aerodynamic force intelligent identification method based on wavelet decomposition, which comprises the following steps: obtaining a balance signal and an ideal step signal in a simulated wind tunnel test process, decomposing the balance signal and the ideal step signal by adopting a wavelet decomposition method to respectively obtain wavelet coefficients of the balance signal and the ideal step signal, and simultaneously inputting the wavelet coefficients into a deep learning model for training to obtain a trained deep learning model; obtaining a balance signal and carrying out wavelet decomposition on the balance signal to obtain a wavelet coefficient of the balance signal, inputting a trained deep learning model to remove interference components, and reconstructing the wavelet coefficient of the balance signal with the interference components removed by adopting a wavelet reconstruction method to obtain a real aerodynamic signal; the hypersonic wind tunnel aerodynamic force intelligent identification method based on wavelet decomposition can effectively filter inertia force and other interference signals and accurately obtain real aerodynamic force signals.
Description
Technical Field
The invention relates to the technical field of shock tunnel force measurement tests, in particular to a hypersonic wind tunnel aerodynamic force intelligent identification method based on wavelet decomposition.
Background
The wind tunnel test is a key technology for development of an air suction hypersonic aircraft, high-precision aerodynamic force measurement is an important part of the air tunnel test, high-speed transient airflow can generate transient impact on an aircraft model installed in a force measuring system in the wind tunnel starting process, so that transient vibration is generated, model inertia force caused by the vibration can be collected by the force measuring system together with aerodynamic force, the effective time is short, the inertia force is difficult to attenuate completely, and finally an output signal of the force measuring system shows oscillation attenuation characteristics. In the wind tunnel test process, the signal transmission and acquisition of the force measuring system inevitably generate noise, mainly high-frequency white noise. In addition, in the test process, the frequency conversion impact can appear at the joint of the aircraft model and the connection part of the force measurement system, so that the frequency conversion signal can appear in the output signal of the force measurement system, and the aerodynamic force identification precision is affected. The current common load identification method for engineering mainly comprises the following steps: the mean value method, the frequency domain method, the time domain method and the traditional neural network method can not effectively reduce the interference of variable frequency signals, and part of methods can only identify one aerodynamic load constant and can not effectively reflect the magnitude and change process of dynamic aerodynamic load in the wind tunnel test process.
With the continuous development of aerospace technology, hypersonic technology is widely focused and studied in great countries of aviation, and scientific problems of hypersonic technology have important strategic significance. The aerodynamic force measurement test with high precision plays a decisive role in designing and optimizing the aerodynamic shape layout of the hypersonic aircraft. The shock wind tunnel force test can provide reliable data for the research of high-temperature real gas effects, and simultaneously provides key technical support for the research of hypersonic aircrafts in China. Currently, there are many unresolved key technical problems in the shock wind tunnel force test, and these problems cause the shock wind tunnel force test to be a challenging research topic. One of the most important problems is that the model force measuring system is subjected to structural inertial vibration caused by pneumatic impact of a transient flow field. When a force measurement test is carried out, a force measurement system formed by a model, a balance and a support is subjected to instantaneous impact to generate structural vibration, and vibration signals cannot be attenuated rapidly in a short time, so that an output signal of the force measurement system contains interference signals generated by inertial vibration, and the accuracy of the force measurement test is seriously affected.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a hypersonic wind tunnel aerodynamic intelligent identification method based on wavelet decomposition.
In order to achieve the aim of the invention, the invention adopts the following technical scheme:
a hypersonic wind tunnel aerodynamic force intelligent identification method based on wavelet decomposition comprises the following steps:
s1, acquiring a balance signal and an ideal step signal which simulate a wind tunnel test process;
S2, decomposing the balance signal and the ideal step signal obtained in the step S1 by adopting a wavelet decomposition method to respectively obtain wavelet coefficients of the balance signal and the ideal step signal;
s3, inputting the wavelet coefficients of the balance signal and the ideal step signal obtained in the step S2 into a deep learning model for training, and obtaining a trained deep learning model;
s4, acquiring a balance signal, carrying out wavelet decomposition on the acquired balance signal to obtain a wavelet coefficient of the balance signal, inputting the wavelet coefficient of the balance signal into the deep learning model trained in the step S3 to remove interference components, and reconstructing the wavelet coefficient of the balance signal from which the interference components are removed by adopting a wavelet reconstruction method to obtain a real aerodynamic signal.
Further, the step S1 specifically includes:
S11, realizing step loading in three directions of XYZ by a test bed through a motor, controlling the motor to load by an upper computer according to an S-shaped force sensor connected with a loading head, releasing a pneumatic device of the loading head to realize unloading when the expected load is reached, and obtaining a balance signal comprising a pre-loading signal, a post-loading stable signal and an post-unloading oscillation signal;
S12, subtracting the mean value of the signal before loading obtained in the step S11 from the mean value of the stable signal after loading obtained in the step S11 to obtain a step amplitude, and manually generating an ideal step signal according to the jump time.
Further, the step S2 specifically includes:
And (3) according to the balance signal and the ideal step signal obtained in the step (S1), decomposing frequency components of the balance signal and the ideal step signal by adopting a wavelet decomposition method to respectively obtain wavelet coefficients of the balance signal and the ideal step signal.
Further, the specific process of disassembling the frequency components of the balance signal and the ideal step signal by adopting the wavelet decomposition method is as follows:
Constructing a decomposition high-pass filter and a decomposition low-pass filter, selecting bior 3.3.3 wavelet parameters for constructing the decomposition high-pass filter and the decomposition low-pass filter, and respectively carrying out convolution operation on the balance signal and the ideal step signal and the constructed decomposition high-pass filter and the decomposition low-pass filter to respectively obtain wavelet detail coefficients and wavelet approximation coefficients of the balance signal and the ideal step signal of primary decomposition; and carrying out convolution operation on the wavelet approximation coefficients of the primary decomposed balance signal and the ideal step signal and the constructed decomposition high-pass filter and the decomposition low-pass filter to obtain the wavelet detail coefficients and the wavelet approximation coefficients of the secondary decomposed balance signal and the ideal step signal respectively.
Further, the step S3 specifically includes:
S31, taking the wavelet coefficient of the balance signal obtained in the step S2 as original data, taking the wavelet coefficient of an ideal step signal as a label of the original data, and inputting the original data endowed with the label into a bidirectional LSTM model as a data set;
S32, training the bidirectional LSTM model by taking a part of the data set in the step S31 as a training set, and evaluating the performance of the trained bidirectional LSTM model by taking the rest part of the data set as a testing set;
s33, selecting a mean square error as a loss function to predict a bidirectional LSTM model in training, and judging the difference degree between a predicted value of the bidirectional LSTM model and a wavelet coefficient of an ideal step signal, namely:
wherein MSE represents mean square error, n represents training times, Y i represents predicted value of bi-directional LSTM model of the ith training time, Wavelet coefficients representing an ideal step signal for training the ith time;
S34, selecting an Adam optimizer to optimize parameters of the bidirectional LSTM model, and generating the bidirectional LSTM model with smaller loss.
Further, the step S4 specifically includes:
s41, acquiring a balance signal, and decomposing the acquired balance signal by adopting a wavelet decomposition method to obtain a wavelet coefficient of the balance signal;
s42, inputting the wavelet coefficient of the balance signal obtained in the step S41 into the trained bidirectional LSTM model in the step S3 to remove interference components;
S43, reconstructing wavelet coefficients of the balance signal with the interference components removed in the step S42 by adopting a wavelet reconstruction method to obtain a real aerodynamic signal.
Further, step S43 specifically includes:
s431, adding the convolution of the wavelet detail coefficient of the balance signal and the recombination high-pass filter and the convolution of the wavelet approximation coefficient of the balance signal and the recombination low-pass filter to obtain the wavelet approximation coefficient of the upper stage of the balance signal;
s432, continuously repeating the operation of the step S431 with the wavelet approximation coefficient of the upper level of the balance signal obtained in the step S431 to obtain a real aerodynamic signal
The invention has the following beneficial effects:
According to the hypersonic wind tunnel aerodynamic force intelligent identification method based on wavelet decomposition, a real aerodynamic force signal can be accurately obtained, and even if the frequency of the aerodynamic force signal is overlapped with that of the aerodynamic force signal, inertia force and other interference signals can be effectively filtered; meanwhile, the invention adopts wavelet decomposition technology and deep learning technology to remove the interference frequency component of each level of wavelet coefficient, so that the filtering process has better interpretation.
Drawings
FIG. 1 is a schematic flow chart of a hypersonic wind tunnel aerodynamic force intelligent identification method based on wavelet decomposition;
FIG. 2 is a schematic diagram of power spectrum of model training data and model training data according to an embodiment;
FIG. 3 is a schematic diagram of an exploded low pass filter and an exploded high pass filter constructed in an embodiment;
FIG. 4 is a diagram of training set and test set loss in an embodiment;
FIG. 5 is a schematic diagram of model predictive results and model predictive results power spectra;
FIG. 6 is a schematic diagram of a constructed recombination low pass filter and recombination high pass filter;
FIG. 7 is a schematic diagram of a level 9 wavelet decomposition detail coefficient reconstructed signal and a power spectrum of the reconstructed signal;
fig. 8 is a schematic diagram of a level 13 wavelet decomposition detail coefficient recombined signal and a power spectrum of the recombined signal.
Detailed Description
The following description of the embodiments of the present invention is provided to facilitate understanding of the present invention by those skilled in the art, but it should be understood that the present invention is not limited to the scope of the embodiments, and all the inventions which make use of the inventive concept are protected by the spirit and scope of the present invention as defined and defined in the appended claims to those skilled in the art.
As shown in FIG. 1, the hypersonic wind tunnel aerodynamic force intelligent identification method based on wavelet decomposition comprises the following steps S1-S4:
s1, acquiring a balance signal and an ideal step signal which simulate a wind tunnel test process.
In an alternative embodiment of the invention, wind tunnel experiments are costly because of the large amount of training data required for deep learning, and the required data volume is difficult to meet. The installation mode of the aircraft and the balance of the experiment table is the same as that of a real wind tunnel test, and the obtained dynamic characteristics are similar to those of the real wind tunnel test. Therefore, a dynamic calibration test bed is adopted to simulate a wind tunnel experiment to obtain deep learning training data.
Specifically, step S1 specifically includes:
S11, realizing step loading in three directions of XYZ by the test bed through the motor, controlling the motor to load by the upper computer according to the S-shaped force sensor connected with the loading head, releasing the pneumatic device of the loading head to realize unloading when the expected load is reached, and obtaining a balance signal comprising a pre-loading signal, a post-loading stable signal and an post-unloading oscillation signal.
S12, subtracting the mean value of the signal before loading obtained in the step S11 from the mean value of the stable signal after loading obtained in the step S11 to obtain a step amplitude, and manually generating an ideal step signal according to the jump time.
The balance signal acquired in the embodiment comprises a pre-loading signal, a post-loading stabilizing signal and an post-unloading oscillating signal. The step load is obtained by subtracting the pre-loading signal from the post-loading stable signal, and the post-unloading oscillating signal comprises the step load and the system inertial vibration, so that the acquired balance signal has the system inertial vibration and interference components. Therefore, the step amplitude is obtained by subtracting the mean value of the signal before loading from the mean value of the stable signal after loading, an ideal step signal is artificially generated according to the jump time, and the generated ideal step signal is used as a label of a balance signal to be input into a deep learning model for training.
As shown in fig. 2, fig. 2 is a schematic diagram of model training data and a power spectrum of the model training data. In fig. 2 (a) is a model training data diagram, the abscissa represents training time, the ordinate represents normal force, the solid line represents an input balance signal, and the broken line represents an ideal step signal as a tag. In fig. 2 (b), a power spectrum diagram of model training data is shown, the abscissa represents frequency, the ordinate represents amplitude, the solid line represents an inputted balance signal, and the broken line represents an ideal step signal as a tag. From fig. 2 (b), it can be derived that the interference frequencies of the input balance signal are mainly 10.3Hz and 56.2Hz.
S2, decomposing the balance signal and the ideal step signal obtained in the step S1 by adopting a wavelet decomposition method to respectively obtain wavelet coefficients of the balance signal and the ideal step signal.
Specifically, step S2 specifically includes:
And (3) according to the balance signal and the ideal step signal obtained in the step (S1), decomposing frequency components of the balance signal and the ideal step signal by adopting a wavelet decomposition method to respectively obtain wavelet coefficients of the balance signal and the ideal step signal.
Specifically, the specific process of disassembling the frequency components of the balance signal and the ideal step signal by adopting the wavelet decomposition method is as follows:
Constructing a decomposition high-pass filter and a decomposition low-pass filter, selecting bior 3.3.3 wavelet parameters for constructing the decomposition high-pass filter and the decomposition low-pass filter, and respectively carrying out convolution operation on the balance signal and the ideal step signal and the constructed decomposition high-pass filter and the decomposition low-pass filter to respectively obtain wavelet detail coefficients and wavelet approximation coefficients of the balance signal and the ideal step signal of primary decomposition; and carrying out convolution operation on the wavelet approximation coefficients of the primary decomposed balance signal and the ideal step signal and the constructed decomposition high-pass filter and the decomposition low-pass filter to obtain the wavelet detail coefficients and the wavelet approximation coefficients of the secondary decomposed balance signal and the ideal step signal respectively.
As shown in fig. 3, fig. 3 is a schematic diagram of a structured split low-pass filter and split high-pass filter. In this embodiment, the frequency components of the obtained balance signal and the ideal step signal are decomposed by a wavelet decomposition method respectively, and are decomposed into 14-level wavelet coefficients, where each level wavelet coefficient includes frequency information and phase information of the original signal.
S3, inputting the wavelet coefficients of the balance signal and the ideal step signal obtained in the step S2 into a deep learning model for training, and obtaining a trained deep learning model.
In the embodiment, the deep learning model adopts a bidirectional LSTM model, the wavelet coefficients of the balance signal and the ideal step signal which are respectively obtained are input into the deep learning model for training, and the hidden layer number, the hidden layer dimension and the dropout value are adjusted according to the training effect of the model in the training of the bidirectional LSTM model. Wherein, the hidden layer number is selected from 1, 2 and 4, the hidden layer dimension is selected from 5 and 10, the dropout value is selected from 0, 0.2 and 0.6, the learning rate is selected from 0.01, 0.001 and 0.0001, the Batchsize is selected from 5 and 10, and the training round number is selected from 1000, 5000 and 10000. Finally, the hidden layer dimension is determined to be 10 by comprehensively considering the loss size and the training time, and the hidden layer comprises two hidden layers, and the dropout value is 0.2. And selecting a mean square error MSE as a loss function, and selecting an Adam optimizer by the optimizer, wherein the learning rate is 0.001, the Batchsize is 10, and the training round number is 10000. In this embodiment, the loss function is an operation function for measuring the difference degree between the predicted value and the tag value of the bidirectional LSTM model, that is, the wavelet coefficient of the ideal step signal, and is a non-negative real value function, and the smaller the loss function is, the better the robustness of the model is. The loss function is mainly used in the training stage of the bidirectional LSTM model, after the training data of each batch is sent into the bidirectional LSTM model, the predicted value is output through forward propagation, and then the difference value between the predicted value and the tag value, namely the wavelet coefficient of the ideal step signal, is calculated by the loss function, and the difference value is also the loss value. After the loss value is obtained, the bidirectional LSTM model updates each parameter through back propagation, so that the loss between the tag value and the predicted value is reduced, the predicted value of the bidirectional LSTM model is close to the tag value direction, and the Adam optimizer optimizes the parameters of the bidirectional LSTM model by adopting a gradient descent method to generate the bidirectional LSTM model with smaller loss, so that the learning purpose is achieved.
Specifically, step S3 specifically includes S31-S34:
S31, taking the wavelet coefficient of the balance signal obtained in the step S2 as original data, taking the wavelet coefficient of the ideal step signal as a label of the original data, and inputting the original data given to the label as a data set into the bidirectional LSTM model.
And S32, training the bidirectional LSTM model by taking a part of the data set in the step S31 as a training set, and evaluating the performance of the trained bidirectional LSTM model by taking the rest part of the data set as a testing set.
S33, selecting a mean square error as a loss function to predict a bidirectional LSTM model in training, and judging the difference degree between a predicted value of the bidirectional LSTM model and a wavelet coefficient of an ideal step signal, namely:
wherein MSE represents mean square error, n represents training times, Y i represents predicted value of bi-directional LSTM model of the ith training time, The wavelet coefficients representing the ideal step signal for training the ith time.
S34, selecting an Adam optimizer to optimize parameters of the bidirectional LSTM model, and generating the bidirectional LSTM model with smaller loss.
As shown in fig. 4, fig. 4 is a schematic diagram of the training set and the test set loss, the abscissa represents the number of training rounds, the ordinate represents the mean square error, the solid line represents the training set loss, and the ordinate represents the test set loss. In this embodiment, the wavelet coefficient of the balance signal obtained in step S2 is used as the original data, the wavelet coefficient of the ideal step signal is used as the label of the original data, the original data given to the label is used as the data set to be input into the bidirectional LSTM model, the total data set includes 180 sets of data sets, 150 sets of data sets are divided into training sets for training the bidirectional LSTM model, the remaining 30 sets are divided into test sets, and the performance of the trained bidirectional LSTM model is evaluated. As can be seen from fig. 4, training to 10000 rounds of test set loss and training set loss were minimal.
S4, acquiring a balance signal, carrying out wavelet decomposition on the acquired balance signal to obtain a wavelet coefficient of the balance signal, inputting the wavelet coefficient of the balance signal into the deep learning model trained in the step S3 to remove interference components, and reconstructing the wavelet coefficient of the balance signal from which the interference components are removed by adopting a wavelet reconstruction method to obtain a real aerodynamic signal.
As shown in fig. 5, fig. 5 is a schematic diagram of model prediction results and power spectra of the model prediction results, and fig. 5 (a) shows a model prediction result diagram, the abscissa shows time, and the ordinate shows normal force; in fig. 5, (b) shows a model predictive result power spectrum, the abscissa shows frequency, and the ordinate shows amplitude; the model predictive signal input to the model is a balance signal. In this embodiment, the obtained balance signal is input into the trained bidirectional LSTM model for prediction, and a real aerodynamic signal can be obtained from fig. 5 (a).
Specifically, step S4 specifically includes S41-S42:
s41, acquiring a balance signal, and decomposing the acquired balance signal by adopting a wavelet decomposition method to obtain a wavelet coefficient of the balance signal.
S42, inputting the wavelet coefficient of the balance signal obtained in the step S41 into the trained bidirectional LSTM model in the step S3 to remove the interference component.
S43, reconstructing wavelet coefficients of the balance signal with the interference components removed in the step S42 by adopting a wavelet reconstruction method to obtain a real aerodynamic signal.
Specifically, step S43 includes steps S431-S432:
s431, the convolution of the wavelet detail coefficient of the balance signal and the recombination high-pass filter and the convolution of the wavelet approximation coefficient of the balance signal and the recombination low-pass filter are added, so that the wavelet approximation coefficient of the upper stage of the balance signal is obtained.
And S432, continuously repeating the operation of the step S431 by using the wavelet approximation coefficient of the upper stage of the balance signal obtained in the step S431 to obtain a real aerodynamic signal.
As shown in fig. 6, fig. 6 is a schematic diagram of a structured recombination low-pass filter and a recombination high-pass filter. In the embodiment, the wavelet coefficient of the balance signal decomposed into 14 levels is reconstructed by adopting a wavelet reconstruction method, so that a real aerodynamic signal is obtained.
As shown in fig. 7, fig. 7 is a schematic diagram of the level 9 wavelet decomposition detail coefficient recombined signal and the power spectrum of the recombined signal. As can be seen from fig. 7, the level 9 wavelet decomposition detail components separate out the interference frequencies of 56.2 Hz.
As shown in fig. 8, fig. 8 is a schematic diagram of a level 13 wavelet decomposition detail coefficient recombined signal and a power spectrum of the recombined signal. As can be seen from fig. 8, the level 13 wavelet decomposition detail components separate out the interference frequencies of 10.3 Hz.
The principles and embodiments of the present invention have been described in detail with reference to specific examples, which are provided to facilitate understanding of the method and core ideas of the present invention; meanwhile, as those skilled in the art will have variations in the specific embodiments and application scope in accordance with the ideas of the present invention, the present description should not be construed as limiting the present invention in view of the above.
Those of ordinary skill in the art will recognize that the embodiments described herein are for the purpose of aiding the reader in understanding the principles of the present invention and should be understood that the scope of the invention is not limited to such specific statements and embodiments. Those of ordinary skill in the art can make various other specific modifications and combinations from the teachings of the present disclosure without departing from the spirit thereof, and such modifications and combinations remain within the scope of the present disclosure.
Claims (5)
1. The hypersonic wind tunnel aerodynamic force intelligent identification method based on wavelet decomposition is characterized by comprising the following steps of:
S1, acquiring a balance signal and an ideal step signal which simulate a wind tunnel test process, wherein the specific process is as follows:
S11, realizing step loading in three directions of XYZ by a test bed through a motor, controlling the motor to load by an upper computer according to an S-shaped force sensor connected with a loading head, releasing a pneumatic device of the loading head to realize unloading when the expected load is reached, and obtaining a balance signal comprising a pre-loading signal, a post-loading stable signal and an post-unloading oscillation signal;
S12, subtracting the mean value of the signal before loading obtained in the step S11 from the mean value of the stable signal after loading obtained in the step S11 to obtain a step amplitude, and manually generating an ideal step signal according to the jump time;
S2, decomposing the balance signal and the ideal step signal obtained in the step S1 by adopting a wavelet decomposition method to respectively obtain wavelet coefficients of the balance signal and the ideal step signal;
s3, inputting the wavelet coefficients of the balance signal and the ideal step signal obtained in the step S2 into a deep learning model for training, and obtaining a trained deep learning model, wherein the specific process is as follows:
S31, taking the wavelet coefficient of the balance signal obtained in the step S2 as original data, taking the wavelet coefficient of an ideal step signal as a label of the original data, and inputting the original data endowed with the label into a bidirectional LSTM model as a data set;
S32, training the bidirectional LSTM model by taking a part of the data set in the step S31 as a training set, and evaluating the performance of the trained bidirectional LSTM model by taking the rest part of the data set as a testing set;
s33, selecting a mean square error as a loss function to predict a bidirectional LSTM model in training, and judging the difference degree between a predicted value of the bidirectional LSTM model and a wavelet coefficient of an ideal step signal, namely:
Wherein MSE represents mean square error, n represents training times, Y t represents predicted value of bi-directional LSTM model of the ith training time, Wavelet coefficients representing an ideal step signal for training the ith time;
S34, selecting an Adam optimizer to optimize parameters of the bidirectional LSTM model, and generating a bidirectional LSTM model with smaller loss;
s4, acquiring a balance signal, carrying out wavelet decomposition on the acquired balance signal to obtain a wavelet coefficient of the balance signal, inputting the wavelet coefficient of the balance signal into the deep learning model trained in the step S3 to remove interference components, and reconstructing the wavelet coefficient of the balance signal from which the interference components are removed by adopting a wavelet reconstruction method to obtain a real aerodynamic signal.
2. The hypersonic wind tunnel aerodynamic force intelligent identification method based on wavelet decomposition according to claim 1, wherein step S2 specifically comprises:
And (3) according to the balance signal and the ideal step signal obtained in the step (S1), decomposing frequency components of the balance signal and the ideal step signal by adopting a wavelet decomposition method to respectively obtain wavelet coefficients of the balance signal and the ideal step signal.
3. The hypersonic wind tunnel aerodynamic force intelligent identification method based on wavelet decomposition according to claim 2, wherein the specific process of decomposing the frequency components of a balance signal and an ideal step signal by adopting a wavelet decomposition method is as follows:
Constructing a decomposition high-pass filter and a decomposition low-pass filter, selecting bior 3.3.3 wavelet parameters for constructing the decomposition high-pass filter and the decomposition low-pass filter, and respectively carrying out convolution operation on the balance signal and the ideal step signal and the constructed decomposition high-pass filter and the decomposition low-pass filter to respectively obtain wavelet detail coefficients and wavelet approximation coefficients of the balance signal and the ideal step signal of primary decomposition; and carrying out convolution operation on the wavelet approximation coefficients of the primary decomposed balance signal and the ideal step signal and the constructed decomposition high-pass filter and the decomposition low-pass filter to obtain the wavelet detail coefficients and the wavelet approximation coefficients of the secondary decomposed balance signal and the ideal step signal respectively.
4. The hypersonic wind tunnel aerodynamic force intelligent identification method based on wavelet decomposition according to claim 1, wherein step S4 specifically comprises:
s41, acquiring a balance signal, and decomposing the acquired balance signal by adopting a wavelet decomposition method to obtain a wavelet coefficient of the balance signal;
s42, inputting the wavelet coefficient of the balance signal obtained in the step S41 into the trained bidirectional LSTM model in the step S3 to remove interference components;
S43, reconstructing wavelet coefficients of the balance signal with the interference components removed in the step S42 by adopting a wavelet reconstruction method to obtain a real aerodynamic signal.
5. The intelligent identification method of hypersonic wind tunnel aerodynamic force based on wavelet decomposition according to claim 4, wherein step S43 specifically comprises:
s431, adding the convolution of the wavelet detail coefficient of the balance signal and the recombination high-pass filter and the convolution of the wavelet approximation coefficient of the balance signal and the recombination low-pass filter to obtain the wavelet approximation coefficient of the upper stage of the balance signal;
And S432, continuously repeating the operation of the step S431 by using the wavelet approximation coefficient of the upper stage of the balance signal obtained in the step S431 to obtain a real aerodynamic signal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311162585.2A CN117288418B (en) | 2023-09-08 | 2023-09-08 | Hypersonic wind tunnel aerodynamic force intelligent identification method based on wavelet decomposition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311162585.2A CN117288418B (en) | 2023-09-08 | 2023-09-08 | Hypersonic wind tunnel aerodynamic force intelligent identification method based on wavelet decomposition |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117288418A CN117288418A (en) | 2023-12-26 |
| CN117288418B true CN117288418B (en) | 2024-05-07 |
Family
ID=89243584
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311162585.2A Active CN117288418B (en) | 2023-09-08 | 2023-09-08 | Hypersonic wind tunnel aerodynamic force intelligent identification method based on wavelet decomposition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117288418B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0340316A1 (en) * | 1988-04-30 | 1989-11-08 | Carl Schenck Ag | Calibrating device for an internal wind tunnel balance |
| CN111473947A (en) * | 2020-05-07 | 2020-07-31 | 中国科学院力学研究所 | Single-vector dynamic calibration method for pulse wind tunnel force measurement system |
| CN113970419A (en) * | 2021-10-13 | 2022-01-25 | 中国科学院力学研究所 | Shock tunnel force measurement balance signal data processing method based on time-frequency transformation |
| CN113970420A (en) * | 2021-10-13 | 2022-01-25 | 中国科学院力学研究所 | Deep learning-based shock tunnel force measurement signal frequency domain analysis method |
-
2023
- 2023-09-08 CN CN202311162585.2A patent/CN117288418B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0340316A1 (en) * | 1988-04-30 | 1989-11-08 | Carl Schenck Ag | Calibrating device for an internal wind tunnel balance |
| CN111473947A (en) * | 2020-05-07 | 2020-07-31 | 中国科学院力学研究所 | Single-vector dynamic calibration method for pulse wind tunnel force measurement system |
| CN113970419A (en) * | 2021-10-13 | 2022-01-25 | 中国科学院力学研究所 | Shock tunnel force measurement balance signal data processing method based on time-frequency transformation |
| CN113970420A (en) * | 2021-10-13 | 2022-01-25 | 中国科学院力学研究所 | Deep learning-based shock tunnel force measurement signal frequency domain analysis method |
Non-Patent Citations (2)
| Title |
|---|
| A new aerodynamic identification technology for short time hypersonic wind tunnel while considering inertial force interference;Shichao Li;《Aerospace science and technology》;20230404;全文 * |
| 基于时频变换的激波风洞天平信号分析与处理;聂少军;《力学学报》;20220131;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117288418A (en) | 2023-12-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102750543B (en) | A kind of Classification of Transient Power Quality Disturbances recognition methods of composing kurtosis based on BUD | |
| Zhang et al. | Fault identification based on PD ultrasonic signal using RNN, DNN and CNN | |
| AU2014202530B2 (en) | Electrical power system stability | |
| CN103235844A (en) | Generation method for complex electromagnetic environment simulation scene | |
| CN111914705A (en) | Signal generation method and device for improving health state evaluation accuracy of reactor | |
| CN111707875A (en) | Electric control pry shielding effectiveness testing device and method for gas gathering station field | |
| CN106971066A (en) | Method based on Neural Network Models To Prediction geomagnetic storm | |
| CN117288418B (en) | Hypersonic wind tunnel aerodynamic force intelligent identification method based on wavelet decomposition | |
| CN111951833A (en) | Voice test method and device, electronic equipment and storage medium | |
| CN111144022A (en) | Satellite random vibration and noise test clipping method based on surface-to-mass ratio | |
| CN113919388B (en) | Electromechanical equipment fault diagnosis method and device integrating signal spectrum amplitude modulation and deep learning | |
| Randall et al. | Updating modal models from response measurements | |
| CN115900802A (en) | Universal dynamic compensation method for multi-type sensors | |
| CN111985408B (en) | Porcelain post insulator fault identification method and device | |
| CN111428342B (en) | Random dynamic load identification method based on frequency domain spectrum decomposition | |
| CN114441111A (en) | Pipeline leakage infrasonic wave signal analysis method and system | |
| Locci et al. | Acoustic Optimization of a Muffler through the Sherpa Algorithm | |
| Xu et al. | Optimization of Damage Equivalent Accelerated Test Spectrum Derivation Using Multiple Non‐Gaussian Vibration Data | |
| Hu et al. | Prediction of Operational Noise Uncertainty in Automotive Micro-Motors Based on Multi-Branch Channel–Spatial Adaptive Weighting Strategy | |
| CN113408357B (en) | Method for generating flutter turbulence response signal to impulse response signal | |
| CN116884417B (en) | Weight distribution-based transformer voiceprint spectrum feature enhancement method and system | |
| CN114444551B (en) | Aerodynamic load identification method based on wavelet transform and convolution self-encoder | |
| Wang et al. | Study on the Simulation Method for Track Fault Diagnosis of Mechanical Parking Equipment | |
| Lv et al. | Technical analysis of transformer vowel print fault diagnosis based on CNN-LSTM network. | |
| CN116894414A (en) | Vibration mode-state characteristic analysis method for three-phase unbalanced running winding of transformer |
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 |