CN115795282A - Shock tube dynamic pressure reconstruction method, device, electronic equipment and storage medium - Google Patents

Abstract

The invention provides a shock tube dynamic pressure reconstruction method, a shock tube dynamic pressure reconstruction device, electronic equipment and a storage medium, wherein the method comprises the following steps: acquiring an initial dynamic pressure response signal, which comprises a vibration signal and a response signal; preprocessing the vibration signal and the response signal based on a variational modal decomposition method and an empirical modal decomposition method to obtain a preprocessed signal and component signals of the signal in different frequency bands; then, a training set is constructed according to the relation between each component signal obtained by the correlation coefficient and the preprocessed signal and the denoising vibration signal; constructing an initial inverse sensor network model based on a Bi-LSTM neural network model, and iteratively training the initial inverse sensor network model based on a training set to obtain a target inverse sensor network model; and reconstructing the dynamic pressure of the shock tube by adopting the target inverse sensing network model to obtain a dynamic pressure reconstruction signal of the target shock tube. The dynamic pressure reconstruction method for the shock tube provided by the invention can be used for reconstructing dynamic pressure signals of the shock tube more reasonably and accurately.

Description

Shock tube dynamic pressure reconstruction method, device, electronic equipment and storage medium

technical field

The invention relates to the technical field of metering and testing, in particular to a shock tube dynamic pressure reconstruction method, device, electronic equipment and storage medium.

Background technique

Shock tube dynamic pressure widely exists in the fields of explosion test, medical instrument, material impact test, aero engine and so on. In actual measurement, the stable duration of dynamic pressure is generally several milliseconds to ten milliseconds. The generation of dynamic pressure in the shock tube is a transient dynamic process, not only the test environment is complex, but also difficult to control, which makes it difficult to accurately estimate the dynamic pressure signal and seriously affects the measurement accuracy of the dynamic pressure signal.

The existing methods regard the dynamic pressure generated by the shock tube as an ideal step pressure, that is, the amplitude of the dynamic pressure is constant. However, the amplitude of the dynamic pressure generated by the actual shock tube fluctuates with time, and there are certain idealized assumptions to characterize the dynamic pressure of the shock tube with a constant amplitude, which will inevitably lead to unreasonable characterization results. Moreover, during the working process of the shock tube, when the incident shock wave reaches the end face of the low-pressure chamber, shock vibration will be generated. The pressure sensor installed on the end face will be excited by the dynamic pressure signal and the vibration signal at the same time. The collected output signal of the pressure sensor is the dynamic pressure The existing methods do not consider the influence of the vibration response, resulting in inaccurate estimation results of the dynamic pressure amplitude.

In summary, the prior art does not consider the fluctuation characteristics of the dynamic pressure signal of the shock tube and the impact of shock vibration on the dynamic pressure reconstruction results when reconstructing the dynamic pressure of the shock tube, resulting in a lack of rationality and consistency in the estimation of the dynamic pressure amplitude. accuracy.

Contents of the invention

In view of this, it is necessary to provide a shock tube dynamic pressure reconstruction method, device, electronic equipment and storage medium to solve the problem that the shock tube dynamic pressure signal is not considered in the prior art when the shock tube dynamic pressure is reconstructed. The fluctuation characteristics of the dynamic pressure amplitude and the influence of shock vibration on the dynamic pressure reconstruction results lead to the lack of rationality and accuracy of the dynamic pressure amplitude estimation results.

In order to solve the above technical problems, on the one hand, the present invention provides a shock tube dynamic pressure reconstruction method, including:

acquiring an initial dynamic pressure response signal, the initial dynamic pressure response signal including a vibration signal and a response signal;

Preprocessing the vibration signal and the response signal based on the variational mode decomposition method and the empirical mode decomposition method to obtain the denoised vibration signal, the preprocessing response signal, and the component signals of the preprocessing response signal in different frequency bands;

Constructing a training set according to the correlation between the component signals of the different frequency bands and the denoising vibration signal and the preprocessing response signal;

Constructing an initial inverse sensor network model based on the Bi-LSTM neural network model, and iteratively training the initial inverse sensor network model based on the training set to obtain a target inverse sensor network model;

The real-time dynamic pressure response signal is obtained and input into the target inverse sensor network model after performing the preprocessing operation to obtain the dynamic pressure reconstruction signal of the target shock tube.

In some possible implementations, the variational mode decomposition method and the empirical mode decomposition method perform preprocessing operations on the vibration signal and the response signal to obtain the denoised vibration signal, the preprocessed response signal, and the preprocessed Process component signals of response signals in different frequency bands, including:

Decompose the vibration signal based on the variational mode decomposition method to obtain a plurality of vibration modal components, respectively calculate the correlation coefficients between the multiple vibration modal components and the vibration signal, remove high-frequency noise components, and obtain the described removal. noise vibration signal;

Decomposing the response signal based on a variational mode decomposition method to obtain a plurality of response modal components, and reconstructing the plurality of response modal components based on the sensor ringing frequency to obtain a plurality of reconstructed signals;

Decomposing the multiple reconstructed signals based on an empirical mode decomposition method to obtain multiple reconstructed signal eigenmode function components;

Calculate the correlation coefficients between the multiple reconstruction signal eigenmode function components and the denoising vibration signal respectively, wherein the reconstruction signal corresponding to the reconstruction signal eigenmode function component with the largest correlation coefficient with the denoising vibration signal is The preprocessing response signal;

The preprocessing response signal is decomposed based on an empirical mode decomposition method to obtain component signals of the preprocessing response signal in different frequency bands.

In some possible implementation manners, the constructing a training set according to the correlation between the component signals of different frequency bands and the denoised vibration signal and the preprocessed response signal respectively includes:

Calculate the correlation coefficients between the component signals of different frequency bands and the preprocessing response signal and the denoised vibration signal, and use the component signal with the largest correlation coefficient with the preprocessing response signal among the component signals of different frequency bands as the vibration signal ring component signal;

Taking the component signals of the component signals of different frequency bands except the ringing frequency component and the component signal whose correlation coefficient value is smaller than the set threshold value with the denoising vibration signal as the noise component signal, and removing the noise component signal to obtain the denoising response signal, Reconstruct response signals, vibration-related component signals and trend signals;

The training set is constructed based on the reconstructed response signal, the vibration-related component signal and the trend component signal.

In some possible implementation manners, the trend component signal corresponds to the amplitude of the target shock tube dynamic pressure reconstruction signal, and can reflect the amplitude characteristics of the target shock tube dynamic pressure reconstruction signal.

In some possible implementation manners, the iterative training of the initial inverse sensor network model based on the training set to obtain the target inverse sensor network model includes:

Construct an initial inverse sensor network model first training set input based on the reconstructed response signal, construct an initial inverse sensor network model second training set input based on the vibration-related component, and initialize an inverse sensor network based on the trend component signal Model training set output;

After setting the number of hidden layers and hyperparameters of the initial inverse sensor network model, iteratively train the initial inverse sensor network model based on the constructed training set;

When the output loss rate of the initial inverse sensor network model is lower than a set loss threshold, the target inverse sensor network model is obtained.

In some possible implementation manners, after setting the number of hidden layers and hyperparameters of the initial inverse sensor network model, iteratively training the initial inverse sensor network model based on the constructed training set, including:

The number of initial hidden layers of the initial inverse sensor network model is set, and the hidden layers include several neuron units;

Setting the hyperparameters of the initial inverse sensor network model, the hyperparameters include: optimizer parameters, learning rate, sequence length and training rounds;

In the iterative training process, the initial hidden layer number, optimizer parameters, learning rate, sequence length and training rounds of the initial inverse sensor network model are adjusted, and the number of each neuron unit node inside the hidden layer is determined. Weights and offsets, so that the output of the initial inverse sensor network model minimizes the root mean square error of the output and reaches the set loss threshold, and completes the iterative training process of the initial inverse sensor network model.

In some possible implementation manners, the real-time dynamic pressure response signal is input into the target inverse sensor network model after performing the preprocessing operation to obtain the dynamic pressure reconstruction signal of the target shock tube, including:

Obtaining a real-time dynamic pressure response signal of the shock tube sensor, performing the preprocessing operation on the real-time response signal to obtain the denoising vibration signal and denoising response signal corresponding to the real-time dynamic pressure response signal;

Input the denoised vibration signal and denoised response signal corresponding to the real-time dynamic pressure response signal into the target inverse sensor network model to obtain the output of the target inverse sensor network model, and divide the output by the pressure sensor magnification and sensitivity , to obtain the dynamic pressure reconstruction signal of the target shock tube.

On the other hand, the present invention also provides a shock tube dynamic pressure reconstruction device, including:

A signal acquisition module, configured to acquire an initial dynamic pressure response signal, where the initial dynamic pressure response signal includes a vibration signal and a response signal;

The signal processing module is used to perform preprocessing operations on the vibration signal and the response signal based on the variational mode decomposition method and the empirical mode decomposition method, to obtain the denoised vibration signal, the preprocessing response signal, and the preprocessing response signal in the Component signals in different frequency bands;

A signal construction module, configured to construct a training set according to the correlation between the component signals of the different frequency bands and the denoising vibration signal and the preprocessing response signal;

A model training module, for constructing an initial inverse sensor network model based on the Bi-LSTM neural network model, and iteratively training the initial inverse sensor network model based on the training set to obtain a target inverse sensor network model;

The target reconstruction module is used to obtain the real-time dynamic pressure response signal and input the target inverse sensor network model after performing the preprocessing operation to obtain the target shock tube dynamic pressure reconstruction signal.

On the other hand, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and operable on the processor. Wave tube dynamic pressure reconstruction method.

Finally, the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the shock tube dynamic pressure reconstruction method described in the above-mentioned implementation manners is realized.

The beneficial effect of adopting the above-mentioned embodiment is: the shock tube dynamic pressure reconstruction method provided by the present invention, on the one hand, utilizes the variational mode decomposition method and the empirical mode decomposition method to decompose and reconstruct the response signal and the vibration signal, And comprehensively consider the correlation between each component signal and the response signal and the vibration signal, and consider the fluctuation of the actual dynamic pressure amplitude generated by the shock tube with time and the influence of the vibration signal, so that the data obtained by preprocessing has a higher Reasonability and accuracy. On the other hand, the inverse sensor network model based on the Bi-LSTM neural network reconstructs the dynamic pressure of the shock tube by using bidirectional input, which makes the signal feature extraction of the model more detailed and further improves The rationality and accuracy of shock tube reconstruction of dynamic pressure are confirmed.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings that need to be used in the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those skilled in the art, other drawings can also be obtained based on these drawings without any creative effort.

Fig. 1 is a schematic flow chart of an embodiment of a shock tube dynamic pressure reconstruction method provided by the present invention;

FIG. 2 is a schematic flow diagram of an embodiment of step S102 in FIG. 1 provided by the present invention;

FIG. 3 is a schematic flowchart of an embodiment of step S103 in FIG. 1 provided by the present invention;

FIG. 4 is a schematic flow diagram of an embodiment of step S104 in FIG. 1 provided by the present invention;

Fig. 5 is a schematic diagram of an embodiment of sensor raw measurement data provided by the present invention;

FIG. 6 is a schematic diagram of an embodiment of constructing training set data provided by the present invention;

Fig. 7 is a schematic diagram of an embodiment of the model training result provided by the present invention;

Fig. 8 is a schematic diagram of an embodiment of a dynamic pressure reconstruction signal provided by the present invention;

Fig. 9 is a schematic flowchart of an embodiment of a shock tube dynamic pressure reconstruction device provided by the present invention;

FIG. 10 is a schematic structural diagram of an embodiment of an electronic device provided by the present invention.

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Apparently, the described embodiments are only some of the embodiments of the present invention, but not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative efforts belong to the protection scope of the present invention.

It should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this disclosure illustrate operations implemented in accordance with some embodiments of the present invention. It should be understood that the operations of the flowcharts may be performed out of order, and steps that do not have a logical context may be performed in reverse order or simultaneously. In addition, those skilled in the art may add one or more other operations to the flowchart, or remove one or more operations from the flowchart under the guidance of the content of the present invention.

Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, or in one or more hardware modules or integrated circuits, or in different networks and/or processor systems and/or microcontroller systems.

Reference herein to an "embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the present invention. The occurrences of this phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is understood explicitly and implicitly by those skilled in the art that the embodiments described herein can be combined with other embodiments.

Before the description of the embodiments, relevant words are defined:

Shock tube: The shock tube is the core equipment for pressure sensor calibration and is used to generate plane shock waves. The so-called shock wave is a sudden change in the pressure somewhere in the gas, and the pressure wave propagates at a high speed. The speed of the wave is related to the strength of the pressure change, the greater the pressure change, the higher the wave speed. During the propagation process, when the wave front reaches a certain place, the pressure, density and temperature of the gas will change suddenly; when the wave front is not everywhere, the gas will not be disturbed by the wave; All are higher than the front of the wave front, and the gas particles flow in the direction of the wave front, and their speed is lower than that of the wave front.

Bi-LSTM neural network: Bidirectional long-term short-term memory neural network, Bi-LSTM neural network is a double-fed neural network, composed of two independent long-term short-term memory (LSTM) neural networks, the input sequence is in positive order and reverse order respectively Input into two long short-term memory networks, followed by feature extraction.

Variational mode decomposition method: (Variational mode decomposition, VMD) is an adaptive, completely non-recursive method of modal variation and signal processing. This technology has the advantage of being able to determine the number of modal decompositions, and its adaptability is to determine the number of modal decompositions of a given sequence according to the actual situation, and the subsequent search and solution process can adaptively match each modal The optimal center frequency and limited bandwidth can realize the effective separation of the intrinsic mode components (IMF), the frequency domain division of the signal, and then obtain the effective decomposition components of the given signal, and finally obtain the optimal solution of the variational problem. It overcomes the problems of endpoint effect and modal component aliasing in the EMD method, and has a more solid mathematical theoretical foundation, which can reduce the non-stationarity of time series with high complexity and strong nonlinearity, and obtain multiple different frequency scales by decomposition And relatively stable subsequences are suitable for non-stationary sequences. The core idea of VMD is to construct and solve variational problems.

Empirical Mode Decomposition method: (Empirical Mode Decomposition, EMD) is based on the time scale characteristics of the data itself to decompose the signal, without pre-setting any basis function, it is a time-frequency domain signal processing method. EMD has obvious advantages in dealing with non-stationary and nonlinear data, and is suitable for analyzing nonlinear and non-stationary signal sequences with a high signal-to-noise ratio. The key to this method is empirical mode decomposition, which decomposes complex signals into a finite number of Intrinsic Mode Functions (IMFs), and each decomposed IMF component contains local characteristic information of different time scales of the original signal.

Based on the description of the above technical terms, the commonly used reconstruction methods for dynamic pressure in the prior art are theoretical calculation method and reverse reconstruction method. The theoretical calculation method uses the propagation characteristics of the incident shock wave in the shock tube and the formula of the moving shock wave to establish a calculation model for the theoretical amplitude of the dynamic pressure of the shock tube, and measures the initial temperature and initial pressure of the low-pressure chamber of the shock tube and the incident shock wave The propagation velocity of the shock tube is obtained to obtain the amplitude of the dynamic pressure of the shock tube; the inverse reconstruction method first estimates the trend of the response signal of the pressure sensor, obtains the stable interval of the trend signal, and then obtains the stable value of the response signal, and finally calculates the stable value of the response signal Divided by the sensitivity of the sensor and the acquisition amplification factor, the stable value of the dynamic pressure signal of the shock tube is obtained. However, the existing technology does not consider the fluctuation characteristics of the shock tube dynamic pressure signal and the impact of shock vibration on the dynamic pressure reconstruction results when reconstructing the dynamic pressure of the shock tube, resulting in a lack of rationality and accuracy in the estimation of the dynamic pressure amplitude. , the present invention aims to propose a shock tube dynamic pressure reconstruction method with higher rationality and accuracy.

Specific embodiments will be described in detail below respectively. It should be noted that the description sequence of the following embodiments is not intended to limit the preferred sequence of the embodiments.

Embodiments of the present invention provide a shock tube dynamic pressure reconstruction method, device, electronic equipment and storage medium.

As shown in Figure 1, Figure 1 is a schematic flow chart of an embodiment of a shock tube dynamic pressure reconstruction method provided by the present invention, the shock tube dynamic pressure reconstruction method includes:

S101. Acquire an initial dynamic pressure response signal, where the initial dynamic pressure response signal includes a vibration signal and a response signal;

S102. Perform a preprocessing operation on the vibration signal and the response signal based on the variational mode decomposition method and the empirical mode decomposition method to obtain the denoised vibration signal, the preprocessing response signal, and the components of the preprocessing response signal in different frequency bands Signal;

S103. Construct a training set according to the correlation between the component signals of different frequency bands and the denoised vibration signal and the preprocessed response signal;

S104. Construct an initial inverse sensor network model based on the Bi-LSTM neural network model, iteratively train the initial inverse sensor network model based on the training set, and obtain a target inverse sensor network model;

S105. Acquire the real-time dynamic pressure response signal and input the target inverse sensor network model after performing the preprocessing operation to obtain the target shock tube dynamic pressure reconstruction signal.

Compared with the prior art, the shock tube dynamic pressure reconstruction method provided by the embodiment of the present invention, on the one hand, utilizes the variational mode decomposition method and the empirical mode decomposition method to decompose and reconstruct the response signal and the vibration signal, And comprehensively consider the correlation between each component signal and the response signal and the vibration signal, and consider the fluctuation of the actual dynamic pressure amplitude generated by the shock tube with time and the influence of the vibration signal, so that the data obtained by preprocessing has a higher Reasonability and accuracy. On the other hand, the inverse sensor network model based on the Bi-LSTM neural network reconstructs the dynamic pressure of the shock tube by using bidirectional input, which makes the signal feature extraction of the model more detailed and further improves The rationality and accuracy of shock tube reconstruction of dynamic pressure are confirmed.

Further, in some embodiments of the present invention, in step S101, when the initial response signal of the shock tube sensor is obtained, since the incident shock wave reaches the end face of the low-pressure chamber during the working process of the shock tube, shock vibration will be generated, and the sensor installed on the end face The pressure sensor above will be excited by the dynamic pressure signal and the vibration signal at the same time. The collected output signal of the pressure sensor is a mixed signal of the dynamic pressure response and the vibration response. The acceleration sensor can be used to collect the vibration signal separately, so the initial response signal includes Vibration signal and response signal.

It should be noted that, in step S105, the real-time response signal of the shock tube sensor is acquired and preprocessed by using the empirical mode decomposition method and the variational mode decomposition method in step S102 to preprocess the real-time response signal.

Further, in some embodiments of the present invention, as shown in FIG. 2, FIG. 2 is a schematic flowchart of an embodiment of step S102 in FIG. 1 provided by the present invention, and step S102 includes:

S201. Decompose the vibration signal based on the variational mode decomposition method to obtain multiple vibration modal components, respectively calculate the correlation coefficients between the multiple vibration modal components and the vibration signal, remove high-frequency noise components, and obtain the obtained The denoised vibration signal;

S202. Decompose the response signal based on a variational mode decomposition method to obtain multiple response modal components, and reconstruct the multiple response modal components based on the sensor ringing frequency to obtain multiple reconstructed signals;

S203. Decompose the multiple reconstructed signals based on an empirical mode decomposition method to obtain multiple reconstructed signal eigenmode function components;

S204. Calculate the correlation coefficients between the plurality of reconstructed signal eigenmode function components and the denoised vibration signal respectively, wherein the reconstruction corresponding to the reconstructed signal eigenmode function component with the largest correlation coefficient of the denoised vibration signal The signal is the preprocessing response signal;

S205. Decompose the preprocessing response signal based on an empirical mode decomposition method to obtain component signals of the preprocessing response signal in different frequency bands.

Further, as shown in FIG. 3, FIG. 3 is a schematic flowchart of an embodiment of step S103 in FIG. 1 provided by the present invention, and step S103 includes:

S301. Calculate the correlation coefficients between the component signals of different frequency bands and the preprocessing response signal and the denoising vibration signal respectively, and calculate the component signal with the largest correlation coefficient with the preprocessing response signal among the component signals of different frequency bands as a ringing component signal;

S301. Among the component signals of different frequency bands, except for the ringing frequency component, the component signal whose correlation coefficient value with the denoising vibration signal is smaller than the set threshold is taken as the noise component signal, and the noise component signal is eliminated to obtain the denoising response signal, reconstructed response signal, vibration-related component signal, and trend signal;

S301. Construct the training set based on the reconstructed response signal, the vibration-related component signal and the trend component signal.

Wherein, the trend component signal corresponds to the amplitude of the dynamic pressure reconstruction signal of the target shock tube, and can reflect the amplitude characteristic of the dynamic pressure reconstruction signal of the target shock tube.

进行分解，VMD可将振动信号

分解为k个窄带bimfs，表示为

，其中心频率为

，

和

可通过求解变分问题得到：In a specific embodiment of the present invention, first use the variational mode decomposition method to analyze the vibration signal

Decomposed, VMD can convert the vibration signal

Decomposed into k narrowband bimfs, denoted as

, whose center frequency is

,

and

It can be obtained by solving the variational problem:

(1)

(1)

(2)

(2)

表示求偏导，

表示狄拉克分布函数，

分别对应分解后第k个模态分量和中心频率。in,

Indicates partial derivation,

represents the Dirac distribution function,

Corresponding to the kth modal component and center frequency after decomposition, respectively.

及其对应的中心频率

，如下：The quadratic penalty function and Lagrange operator are used to solve the above constrained optimization problem, and the alternating direction method of Lagrange operator is used to calculate the decomposition mode

and its corresponding center frequency

,as follows:

(3)

(3)

(4)

(4)

分别对应

的傅里叶变换。In the formula,

Corresponding respectively

The Fourier transform of .

可表示为：The obtained vibration mode components

Can be expressed as:

(5)

(5)

表示k个模态分量。In the formula,

Denotes k modal components.

与各振动模态分量

之间的相关系数CC，如下：Get calculated vibration signal

and each vibration mode component

The correlation coefficient CC between is as follows:

(6)

(6)

和

表示振动信号和第i个模态分量对应的离散信号；

和

分别表示

和

的平均值；N为信号长度。In the formula,

and

Represents the vibration signal and the discrete signal corresponding to the i-th modal component;

and

Respectively

and

The average value of; N is the signal length.

，可表示为：According to the meaning of the correlation coefficient, take the component with the correlation coefficient less than 0.2 as the noise component signal, and remove it to obtain the denoised vibration signal

, which can be expressed as:

(7)

(7)

为高频噪声分量。In the formula,

is the high-frequency noise component.

分解为K个响应模态分量

，首先将中心频率小于等于传感器振铃频率的几个分量进行重构，得到基信号

；剩余分量依据中心频率大小依次加入到基信号中进行信号重构，基信号

以及重构信号

表达式为：Further, using the variational mode decomposition method, the response signal

Decomposed into K response modal components

, first reconstruct several components whose center frequency is less than or equal to the ringing frequency of the sensor to obtain the base signal

; The remaining components are sequentially added to the base signal according to the size of the center frequency for signal reconstruction, and the base signal

and reconstruct the signal

The expression is:

(8)

(8)

(9)

(9)

；

；

代表第J个重构信号。In the formula,

;

;

represents the Jth reconstructed signal.

The empirical mode decomposition method is used to process the above-mentioned reconstructed signals in turn to obtain a series of narrow-band components, which are called intrinsic mode functions.

Among them, each eigenmode function must meet the following two conditions: (1) the number of extreme points and zero-crossing points is equal or differs by at most one in the entire data set; (2) at any time, the number of local maximum points The estimated upper envelope and the estimated lower envelope of the local minimum points have a mean of zero.

The basic steps of decomposition are as follows:

的局部极小值点和局部极大值点；Step (1): Identify Reconstructed Signals

The local minimum and local maximum points of ;

的下包络线

和上包络线

，计算上包络线和下包络线的均值为：Step (2): Use cubic spline curves to connect all local minimum points and local maximum points respectively to obtain

lower envelope of

and upper envelope

, calculate the mean of the upper and lower envelopes as:

(10)

(10)

中减去

，得到差值信号为：Step (3): From the signal

Subtract from

, the difference signal obtained is:

(11)

(11)

满足本征模态函数的两个条件，则

为

的第一个本征模态函数分量；否则，令

，重复步骤(1)至步骤(3)计算过程k次，直到得到的

满足本征模态函数的两个条件，此时

的第一个本征模态函数分量

为：if

Satisfying the two conditions of the eigenmode function, then

for

The first eigenmode function component of ; otherwise, let

, repeat the calculation process from step (1) to step (3) k times until the obtained

Satisfy the two conditions of the eigenmode function, at this time

The first eigenmode function component of

for:

(12)

(12)

中减去

，得到残余信号

为Step (4): Reconstruct the signal from

Subtract from

, to get the residual signal

for

(13)

(13)

，重复步骤(1)到步骤(4)计算过程i次，得到第i个本征模态函数分量为：make

, repeat the calculation process from step (1) to step (4) i times, and get the i-th eigenmode function component as:

(14)

(14)

成为单调函数或者只包含一个极值点，此时从

中无法再分解出本征模态函数分量。综合式(13)和式(14)，重构信号

表示为:Continue the above decomposition process until the final residual component

becomes a monotonic function or contains only one extreme point, at this time from

It is no longer possible to decompose the eigenmode function components in . Synthesize (13) and (14), reconstruct the signal

Expressed as:

(15)

(15)

被分解为个h个重构信号的本征模态函数分量IMF和一个残余分量，并且这些分量的频段由高到低变化。Therefore, the reconstructed signal

It is decomposed into an intrinsic mode function component IMF of h reconstructed signals and a residual component, and the frequency bands of these components change from high to low.

与去噪振动信号

之间的相关系数，找到其中的最大相关系数，其对应的重构信号即为预处理信号

。如下：Calculate all component signals obtained by the empirical mode decomposition method

with denoised vibration signal

The correlation coefficient between them, find the maximum correlation coefficient among them, and the corresponding reconstructed signal is the preprocessing signal

. as follows:

(16)

(16)

，

为

分解的本征模态分量

为趋势分量。In the formula,

,

for

Decomposed eigenmode components

is the trend component.

和

；将

中相关系数最大的分量信号作为振铃分量信号，并为最大保留振动信号成分，将

中除振铃频率分量外，数值小于0.1的分量信号作为噪声分量信号，予以剔除，得到压力传感器重构响应信号

和去噪响应信号

。如下：Based on the idea of retaining the vibration signal components to the greatest extent, the correlation coefficient between each imf(t) component signal and the preprocessing signal and the denoising vibration signal is calculated separately, which is denoted as

and

;Will

The component signal with the largest correlation coefficient is used as the ringing component signal, and the vibration signal component is reserved for the maximum, and the

In addition to the ringing frequency component, the component signal whose value is less than 0.1 is used as the noise component signal and eliminated to obtain the reconstructed response signal of the pressure sensor

and the denoised response signal

. as follows:

(17)

(17)

(18)

(18)

为振动相关分量信号，

，其中

。In the formula,

is the vibration-related component signal,

,in

.

In the embodiment of the present invention, the response signal and the vibration signal are decomposed and reconstructed by using the variational mode decomposition method and the empirical mode decomposition method, and the correlation between each component signal and the response signal and the vibration signal is considered comprehensively. The fluctuation of the actual dynamic pressure amplitude generated by the shock tube with time and the influence of the vibration signal make the data obtained by preprocessing more reasonable and accurate

Further, in some embodiments of the present invention, as shown in FIG. 4, FIG. 4 is a schematic flowchart of an embodiment of step S104 in FIG. 1 provided by the present invention, and step S103 includes:

S401. Construct the first training set input of the initial inverse sensor network model based on the reconstructed response signal, construct the second training set input of the initial inverse sensor network model based on the vibration-related components, and initially inversely propagate the initial inverse sensor network model based on the trend component signal Sense network model training set output;

S402. After setting the number of hidden layers and hyperparameters of the initial inverse sensor network model, iteratively train the initial inverse sensor network model based on the constructed training set;

S403. Obtain the target inverse sensor network model when the output loss rate of the initial inverse sensor network model is lower than a set loss threshold.

Wherein, step S402 specifically includes:

The number of initial hidden layers of the initial inverse sensor network model is set, and the hidden layers include several neuron units;

Setting the hyperparameters of the initial inverse sensor network model, the hyperparameters include: optimizer parameters, learning rate, sequence length and training rounds;

In the iterative training process, the initial hidden layer number, optimizer parameters, learning rate, sequence length and training rounds of the initial inverse sensor network model are adjusted, and the number of each neuron unit node inside the hidden layer is determined. Weights and offsets, so that the output of the initial inverse sensor network model minimizes the root mean square error of the output and reaches the set loss threshold, and completes the iterative training process of the initial inverse sensor network model.

In the specific embodiment of the present invention, the initial response signal of the shock tube is subjected to the variational mode decomposition method and the empirical mode decomposition method to obtain the relevant data required for building the inverse sensor network model training set, and the reconstruction of the pressure sensor The first training set input of the inverse sensor network is constructed by the response signal, the second training set input of the inverse sensor network is constructed by the vibration related component, and the output of the inverse sensor network training set is constructed by the trend component signal.

Among them, the establishment of the initial inverse sensor network model based on the Bi-LSTM neural network mainly includes three parts: setting the number of hidden layers, setting hyperparameters, and learning and training.

The hidden layer performs the internal information transmission function of the neural network, that is, the information passed in from the input layer is processed by multiple hidden layers to obtain the output layer; the hidden layer is composed of several neural units, and a working neural unit consists of a forget gate, an input gate, a temporary Cell state, cell state, output gate and hidden layer state. Among them, the forget gate determines how much of the cell unit state at the previous moment is retained to the current moment, and the input gate determines how much of the network input is saved to the cell unit state at the current moment. The temporary cell state and the input gate work together to update the cell unit state, and the cell state The update of the state of the next cell is provided, and the output gate cooperates with the state of the hidden layer and the state of the cell to provide an update of the state of the hidden layer of the next cell unit. It works as follows:

(19)

(19)

(20)

(20)

(21)

(twenty one)

(22)

(twenty two)

(23)

(twenty three)

(24)

(twenty four)

、

、

、

、

、

分别为遗忘门、输入门、临时细胞状态、细胞状态、输出门和隐层状态，

为权重、

为偏置。In the formula:

,

,

,

,

,

They are forget gate, input gate, temporary cell state, cell state, output gate and hidden layer state,

for the weight,

for the bias.

Hyperparameter setting is to set the optimizer, learning rate, sequence length, training rounds, etc.;

Neural network learning is to determine the weights and biases of each neural unit node in the hidden layer based on the training samples. Neural network training is to find the appropriate weights and biases to minimize the output root mean square error; when the model output loss rate is low At a given loss threshold, the identification of the inverse sensor network model is completed, and the target inverse sensor network model is obtained.

It should be noted that when establishing the inverse sensor network model, adjust the hyperparameters in the inverse sensor network, including the number of hidden unit layers, learning rate, learning decline rate, optimizer and training times, etc., so that the model output loss rate is lower than The set loss threshold is used to complete the identification of the inverse sensor network model.

The embodiment of the present invention builds an inverse sensor network model based on a bidirectional long-short-term memory neural network, and performs iterative training through the training set to determine the number of hidden layers and hyperparameters of the neural network, so that the loss rate of the model output is reduced, and the shock wave The accuracy and rationality of pipe reconstruction dynamic pressure are further improved.

Further, in some embodiments of the present invention, step S105 includes:

Obtaining a real-time response signal of the shock tube sensor, performing the preprocessing operation on the real-time response signal to obtain the denoising vibration signal and denoising response signal corresponding to the real-time response signal;

Input the denoising vibration signal and denoising response signal corresponding to the real-time response signal into the target inverse sensor network model to obtain the output of the target inverse sensor network model, and divide the output by the pressure sensor magnification and sensitivity to obtain The dynamic pressure reconstruction signal of the target shock tube.

In a specific embodiment of the present invention, for the response signal of the shock tube sensor obtained in real time, the same empirical mode decomposition method and variational mode decomposition method in the above embodiment are used to preprocess the real-time response signal, and the obtained The denoising response signal and the denoising vibration signal are used as the input of the target inverse sensor network model, and the output of the model divided by the magnification and sensitivity of the pressure sensor is the dynamic pressure reconstruction signal of the target shock tube. Among them, the dynamic pressure reconstruction signal The relationship between and the output of the inverse sensor network model is:

(24)

(twenty four)

为放大系数，S为灵敏度。In the formula,

is the amplification factor, and S is the sensitivity.

The inverse sensor network model constructed based on the Bi-LSTM neural network in the embodiment of the present invention reconstructs the dynamic pressure of the shock tube by using bidirectional input, so that the model extracts signal features more carefully, and further improves the reconstruction of the shock tube. Reasonability and accuracy of dynamic pressure.

In order to more intuitively reflect the rationality and accuracy of the shock tube dynamic pressure reconstruction method provided by the present invention, the dynamic pressure response data and The vibration signal data is analyzed. The sensitivity of the pressure sensor is 0.16V/MPa, the magnification factor is 50, and the sampling frequency of the data is 5MHz. Dynamic pressure reconstruction is performed:

The raw measurement data of the ENDEVCO 8510B PR pressure sensor and the YA1102 ICP type acceleration sensor are shown in Figure 5, which is a schematic diagram of an embodiment of the sensor raw measurement data provided by the present invention.

Utilize the same empirical mode decomposition method and variational mode decomposition method in the above-mentioned embodiment to process the response signal and the vibration signal in Fig. 5, obtain the inverse sensor network training set as shown in Fig. 6, Fig. 6 is the present invention A schematic diagram of an embodiment of constructing training set data is provided.

The training result of the target inverse sensor network model obtained through the iterative training of the training set in FIG. 6 is shown in FIG. 7 , which is a schematic diagram of an embodiment of the model training result provided by the present invention.

The dynamic pressure reconstruction signal obtained by using the target inverse sensor network model obtained in FIG. 7 is shown in FIG. 8 , which is a schematic diagram of an embodiment of the dynamic pressure reconstruction signal provided by the present invention.

It can be seen from the above-mentioned Figures 5 to 8 that the shock tube dynamic pressure reconstruction signal obtained by the shock tube dynamic pressure reconstruction method provided by the embodiment of the present invention takes into account the influence of the vibration signal and the dynamic pressure of the shock tube at the same time. The characteristic of the pressure amplitude changing with time makes the dynamic pressure reconstruction signal of the shock tube more reasonable and accurate to restore the dynamic pressure signal generated by the shock tube.

In order to better implement the shock tube dynamic pressure reconstruction method in the embodiment of the present invention, on the basis of the shock tube dynamic pressure reconstruction method, correspondingly, the embodiment of the present invention also provides a shock tube dynamic pressure The reconstruction device, as shown in Figure 9, the shock tube dynamic pressure reconstruction device 900 includes:

A signal acquisition module 901, configured to acquire an initial dynamic pressure response signal, where the initial dynamic pressure response signal includes a vibration signal and a response signal;

The signal processing module 902 is configured to perform preprocessing operations on the vibration signal and the response signal based on the variational mode decomposition method and the empirical mode decomposition method, to obtain a denoised vibration signal, a preprocessing response signal, and the preprocessing response signal Component signals in different frequency bands;

A signal construction module 903, configured to construct a training set according to the correlation between the component signals of the different frequency bands and the denoising vibration signal and the preprocessing response signal;

The model training module 904 is used to construct an initial inverse sensor network model based on the Bi-LSTM neural network model, iteratively trains the initial inverse sensor network model based on the training set, and obtains a target inverse sensor network model;

The target reconstruction module 905 is used to obtain the real-time dynamic pressure response signal and input the target inverse sensor network model after performing the preprocessing operation to obtain the target shock tube dynamic pressure reconstruction signal.

The shock tube dynamic pressure reconstruction device 900 provided in the above embodiment can realize the technical solution described in the above embodiment of the shock tube dynamic pressure reconstruction method. Corresponding content in the reconstruction method embodiment will not be repeated here.

As shown in FIG. 10 , the present invention also provides an electronic device 1000 correspondingly. The electronic device 1000 includes a processor 1001 , a memory 1002 and a display 1003 . FIG. 10 shows only some components of the electronic device 1000, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.

In some embodiments, the processor 1001 may be a central processing unit (Central Processing Unit, CPU), a microprocessor or other data processing chips, which are used to run program codes or process data stored in the memory 1002, such as the excitation in the present invention Wave tube dynamic pressure reconstruction procedure.

In some embodiments, processor 1001 may be a single server or a group of servers. Server groups can be centralized or distributed. In some embodiments, processor 1001 may be local or remote. In some embodiments, the processor 1001 can be implemented on a cloud platform. In an embodiment, the cloud platform may include private cloud, public cloud, hybrid cloud, community cloud, distributed cloud, inter-internal, multi-cloud, etc., or any combination thereof.

The storage 1002 may be an internal storage unit of the electronic device 1000 in some embodiments, such as a hard disk or a memory of the electronic device 1000 . The memory 1002 may also be an external storage device of the electronic device 1000 in other embodiments, such as a plug-in hard disk equipped on the electronic device 1000, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, flash memory card (Flash Card), etc.

Further, the memory 1002 may also include both an internal storage unit of the electronic device 1000 and an external storage device. The memory 1002 is used to store application software and various data installed in the electronic device 1000 .

In some embodiments, the display 1003 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode, organic light-emitting diode) touch panel, and the like. The display 1003 is used for displaying information on the electronic device 1000 and for displaying a visualized user interface. The components 1001-1003 of the electronic device 1000 communicate with each other through a system bus.

In one embodiment, when the processor 1001 executes the shock tube dynamic pressure reconstruction program in the memory 1002, the following steps can be implemented:

acquiring an initial dynamic pressure response signal, the initial dynamic pressure response signal including a vibration signal and a response signal;

Preprocessing the vibration signal and the response signal based on the variational mode decomposition method and the empirical mode decomposition method to obtain the denoised vibration signal, the preprocessing response signal, and the component signals of the preprocessing response signal in different frequency bands;

Constructing a training set according to the correlation between the component signals of the different frequency bands and the denoising vibration signal and the preprocessing response signal;

Constructing an initial inverse sensor network model based on the Bi-LSTM neural network model, and iteratively training the initial inverse sensor network model based on the training set to obtain a target inverse sensor network model;

The real-time dynamic pressure response signal is obtained and input into the target inverse sensor network model after performing the preprocessing operation to obtain the dynamic pressure reconstruction signal of the target shock tube.

It should be understood that, when the processor 1001 executes the shock tube dynamic pressure reconstruction program in the memory 1002, in addition to the above functions, other functions may also be implemented. For details, please refer to the descriptions of the corresponding method embodiments above.

Further, the embodiment of the present invention does not specifically limit the type of the mentioned electronic device 1000, the electronic device 1000 may be a mobile phone, a tablet computer, a personal digital assistant (personal digital assistant, PDA), a wearable device, a laptop computer (laptop) and other portable electronic devices. Exemplary embodiments of portable electronic devices include, but are not limited to, portable electronic devices running IOS, android, microsoft, or other operating systems. The aforementioned portable electronic device may also be other portable electronic devices, such as a laptop computer (laptop) with a touch-sensitive surface (such as a touch panel). It should also be understood that, in some other embodiments of the present invention, the electronic device 1000 may not be a portable electronic device, but a desktop computer with a touch-sensitive surface (such as a touch panel).

Correspondingly, the embodiments of the present application also provide a computer-readable storage medium, which is used to store computer-readable programs or instructions, and when the programs or instructions are executed by a processor, the above-mentioned method embodiments can be implemented. Steps or functions in the shock tube dynamic pressure reconstruction method provided.

Those skilled in the art can understand that to realize all or part of the process of the method in the above embodiments, it can be completed by instructing related hardware (such as a processor, a controller, etc.) through a computer program, and the computer program can be stored in a computer-readable storage medium . Wherein, the computer-readable storage medium is a magnetic disk, an optical disk, a read-only memory or a random access memory, and the like.

The shock tube dynamic pressure reconstruction method, device, electronic equipment and storage medium provided by the present invention have been introduced in detail above. In this paper, specific examples have been used to illustrate the principle and implementation of the present invention. The description of the above embodiments It is only used to help understand the method of the present invention and its core idea; at the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation and scope of application. In summary, this The content of the description should not be construed as limiting the present invention.

Claims (10)

1. A shock tube dynamic pressure reconstruction method is characterized by comprising the following steps:

acquiring an initial dynamic pressure response signal, wherein the initial dynamic pressure response signal comprises a vibration signal and a response signal;

preprocessing the vibration signal and the response signal based on a variational modal decomposition method and an empirical modal decomposition method to obtain a de-noised vibration signal, a preprocessed response signal and component signals of the preprocessed response signal in different frequency bands;

constructing a training set according to the correlation between the component signals of different frequency bands and the denoising vibration signal and the preprocessing response signal respectively;

constructing an initial inverse sensing network model based on a Bi-LSTM neural network model, and iteratively training the initial inverse sensing network model based on the training set to obtain a target inverse sensing network model;

and acquiring a real-time dynamic pressure response signal, performing the preprocessing operation, and inputting the signal into the target inverse sensing network model to obtain a dynamic pressure reconstruction signal of the target shock tube.

2. The shock tube dynamic pressure reconstruction method according to claim 1, wherein the preprocessing operation is performed on the vibration signal and the response signal based on a variational modal decomposition method and an empirical modal decomposition method to obtain a de-noised vibration signal, a preprocessed response signal and component signals of the preprocessed response signal in different frequency bands, and the method comprises:

decomposing the vibration signal based on a variational modal decomposition method to obtain a plurality of vibration modal components, respectively calculating correlation coefficients of the plurality of vibration modal components and the vibration signal, and removing a high-frequency noise component to obtain the de-noising vibration signal;

decomposing the response signals based on a variational modal decomposition method to obtain a plurality of response modal components, and reconstructing the plurality of response modal components based on the sensor ringing frequency to obtain a plurality of reconstruction signals;

decomposing the plurality of reconstruction signals based on an empirical mode decomposition method to obtain a plurality of reconstruction signal intrinsic mode function components;

calculating correlation coefficients of the eigenmode function components of the plurality of reconstruction signals and the de-noising vibration signal respectively, wherein the reconstruction signal corresponding to the eigenmode function component of the reconstruction signal with the maximum correlation coefficient of the de-noising vibration signal is the preprocessing response signal;

and decomposing the preprocessing response signal based on an empirical mode decomposition method to obtain component signals of the preprocessing response signal in different frequency bands.

3. The shock tube dynamic pressure reconstruction method according to claim 2, wherein the constructing a training set according to the correlations between the component signals of the different frequency bands and the de-noising vibration signal and the pre-processing response signal respectively comprises:

respectively calculating correlation coefficients among the component signals of different frequency bands, the preprocessing response signal and the denoising vibration signal, and taking the component signal with the maximum correlation coefficient with the preprocessing response signal in the component signals of different frequency bands as a ringing component signal;

taking the component signals, except the ringing frequency component, of the component signals of different frequency bands, of which the correlation coefficient value with the denoising vibration signal is smaller than a set threshold value, as noise component signals, and removing the noise component signals to obtain denoising response signals, reconstruction response signals, vibration correlation component signals and trend signals;

constructing the training set based on the reconstructed response signal, the vibration-related component signal, and the trend component signal.

4. The shock tube dynamic pressure reconstruction method according to claim 3, wherein the trend component signal corresponds to the target shock tube dynamic pressure reconstruction signal amplitude.

5. The shock tube dynamic pressure reconstruction method according to claim 3, wherein the iteratively training an initial inverse sensor network model based on the training set to obtain a target inverse sensor network model comprises:

constructing a first training set input of an initial inverse sensing network model based on the reconstruction response signal, constructing a second training set input of the initial inverse sensing network model based on the vibration related component, and outputting the initial inverse sensing network model training set based on the trend component signal;

after the number of hidden layers and the hyper-parameters of the initial inverse sensor network model are set, carrying out iterative training on the initial inverse sensor network model based on a constructed training set;

and when the output loss rate of the initial inverse sensor network model is lower than a set loss threshold value, obtaining the target inverse sensor network model.

6. The shock tube dynamic pressure reconstruction method according to claim 5, wherein after the number of hidden layers and the hyper-parameter of the initial inverse sensor network model are set, iterative training is performed on the initial inverse sensor network model based on a constructed training set, and the iterative training comprises:

setting the number of initial hidden layer layers of the initial inverse sensor network model, wherein the hidden layers comprise a plurality of neuron units;

setting hyper-parameters of the initial inverse sensor network model, wherein the hyper-parameters comprise: optimizer parameters, learning rate, sequence length and training round;

in the iterative training process, the number of initial hidden layer layers, the parameters of an optimizer, the learning rate, the sequence length and the training round of the initial inverse sensor network model are adjusted, and the weight and the bias of each neuron unit node in the hidden layer are determined, so that the output of the initial inverse sensor network model enables the output root mean square error to be minimum and reaches the set loss threshold, and the iterative training process of the initial inverse sensor network model is completed.

7. The shock tube dynamic pressure reconstruction method according to claim 6, wherein the obtaining of the dynamic pressure real-time response signal and the preprocessing operation are performed before inputting the dynamic pressure real-time response signal into the target inverse sensor network model to obtain a dynamic pressure reconstruction signal of the target shock tube comprises:

acquiring a real-time dynamic pressure response signal of the shock tube sensor, and performing the preprocessing operation on the real-time dynamic pressure response signal to obtain a denoising vibration signal and a denoising response signal corresponding to the real-time dynamic pressure response signal;

and inputting the de-noising vibration signal and the de-noising response signal corresponding to the real-time dynamic pressure response signal into the target inverse sensing network model to obtain target inverse sensing network model output, and dividing the output by the amplification factor and the sensitivity of the pressure sensor to obtain a dynamic pressure reconstruction signal of the target shock tube.

8. A shock tube dynamic pressure reconstruction device, comprising:

the signal acquisition module is used for acquiring a dynamic pressure initial response signal, wherein the initial dynamic pressure response signal comprises a vibration signal and a response signal;

the signal processing module is used for preprocessing the vibration signal and the response signal based on a variational modal decomposition method and an empirical modal decomposition method to obtain a de-noised vibration signal, a preprocessed response signal and component signals of the preprocessed response signal in different frequency bands;

the signal construction module is used for constructing a training set according to the correlations between the component signals of the different frequency bands and the denoising vibration signal and the preprocessing response signal respectively;

the model training module is used for constructing an initial inverse sensor network model based on the Bi-LSTM neural network model, and iteratively training the initial inverse sensor network model based on the training set to obtain a target inverse sensor network model;

and the target reconstruction module is used for acquiring a real-time dynamic pressure response signal, performing the preprocessing operation and inputting the signal into the target inverse sensing network model to obtain a dynamic pressure reconstruction signal of the target shock tube.

9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor, when executing the program, implements the shock tube dynamic pressure reconstruction method according to any one of claims 1 to 7.

10. A computer-readable storage medium on which a computer program is stored, which, when being executed by a processor, carries out a shock tube dynamic pressure reconstruction method according to any one of claims 1 to 7.

Images