Signal optimization method, device and medium based on swarm intelligence and transient extraction

Abstract

The application relates to the field of signal processing and discloses a signal optimization method and device based on swarm intelligence and transient extraction and a medium, the method comprising the following steps: acquiring a non-stationary signal of a device; adopting a short-time fractional Fourier transform algorithm TEPSTFrFT based on transient extraction thought coupling to process the non-stationary signal, so as to obtain a final signal optimization result; and completing device fault diagnosis by using the final signal optimization result. Compared with a traditional STFrFT algorithm, the TEPSTFrFT algorithm improves the precision of fractional transformation, improves the aggregation degree of signal time-frequency distribution, and can improve the precision of device fault signal feature anomaly detection.

Description

Technical Field

[0001] This invention relates to the field of signal processing, and in particular to a signal optimization method, device, and medium based on swarm intelligence and transient extraction. Background Technology

[0002] In the field of signal processing, traditional two-dimensional traversal search time-frequency analysis methods have long suffered from problems such as an imbalance between accuracy and computational cost, poor time-frequency distribution clustering, and weak noise resistance. With the increasing demands for real-time performance and accuracy in signal analysis from scenarios such as industrial equipment monitoring and communication, traditional methods are struggling to meet these requirements.

[0003] For the analysis of non-stationary signals, time-frequency analysis (TFA) is the most efficient method.

[0004] TFA can transform a one-dimensional time-domain signal waveform into a two-dimensional time-frequency (TF) plane, providing joint distribution data for the time and frequency domains. This method can clearly depict the relationship between frequency and time and show the local variations of non-stationary signals. The clustering and energy concentration of the obtained time-frequency distribution are considered core performance parameters of a specific TFA technique. High-quality TFR can accurately extract instantaneous characteristics from unstable signals.

[0005] Classic time-frequency analysis methods, represented by the Fourier Transform (FT), cannot obtain specific and accurate time-frequency domain feature information, and cannot take into account the analysis of both the time and frequency domains.

[0006] Transient signals with a wide frequency range and short duration are defined as pulse-like signals. These unstable signals are common in many fields, such as vibration signals caused by mechanical failures, guided waves for structural health monitoring, and ultrasonic signals for distance measurement.

[0007] Accurate description of the instantaneous characteristics of recorded signals is crucial for accurately assessing structural health, measuring distances, and detecting mechanical faults. However, due to the influence of nonlinear systems or media, recorded signals often exhibit nonlinear frequency-varying behavior. Therefore, analyzing pulse-like signals becomes particularly challenging in practical applications. Exploring efficient signal processing methods is essential for effectively handling these signals. However, in evaluating multi-component signals, high-quality time-frequency distributions (TFRs) cannot be provided due to Heisenberg's uncertainty principle or unforeseen cross-factors. Consequently, traditional TFA methods cannot effectively represent the instantaneous properties of pulse-like signals, hindering the technical challenges of using these signals for equipment diagnosis and evaluation. Summary of the Invention

[0008] The purpose of this invention is to propose a signal optimization method, device, and medium based on swarm intelligence and transient extraction, which solves the technical problem that the existing traditional TFA method cannot effectively display the instantaneous properties of non-stationary signals, resulting in incomplete signal analysis results and making it difficult to use the corresponding signals for equipment diagnosis and evaluation.

[0009] Specifically, this invention provides a signal optimization method based on swarm intelligence and transient extraction, comprising the following steps:

[0010] S1. Acquire non-stationary signals from the device;

[0011] S2. The non-stationary signal is processed using the TEPSTFrFT algorithm, which is based on the transient extraction concept and coupled with the short-time fractional Fourier transform, to obtain the final signal optimization result.

[0012] S3. Use the final signal optimization results to complete equipment fault diagnosis.

[0013] Step S2 specifically includes:

[0014] S21. The fractional Fourier transform (PSFrFT) method optimized by particle swarm optimization is used to process non-stationary signals and obtain PSFrFT results.

[0015] S22. Using the idea of ​​transient extraction, the expression of the PSTFrFT algorithm is derived to obtain a new algorithm TEPSTFrFT based on PSTFrFT. The new algorithm is used for signal processing to obtain the final signal optimization result.

[0016] Furthermore, step S21 is as follows:

[0017] S211. The Particle Swarm Optimization (PSO) algorithm is used to process the non-stationary signal to obtain the optimal order of the non-stationary signal.

[0018] S212. Rotate the non-stationary signal to the optimal order and truncate it using different window functions to obtain the truncated signal.

[0019] S213. Perform a Fourier transform on the truncated signal to obtain the PSFrFT result.

[0020] Further, step S211 is as follows: the position dimension of the particle swarm algorithm (PSO) is set to 2, the first element is the signal order, the second element is the window length, the peak amplitude of the time-frequency distribution (TFR) of the fractional Fourier transform (FrFT) is used as the fitness function, and the optimal order of the non-stationary signal is obtained by using the particle swarm algorithm.

[0021] Furthermore, step S22 is as follows:

[0022] S221. The expression of the PSTFrFT algorithm is derived using the idea of ​​transient extraction, and a TEO operator based on PSTFrFT is constructed.

[0023] S222. Multiply the TEO operator with the PSFrFT result to obtain the final signal optimization result.

[0024] Further, in step S221, the expression for the TEO operator is as follows:

[0025]

[0026]

[0027] in, Represents the Dirac function, t For time points, For frequency points; G tg ( t , ω ) is a signal t · g ( t The result of STFT performed using a window function; G ( t , ω ) is a signal g ( t The result of STFT performed using a window function; t 0 is G ( t , ω The moment corresponding to the maximum amplitude.

[0028] Furthermore, in step S222, the final signal optimization result is as follows: .

[0029] A storage device that stores instructions and data for implementing a signal optimization method based on swarm intelligence and transient extraction.

[0030] A signal optimization device based on swarm intelligence and transient extraction includes: a processor and a storage device; the processor loads and executes instructions and data in the storage device to implement a signal optimization method based on swarm intelligence and transient extraction.

[0031] The beneficial effects provided by this invention are:

[0032] 1. Significantly improves the accuracy of time-frequency analysis

[0033] By optimizing the fractional Fourier transform (FrFT) using the particle swarm optimization (PSO) algorithm, and adaptively searching for the optimal order and window length, the problem of high computational cost and low accuracy (only 0.1 order accuracy) of traditional traversal search is solved, resulting in higher clustering of the time-frequency distribution (TFR).

[0034] 2. Enhance time-frequency distribution clustering and noise resistance

[0035] By introducing transient extraction technology (TET) and constructing a synchronous extraction operator (TEO), the coefficients of the time-frequency ridge position are accurately extracted, and noise interference is filtered out.

[0036] The TEO operator locates the peak energy of transient signals using the Dirac function. Combined with PSFrFT, it forms the TEPSTFrFT algorithm, further concentrating time-frequency energy and reducing spectral leakage.

[0037] 3. Improve the accuracy of fault diagnosis

[0038] It is suitable for high-noise scenarios such as mechanical vibration and ultrasonic ranging, providing a reliable basis for equipment health monitoring.

[0039] 4. Significantly reduces computational complexity

[0040] PSO optimizes the order and window length simultaneously (two-dimensional vector search), avoiding the full-order traversal of traditional algorithms, reducing computational load, and supporting real-time signal processing. Attached Figure Description

[0041] Figure 1 This is a simplified schematic diagram of the method flow of the present invention;

[0042] Figure 2 This is the flowchart of the particle swarm optimization algorithm;

[0043] Figure 3 This is the flowchart of the TEPSTFrFT algorithm;

[0044] Figure 4This is a schematic diagram illustrating the processing results of the three algorithms: STFrFT, PSTFrFT, and TEPSTFrFT.

[0045] Figure 5 This is a schematic diagram of the processing results of the bearing outer ring fault signal optimization algorithm;

[0046] Figure 6 This is a schematic diagram of the hardware device used in this application. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0048] Before formally describing this invention, a general overview of the invention's solution will be provided for ease of understanding. First, the relevant technical terms will be explained uniformly:

[0049] 1. Time-frequency analysis (TFA): A highly efficient method for analyzing non-stationary signals, TFA transforms a one-dimensional time-domain signal waveform into a two-dimensional time-frequency (TF) plane, providing joint distribution data for both the time and frequency domains. This method clearly depicts the relationship between frequency and time and reveals the local variations of non-stationary signals. The clustering and energy concentration of the obtained time-frequency distribution are considered core performance parameters of specific TFA techniques. High-quality TFA can accurately extract instantaneous characteristics from unstable signals.

[0050] 2. Fourier Transform (FT): A mathematical operation with wide applications in mathematics, physics, computer science, engineering, and other fields. It is a generalization of Fourier series and describes the amplitudes of sine or cosine functions of different frequencies contained within a function. Within a specific function space, the Fourier transform of a function possesses desirable properties, such as the ability to convert between differentials and polynomial products, and between function convolutions and products. Therefore, it can be used to solve partial differential equations, address norm control, and prove inequalities.

[0051] 3. Time-Frequency Representation (TFR) is a view of a signal (considered as a function of time) represented in terms of time and frequency.

[0052] 4. Time-frequency distribution (TFD) refers to the analysis of the time-frequency domain provided by the time-frequency representation (TFR).

[0053] 5. Particle swarm optimization (PSO) is an evolutionary computational technique originating from studies of bird flocks' foraging behavior. The basic idea of ​​PSO is to find the optimal solution through cooperation and information sharing among individuals within the group. Its advantages lie in its simplicity and ease of implementation, and the lack of extensive parameter tuning.

[0054] Particle swarm optimization (PSO) simulates birds in a flock by designing massless particles. These particles have only two attributes: velocity and position. Velocity represents the speed of movement, and position represents the direction of movement. Each particle independently searches for the optimal solution in the search space, recording it as its current individual extreme value. This individual extreme value is shared with all other particles in the swarm. The optimal individual extreme value is then used as the current global optimal solution for the entire swarm. All particles in the swarm adjust their velocity and position based on their own current individual extreme value and the shared global optimal solution.

[0055] 6. Transient Extraction Transform (TET): This is a new time-frequency analysis (TFA) post-processing technique. Its main principle is to establish a synchronous extraction operator (TEO) based on STFT to extract the TF coefficients at the ridge positions in the TF distribution of STFT, thereby increasing the TF concentration.

[0056] 7. The short-time Fourier transform (STFT) is a mathematical transformation of the Fourier transform (FT) correlation, used to determine the frequency and phase of the sinusoidal wave in the local region of a time-varying signal.

[0057] 8. Fractional Fourier Transform (FrFT): A linear transformation that generalizes the continuous Fourier transform. It can be thought of as the nth power of the Fourier transform and is used in filter design, signal analysis, phase retrieval, and pattern recognition.

[0058] Please refer to Figure 1 The present invention provides a signal optimization method based on swarm intelligence and transient extraction, comprising the following steps:

[0059] S1. Acquire non-stationary signals from the device;

[0060] It should be noted that the non-stationary signal in the device of this invention can be a bearing vibration signal, a ranging ultrasonic signal, or a non-stationary signal from other related devices; this is only for illustrative purposes and is not intended to be particularly limiting. Pulse-like signals, as a type of non-stationary signal, are transient signals with a wide frequency range and short duration. Analyzing pulse-like signals becomes particularly difficult in practical applications. Therefore, exploring efficient signal processing methods is crucial for effectively processing these signals.

[0061] S2. The non-stationary signal is processed using the TEPSTFrFT algorithm, which is based on the transient extraction concept and coupled with the short-time fractional Fourier transform, to obtain the final signal optimization result.

[0062] It should be noted that step S2 specifically includes:

[0063] S21. The fractional Fourier transform (PSFrFT) method optimized by particle swarm optimization is used to process non-stationary signals and obtain PSFrFT results.

[0064] Step S21 is as follows:

[0065] S211. The Particle Swarm Optimization (PSO) algorithm is used to process the non-stationary signal to obtain the optimal order of the non-stationary signal.

[0066] It should be noted that the process of traditional algorithms finding the optimal order is essentially a process of determining the order. p The algorithm iterates from 0 to 2 with a step size of 0.1, calculates the FrFT at each order, and obtains the optimal order value during the traversal by comparison. The advantage of this algorithm is that it is based on a strong principle and the results are highly reliable. However, it also has significant disadvantages, namely, it involves a large amount of computation, is time-consuming, and the precision of the optimal order is only 0.1. To obtain higher precision, more computation and time are required.

[0067] To address this issue, this invention, based on time-frequency analysis, utilizes the particle swarm optimization algorithm from swarm intelligence algorithms to optimize the process of obtaining the optimal order, thereby reducing the computational load and obtaining a more accurate optimal order, thus enabling the time-frequency distribution to exhibit better aggregation.

[0068] Please refer to Figure 2 , Figure 2 This is a flowchart of the particle swarm optimization algorithm.

[0069] PSO is initialized with a swarm of random particles (random solutions), and then the optimal solution is found through iteration. In each iteration, the particles update themselves by tracking two "extreme values" (pbest, gbest). After finding these two optimal values, the particles update their velocity and position using equations (1) and (2).

[0070] (1)

[0071] (2)

[0072] in, i =1,2,..., N N is the total number of particles; v i For particle velocity, rand () is a random number between (0,1). x i The current particle position. c 1. c 2 is the learning factor; the first part of equation (1) is called the memory term, which represents the influence of the magnitude and direction of the previous velocity; the second part of equation (1) is called the self-cognition term, which is a vector pointing from the current point to the particle's own best point, representing the part of the particle's action derived from its own experience; the third part of equation (1) is called the group cognition term, which is a vector pointing from the current point to the best point of the population, reflecting the cooperation and knowledge sharing among particles. The particle determines its next movement based on its own experience and the best experience among its peers. Based on the above two formulas, the standard form of PSO is formed.

[0073] (3)

[0074] As shown in equation (3), a non-negative inertia factor is introduced. ω A larger value indicates strong global optimization ability but weak local optimization ability; a smaller value indicates weak global optimization ability but strong local optimization ability. (Dynamic) ω It can achieve better optimization results than a fixed value. Dynamic ω The weight can change linearly during the PSO search process, or it can change dynamically according to a certain performance metric of PSO. Currently, the most commonly used strategy is the Linearly Decreasing Weight (LDW) strategy.

[0075] (4)

[0076] in G k The maximum number of iterations, ω ini , ω end These are the initial inertia weights and the inertia weights at the maximum number of iterations, respectively. ω With the introduction of PSO, the performance of the PSO algorithm has been greatly improved. It can adjust the global and local search capabilities for different search problems, and has also enabled the PSO algorithm to be successfully applied to many practical problems.

[0077] It should be noted that in this invention, the dimension of the position in the particle swarm optimization algorithm is set to 2, making it a two-dimensional vector. The first element is the order, and the second element is the window length. The peak amplitude of the TFR of FrFT or STFrFT is set as the fitness function. The mathematical model of the swarm intelligence algorithm is used to find the parameters corresponding to the maximum amplitude at each order and window length to obtain the final optimal order and optimal window length. This invention names this optimization algorithm the PSTFrFT algorithm.

[0078] S212. Rotate the non-stationary signal to the optimal order and truncate it using different window functions to obtain the truncated signal.

[0079] S213. Perform a Fourier transform on the truncated signal to obtain the PSFrFT result.

[0080] S22. Using the idea of ​​transient extraction, the expression of the PSTFrFT algorithm is derived to obtain a new algorithm TEPSTFrFT based on PSTFrFT. The new algorithm is used for signal processing to obtain the final signal optimization result.

[0081] It should be noted that, in order to further improve the clustering of time-frequency distribution, enhance noise resistance, reduce Rayleigh entropy, and obtain a better time-frequency distribution TFR, this invention further proposes a short-time fractional Fourier transform algorithm TEPSTFrFT based on the transient extraction idea.

[0082] Transient extraction (TET) technology is a new TFA post-processing technology. Its main principle is to establish a synchronous extraction operator (TEO) based on STFT to extract the TF coefficients at the ridge positions in the TF distribution of STFT and improve the TF concentration.

[0083] The expression for STFT is given by equation (5):

[0084] (5)

[0085] in g ( u - t () is a sliding window. s ( u () is a signal. In the time domain, the Dirac function δ ( t The Dirac signal should have an optimal time position because it only appears at one point in time. Therefore, the Dirac function can be regarded as an ideal model of a signal with transient characteristics. Typically, the Dirac signal can be expressed as equation (6):

[0086] (6)

[0087] For the STFT method, even for a simple Dirac signal, an ideal description cannot be achieved. To explore the TF energy distribution of the STFT results more specifically, substituting equation (6) into equation (5) yields equation (7):

[0088] (7)

[0089] Because of | e -iωt0 |=0, therefore the energy distribution of the Dirac function STFT result can be expressed as equation (8):

[0090] (8)

[0091] From equation (8), it can be seen that, due to the window function g (·) is compact in the time domain, therefore | g ( t , ω The energy distribution is concentrated in t = t It reaches its maximum value at time 0. A · g (0). Another point to note is that the STFT of a Dirac function is composed of a series of Dirac functions with the same group delay, where GD is equal to 0. To accurately estimate the GD of each Dirac function, the following should be calculated first: G ( t , ω The derivative of the frequency variable. The result is given by equation (9):

[0092] (9)

[0093] Therefore, for any ( t , ω )make G ( t , ω )≠0, its STFT result has a two-dimensional group delay t 0( t , ω It can be expressed as equation (10):

[0094] (10)

[0095] For the TF representation of an ideal Dirac signal, energy should only appear in time. t Instead of spreading to a large region, the TF coefficients are reduced to 0. This prompts the removal of the TF coefficients from the remaining regions, retaining only those with a time of 0. t The TF coefficient when it is 0.

[0096] To achieve this goal, a post-processing procedure called the Transient Extraction Operator (TEO) is established.

[0097] (11)

[0098] TEO is a two-dimensional binary representation, only when... t = t Time 0 is 1, so it can be used to extract... g ( t , ω The TF coefficient at that moment. Therefore, the transient extraction TET can be expressed as equation (12):

[0099] (12)

[0100] In discrete data processing, partial derivatives are usually approximated by finite difference operators, but more accurate parameter estimation can be achieved through equation (13):

[0101] (13)

[0102] in G tg ( t , ω ) is a signal t · g ( t The result of STFT is the window function. Substituting equation (13) into equation (11) yields equation (14):

[0103] (14)

[0104] Therefore, equation (12) can be written as equation (15):

[0105] (15)

[0106] The TET algorithm can be used for further optimization of the TF distribution in PSTFrFT. Figure 3 The main steps and workflow of the TEPSTFrFT algorithm are presented, which can explain the TEPSTFrFT algorithm more intuitively, as follows:

[0107] (1) Calculate the optimal order of the signal using the proposed swarm intelligence optimization algorithm;

[0108] (2) Calculate the optimal order below each g ( t )and t · g ( t The PSTF-rFT distribution of the window function. G ( t , ω )and G tg ( t , ω );

[0109] (3) Calculate the instantaneous group delay estimate of the signal according to equation (16). t 0( t , ω ):

[0110] (16)

[0111] (4) Establish the TEO operator as shown in equation (17):

[0112] (17)

[0113] in ε This is an adjustable precision parameter.

[0114] (5) The TEPSTFrFT expression is obtained as shown in equation (18):

[0115] (18)

[0116] As one embodiment, this invention uses the traditional STFrFT algorithm, the improved PSTFrFT algorithm, and the TEPSTFrFT algorithm to process four types of non-stationary signals. Specific information such as the expression of the test function is shown in Table 1 below:

[0117] Table 1 Test Signal Information Table

[0118]

[0119] The frequency curves of each signal and the processed time-frequency distribution results are as follows: Figure 4 As shown.

[0120] from Figure 4 It can be seen that all algorithms perform well in processing different types of non-stationary signals. Among them, compared to the traditional STFrFT algorithm, the improved PSTFrFT algorithm has a smaller Rayleigh entropy and a more concentrated time-frequency clustering. The TEPSTFrFT algorithm proposed in this invention has a significantly improved time-frequency clustering compared to the STFrFT algorithm. For the four non-stationary signals, the Rayleigh entropy of the TEPSTFrFT algorithm proposed in this invention is consistently the lowest, indicating that its clustering performance is superior to other proposed new algorithms and classic algorithms.

[0121] In summary, the PSTFrFT and TEPSTFrFT algorithms proposed in this invention improve the accuracy of fractional-order transforms and enhance the concentration of signal time-frequency distribution compared to the traditional STFrFT algorithm. Therefore, the optimization method of this invention is effective.

[0122] S3. Use the final signal optimization results to complete equipment fault diagnosis.

[0123] Taking bearing fault analysis as an example, high-concentration time-frequency slices are extracted, and the peak intervals that appear periodically are calculated to complete the quantitative calculation of the fault center frequency.

[0124] In practical engineering applications, signals often contain interference from unexpected sources. This invention employs PSTFrFT and TEPSTFrFT algorithms to process bearing fault signals, verifying the algorithms' focus and anti-interference capabilities. Because the moving parts of the bearing repeatedly pass through the defect location, a series of pulses are periodically generated. The time interval between two consecutive pulses is determined by both the rotational speed and the fault type. Considering that pulses often have a wide bandwidth, there should exist a frequency point where the TF amplitude is most significant; this frequency point can be used to describe the interval between pulses. Vibration signals from the Bearing Data Center at Case Western Reserve University are used to test the effectiveness of the fault signal anomaly detection method proposed in this invention.

[0125] The outer ring fault signal of the bearing was selected for testing. The outer ring fault data came from the bearing accelerometer data on the drive end of the bearing test bench. The bearing tested was a 6205-2RS JEM SKF. During the test, the tolerance diameter was 0.007, the motor speed and sampling rate were set to 1750 rpm and 12 kHz, respectively. According to the bearing parameters provided by the manufacturer, the fault characteristic frequency (FCF) at the current shaft speed was 104.19 Hz, and the time interval between two consecutive pulses was 9.59 ms. Its time-domain waveform, time-frequency distribution under various algorithms, and high-concentration time-frequency slice are shown below. Figure 5 As shown.

[0126] exist Figure 5 In the diagram, the yellow areas represent the periodic fault points of the bearing vibration. While the PSTFrFT algorithm clearly shows the periodic faults in the bearing, the insufficient time-frequency clustering and resolution result in a blurred TF distribution map, making it impossible to accurately locate the center frequency of the fault. Furthermore, the high-density time-frequency slice image on the right also shows significant spectral leakage in the TF slice image, hindering precise location. In contrast, the TEPSTFrFT algorithm provides clearer and more concentrated results compared to the PSTFrFT algorithm, effectively filtering out noise and significantly improving noise immunity.

[0127] Based on the data plotted in the TEPSTFrFT high-concentration time-frequency slice, the time interval between the two consecutive bearing fault pulses was calculated to be 9.508 ms, the calculated fault center frequency was 105.17 Hz, and the analysis error was 0.94%.

[0128] The results show that the optimized algorithm TEPSTFrFT proposed in this invention can provide effective analysis for the detection of abnormal bearing fault characteristics.

[0129] Please see Figure 6 , Figure 6 This is a schematic diagram of the hardware device in operation according to an embodiment of the present invention. The hardware device specifically includes: a signal optimization device 401 based on swarm intelligence and transient extraction, a processor 402, and a storage device 403.

[0130] A signal optimization device 401 based on swarm intelligence and transient extraction: The signal optimization device 401 based on swarm intelligence and transient extraction implements the signal optimization method based on swarm intelligence and transient extraction.

[0131] Processor 402: The processor 402 loads and executes the instructions and data in the storage device 403 to implement the signal optimization method based on swarm intelligence and transient extraction.

[0132] Storage device 403: The storage device 403 stores instructions and data; the storage device 403 is used to implement the signal optimization method based on swarm intelligence and transient extraction.

[0133] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A swarm intelligence and transient extraction based signal optimization method, characterized in that: The method comprises the following steps: S1, acquiring a non-stationary signal of a device; S2, processing the non-stationary signal by using a TEPSTFrFT algorithm based on transient extraction and coupling short-time fractional Fourier transform to obtain a final signal optimization result; S3, completing device fault diagnosis by using the final signal optimization result; Step S2 specifically comprises: S21, processing the non-stationary signal by using a PSFrFT method optimized by a particle swarm algorithm to obtain a PSFrFT result; Step S21 specifically comprises the following steps: S211, processing the non-stationary signal by using a particle swarm algorithm PSO to obtain an optimal order of the non-stationary signal; S212, rotating the non-stationary signal to the optimal order and truncating the non-stationary signal by using different window functions to obtain a truncated signal; S213, performing Fourier transform on the truncated signal to obtain the PSFrFT result; S22, deducing a new algorithm TEPSTFrFT based on the PSTFrFT by using the idea of transient extraction to the expression of the PSTFrFT algorithm, and using the new algorithm to process the signal to obtain a final signal optimization result; Step S22 specifically comprises the following steps: S221, deducing the TEO operator based on the PSTFrFT by using the idea of transient extraction to the expression of the PSTFrFT algorithm; S222, multiplying the TEO operator and the PSFrFT result to obtain the final signal optimization result.

2. The signal optimization method based on swarm intelligence and transient extraction as claimed in claim 1, wherein: Step S211 specifically comprises the following steps: setting the position dimension of the particle swarm algorithm PSO as 2, setting the first element as the signal order and the second element as the window length, taking the time-frequency distribution TFR amplitude peak value of the fractional Fourier transform FrFT as the adaptive function, and obtaining the optimal order of the non-stationary signal by using the particle swarm algorithm.

3. The signal optimization method based on swarm intelligence and transient extraction as claimed in claim 1, wherein: In step S221, the expression of the TEO operator is as follows: wherein, denotes the Dirac function, t is a time point, is a frequency point; G tg ( t , ω ) is the STFT result of the signal with t · g ( t ) as the window function; G ( t , ω ) is the STFT result of the signal with g ( t ) as the window function; t 0 is G ( t , ω ) the time point corresponding to the maximum amplitude.

4. The swarm intelligence and transient extraction based signal optimization method of claim 3, wherein: In step S222, the final signal optimization result is as follows: .

5. A storage device, characterized by: The storage device stores instructions and data for implementing the signal optimization method based on swarm intelligence and transient extraction according to any one of claims 1-4.

6. A signal optimization device based on swarm intelligence and transient extraction, characterized by: It comprises: A processor and a storage device; the processor loads and executes the instructions and data in the storage device to implement the signal optimization method based on swarm intelligence and transient extraction according to any one of claims 1-4.

Images