CN113589686B - GSA-IFCM-based unit cycle time sequence self-adaptive extraction method - Google Patents

Abstract

The present invention discloses an adaptive extraction method of unit period time series based on GSA-IFCM, which includes steps S1. During the operation of the naval gun bomb supplying mechanism, the sensor is used to first obtain the original vibration signal of the naval gun bomb supplying mechanism; S2. Use the GSA‑IFCM algorithm to cluster the maximum values of the envelope of the original vibration signal and extract the center point of the unit period time series in the original vibration signal; S3. After obtaining the center point of the unit period time series in the original vibration signal , use t‑MSV to process the center point, obtain the length of the unit period time series in the forward and backward direction of the center point, find the final wave bottom by setting a threshold to determine the start and end positions of the unit period time series, and then extract the unit period time series; It has been verified that this method can effectively overcome the influence of data noise, improve the accuracy of clustering results, and accurately extract unit period time series.

Description

Adaptive extraction method of unit period time series based on GSA-IFCM

Technical field

The invention relates to the technical field of mechanical equipment health status detection methods, and specifically relates to an adaptive extraction method of unit period time series based on GSA-IFCM.

Background technique

The ammunition supply and delivery mechanism of medium and large-caliber naval guns usually performs cyclic reciprocating motion during operation. The process is complex and often accompanied by severe impact, friction and vibration. Its working health has always restricted the performance and practicality of the ammunition supply and delivery mechanism. Obstacles have also become the focus in the field of mechanical equipment condition monitoring and fault diagnosis; the prerequisite for assessing its health status is the collection of condition monitoring data. In a strict mathematical sense, monitoring data are non-periodic and non-stationary random signals, but from a macro perspective Considering the working principle and process, during continuous operation, the monitoring data shows obvious approximate periodicity. Therefore, the vibration signal of the bomb supply mechanism is a typical approximate periodic signal, and the approximate periodic signal is a special approximate periodic time series. ;

Currently, regarding the problem of period estimation and extraction of approximate periodic time series, Wu Shujin et al. analyzed the approximate periodic time series for the first time, gave a detailed mathematical definition, and proposed the use of moment estimation method to estimate the period of the time series; Hong Shuhui et al. used The fitting estimation method extracts two-dimensional data that can reflect time transformation from existing time series data, and more truly reflects the periodic changes of approximately periodic time series; the above method is better for approximate periodic time series with a small change range, and The ammunition feeding mechanism reciprocates, the working rhythm is inconsistent, and the instantaneous impact is large. The signal amplitude is transient and has a large change range, which makes the above method unable to effectively extract the unit period time series. Chen Renxiang et al. used the phase difference spectrum correction-cross-correlation method to The approximately periodic vibration signal of the harmonic reducer is segmented to construct unit period data samples to accurately depict the operating status information of the reducer. However, this method only targets vibration signals whose envelope is approximately a sinusoidal curve, and the approximately periodic impact signal cannot pass through Compute and efficiently construct envelopes;

At present, there is little research on adaptive extraction methods for approximate periodic time series extraction. Most of them require manual interception of unit period time series, which relies on expert experience and is inefficient.

Contents of the invention

In view of the above existing problems, the present invention aims to provide an adaptive extraction method of unit period time series based on GSA-IFCM. During the operation of the medium and large caliber naval gun supply and delivery mechanism, this method first uses improved fuzzy C-means The clustering algorithm is combined with the genetic simulated annealing algorithm to extract the cluster center of the maximum value of the approximate periodic signal; then the time window energy method is used to extract the unit period time series. This method can effectively overcome the influence of data noise and improve Accuracy of clustering results, accurate extraction of unit period time series.

In order to achieve the above objects, the technical solutions adopted by the present invention are as follows:

Adaptive extraction method of unit period time series based on GSA-IFCM, including steps

S1. During the operation of the naval gun's bomb supply and transport mechanism, the sensor is first used to obtain the original vibration signal of the naval gun's bomb supply and transport mechanism;

S2. Use the GSA-IFCM algorithm to perform clustering processing on the maximum values of the envelope of the original vibration signal, and extract the center point of the unit period time series in the original vibration signal;

S3. After obtaining the center point of the unit period time series in the original vibration signal, use t-MSV to process the center point, obtain the length of the unit period time series in the forward and backward directions of the center point, and find the final wave bottom by setting the threshold to determine the unit. The starting and ending positions of the periodic time series are then extracted to the unit period time series.

Preferably, the process of obtaining the unit period time series center point in step S2 includes:

S201. Use the envelope function in MATLAB software to calculate the effective envelope of the original vibration signal;

S202. Extract the maximum value of the vibration signal envelope, and then obtain the effective maximum value point of the original vibration signal;

S203. Use the horizontal and vertical coordinates of the maximum value point as the input vector, and set the number of categories at the same time. Iteratively calculate and continuously correct the clustering center through the GSA-IFCM algorithm until the optimal clustering center is obtained, thereby obtaining the unit period time series. center point.

Preferably, the process of obtaining the unit period time series center point in step S203 includes:

S2031. Set _the number of clusters c and _{the time sample data} of _length n category;

S2032. Initialize the genetic annealing algorithm control parameters: population size sizepop, maximum number of evolutions MAXGEN, crossover probability P, mutation probability P _m , annealing initial temperature T ₀ , temperature cooling coefficient k, termination temperature T _end ;

S2033. Initialize the weighting parameter b and set the weighting parameter step size b _s ;

S2034. Randomly initialize c clustering centers and generate the initial population Po. For each clustering center, use equation (3) to calculate the membership degree of each sample and the fitness value _fi of each individual, where i=1, 2,…,sizepop;

S2035. Set the loop count variable gen=0;

S2036. Implement selection, crossover and mutation genetic operations on the population Po, and use formula (3) and formula (4) to calculate c cluster centers, the membership degree of each sample, and the fitness value f of each individual for the newly generated individuals. _i ', and the variance value Var of the abscissa of each cluster center; if f _i '>f _i , replace the old individual with the new individual, otherwise, use probability P=exp(( _{fi -fi} ')T) Accept the new individual and discard the old;

S2037. Check whether gen reaches the maximum number of evolutions MAXGEN; if gen<MAXGEN, then gen=gen+1, go to S2038; otherwise, go to S2036;

S2038. Detect whether T reaches the end temperature T _end ; if _Ti <T _end , go to S2039; otherwise, perform the cooling operation Ti ₊₁ = kT _i , go to S2035;

S2039. Detect whether the variance value Var reaches the termination variance Var _end ; if Var < Var _end , the algorithm ends successfully and returns the global optimal solution; otherwise, the weighted parameters increase the step size b=b+b _s and go to S2034.

Preferably, the calculation process of the cluster center in step S2034 includes:

(1) The fitness value is obtained by using the ranking fitness distribution function: Fint V=ranking(J _b ). The smaller J _b is, the higher the individual fitness is. The objective function J _b is:

Among them, in formula (1): d _ik is the Euclidean distance, used to measure the distance between the i-th sample x _i and the k-th category center point; m is the feature number of the sample; b is a weighting parameter, the value range is 1≤b≤∞; u _k ( _xi ) is the membership degree of sample x _i to class A _k (abbreviated as u _ik ), and its calculation formula is:

Among them, in formula (2), the sum of the membership values of a sample for each cluster is 1, that is

(2) Assume I _k ={i|2≤c≤n;d _ik =0}, for all i categories, i∈I _k , u _ik =0; then the iteration equation for updating the cluster center v _ij is:

Preferably, the specific process of using t-MSV to process the center point and extracting the unit period time series in step S3 includes:

S301. Set a time window with a smaller width, use the time window width as the step size, and use the cluster center as the starting point to circularly calculate the energy value of the time series in the window. The calculation formula of the parameter mean square value that represents the signal energy value is:

^Among them, in equation (5): X _i (t) is the energy value of a time series of length N, and _Y In the shock and oscillation stage, when the mean square value decreases monotonically, it means that the energy of the signal is declining. When the mean square value stabilizes at a certain value, it means that the energy of the signal tends to be stable. Therefore, the wave bottom corresponding to each wave peak is the starting point of the impact signal. start and end positions;

S302. Calculate the time corresponding to the wave bottom of the time window energy curve, and set the corresponding threshold. If the energy in the time window fails to detect an increase in energy within the threshold, it is determined that this wave bottom is the final wave bottom, and the time corresponding to this wave bottom is is the optimal starting position of the periodic time series, and its calculation formula is:

x _f =max(x _m )+x _h (6)

s.t.f(x+ε)′≤0 (7)

Among them, in equations (6) and (7), x _m is the time corresponding to the signal peak, x _h is the wave width of the wave, and x _f is the time corresponding to the final wave bottom in the forward direction;

S303. Obtaining the sample length backward is the same as step S302, obtaining the time corresponding to the final wave bottom in the backward direction, determining the start and end positions of the unit period sequence, and then extracting the start and end sequence of the unit period.

The beneficial effects of the present invention are: the present invention discloses an adaptive extraction method of unit period time series based on GSA-IFCM. Compared with the existing technology, the improvements of the present invention are:

(1) Aiming at the problem that it is difficult to extract the unit period time series representing the approximately periodic signal during the reciprocating motion of the bomb feeding mechanism, the present invention proposes an adaptive extraction method of unit period time series based on GSA-IFCM. This method first adopts The improved fuzzy C-means clustering algorithm is combined with the genetic simulated annealing algorithm to extract the cluster center of the maximum value of the approximate periodic signal; then the time window energy method is used to extract the unit period time series;

(2) This method uses the combination of GSA and IFCM to effectively overcome the impact of data noise. The clustering results have stronger stability and better objective function values. It is effective for clustering randomly distributed data containing noise. Obvious superiority;

(3) This method uses the time window to obtain the mean square value of the time domain characteristic parameters of the signal, which can represent the energy trend of the signal. Compared with other time domain characteristic parameters, it can effectively show the vibration pattern of the original signal and can accurately extract the unit period. Time series has the advantage of being able to effectively overcome the influence of data noise, improve the accuracy of clustering results, and accurately extract unit period time series.

Description of the drawings

Figure 1 is an algorithm flow chart of the adaptive extraction method of unit period time series based on GSA-IFCM of the present invention.

Figure 2 is an algorithm flow chart of the GSA-IFCM algorithm of the present invention.

Figure 3 is an algorithm flow chart of the t-MSV algorithm of the present invention.

Figure 4 is a diagram of the test bench for the bomb feeding mechanism according to Embodiment 1 of the present invention.

Figure 5 is a time domain waveform diagram of an acceleration signal in Embodiment 1 of the present invention.

Figure 6 is the original signal and envelope diagram of Embodiment 1 of the present invention.

Figure 7 is a diagram of the original signal and the maximum value point in Embodiment 1 of the present invention.

Figure 8 is a diagram of maximum value points and cluster centers in Embodiment 1 of the present invention.

Figure 9 is a windowed view of the original signal in Embodiment 1 of the present invention.

Figure 10 is an energy curve diagram of Embodiment 1 of the present invention.

Figure 11 is the unit period time series extracted in Embodiment 1 of the present invention.

Figure 12 is a time distance curve diagram of the standard FCM algorithm, the GSA-based FCM algorithm and the GSA-IFCM algorithm in Embodiment 1 of the present invention.

Figure 13 is a time domain characteristic time window curve diagram of Embodiment 1 of the present invention.

Among them: in Figure 12, Figure (a) shows the standard FCM algorithm, GSA-based FCM algorithm and GSA-IFCM algorithm-based time distance curve under normal conditions; Figure (b) shows The standard FCM algorithm, GSA-based FCM algorithm and GSA-IFCM algorithm-based time distance curve when roller cracks occur in the naval gun bomb supply mechanism; Figure (c) shows the standard when the naval gun bomb supply mechanism produces slide wear. FCM algorithm, GSA-based FCM algorithm and time distance curve chart based on GSA-IFCM algorithm.

Detailed ways

In order to enable those of ordinary skill in the art to better understand the technical solutions of the present invention, the technical solutions of the present invention are further described below in conjunction with the accompanying drawings and examples.

Referring to the adaptive extraction method of unit period time series based on GSA-IFCM shown in Figures 1-3, it is characterized by: including steps

S1. During the operation of the naval gun bomb supply and transport mechanism, install a sensor on the naval gun bomb supply and transport mechanism, and use the sensor to first obtain the original vibration signal of the naval gun bomb supply and transport mechanism;

S2. Use the GSA-IFCM algorithm to perform clustering processing on the maximum values of the envelope of the original vibration signal, and extract the center point of the unit period time series in the original vibration signal, which specifically includes:

S201. In the MATLAB software, use the envelope function in the MATLAB software to calculate the effective envelope of the original vibration signal;

S202. Extract the maximum value of the vibration signal envelope, and then obtain the effective maximum value point of the original vibration signal;

S203. Use the horizontal and vertical coordinates of the maximum value point as the input vector, and set the number of categories at the same time. Iteratively calculate and continuously correct the clustering center through the GSA-IFCM algorithm until the optimal clustering center is obtained, thereby obtaining the unit period time series. Center point, as can be seen from the above steps, the GSA-IFCM algorithm is the key to obtaining the clustering center point. Its flow chart is shown in Figure 2, which includes the steps:

S2031. Set _the number of clusters c and _{the time sample data} of _length n category;

S2032. Initialize the genetic annealing algorithm control parameters: population size sizepop, maximum number of evolutions MAXGEN, crossover probability P, mutation probability P _m , annealing initial temperature T ₀ , temperature cooling coefficient k, termination temperature T _end ;

S2033. Initialize the weighting parameter b and set the weighting parameter step size b _s ;

S2034. Randomly initialize c clustering centers and generate the initial population Po. For each clustering center, use equation (3) to calculate the membership degree of each sample and the fitness value _fi of each individual, where i=1, 2,…,sizepop; its specific process includes:

(1) The fitness value is obtained by using the ranking fitness distribution function: FintV=ranking(J _b ). The smaller J _b is, the higher the individual fitness is. The objective function J _b is:

Among them, in formula (1): d _ik is the Euclidean distance, used to measure the distance between the i-th sample x _i and the k-th category center point; m is the feature number of the sample; b is a weighting parameter, the value range is 1≤b≤∞; u _k ( _xi ) is the membership degree of sample x _i to class A _k (abbreviated as u _ik ), and its calculation formula is:

Among them, in formula (2), the sum of the membership values of a sample for each cluster is 1, that is

Assume I _k ={i|2≤c≤n;d _ik =0}, for all i categories, i∈I _k , u _ik =0; then the iterative equation for updating the cluster center v _ij is:

S2035. Set the loop count variable gen=0;

S2036. Perform genetic operations such as selection, crossover, and mutation on the population Po, and use equation (3) and equation (4) to calculate c cluster centers, the membership degree of each sample, and the fitness value of each individual for the newly generated individuals. f _i ', and the variance value Var of the abscissa of each cluster center; if f _i '>f _i , replace the old individual with the new individual, otherwise, use the probability P=exp(( _{fi -fi} ')T ) accept new individuals and abandon old ones;

S2037. Check whether gen reaches the maximum number of evolutions MAXGEN; if gen<MAXGEN, then gen=gen+1, go to S2038; otherwise, go to S2036;

S2038. Detect whether T reaches the end temperature T _end ; if _Ti <T _end , go to S2039; otherwise, perform the cooling operation Ti ₊₁ = kT _i , go to S2035;

S2039. Check whether the variance value Var reaches the end variance Var _end ; if Var < Var _end , the algorithm ends successfully and returns the global optimal solution; otherwise, the weighted parameter increases by the step size b=b+b _s and goes to S2034;

S3. After obtaining the clustering center of the maximum value of the approximate periodic signal in the original vibration signal, use t-MSV to process the center point, obtain the length of the unit period time series in the forward and backward direction of the center point, and find the final wave by setting the threshold The bottom is used to determine the starting and ending positions of the unit period time series, and then the unit period time series is extracted. The specific process includes:

S301. Set a time window with a smaller width, use the time window width as the step size, and use the cluster center as the starting point to circularly calculate the energy value of the time series in the window. The calculation formula of the parameter mean square value that represents the signal energy value is:

^Among them, in equation (5): X _i (t) is the energy value of a time series of length N, and _Y In the shock and oscillation stage, when the mean square value decreases monotonically, it means that the energy of the signal is declining. When the mean square value stabilizes at a certain value, it means that the energy of the signal tends to be stable. Therefore, the wave bottom corresponding to each wave peak is the starting point of the impact signal. start and end positions;

S302. Calculate the time corresponding to the wave bottom of the time window energy curve, and set the corresponding threshold. If the energy in the time window fails to detect an increase in energy within the threshold, it is determined that this wave bottom is the final wave bottom, and the time corresponding to this wave bottom is is the optimal starting position of the periodic time series, and its calculation formula is:

x _f =max(x _m )+x _h (6)

s.t.f(x+ε)′≤0 (7)

Among them, in equations (6) and (7), x _m is the time corresponding to the signal peak, x _h is the wave width of the wave, and x _f is the time corresponding to the final wave bottom in the forward direction;

S303. Obtaining the sample length backward is the same as step S302. Obtain the sample length in the forward and backward direction of the cluster center, that is, accurately extract the unit period time series. The calculation formula is as follows:

x＝x _f +x _b (8)

Among them, in equation (8), x _b is the time corresponding to the final wave bottom in the backward direction, and x is the unit period time series.

Embodiment 1: As shown in Figure 4-13: the algorithm of the unit period time series adaptive extraction method based on GSA-IFCM of the present invention also includes steps

S4. Test and result analysis process

S401. Test introduction

This test data is collected from a certain type of ammunition supply and delivery mechanism test bench, as shown in Figure 4;

In the test, 6 vibration acceleration sensors were arranged near the swinging machine slide of the bench device and the plate press located above the rollers. The sensor type was an ICP acceleration sensor, the sampling frequency was 10kHz, and a 32-channel LMS signal acquisition system was used. The normal operation of the samples was collected separately. There are a total of 24 sets of vibration acceleration signals in three states: roller crack and skateboard wear. Each set of signals contains 20 cyclic actions. The time domain waveforms of the vibration acceleration signals in the three states are shown in Figure 5;

It can be seen from Figure 5 that the vibration acceleration signal measured by the bomb delivery mechanism has large impact vibration, among which the impact is most significant when the roller is cracked. At the same time, only from the original vibration acceleration signal, the period of each unit period time series Neither the length nor the maximum amplitude are fixed values, and the intervals between adjacent periods are not equal, so a fixed period cycle cannot be used to extract the unit period time series;

S402. Adaptive extraction method of test bench vibration signal

The extraction method proposed by the present invention is applied to the above test data to select the vibration signal under normal working conditions. The specific steps are as follows:

(1) Calculate the effective envelope of the original vibration signal as shown in Figure 6. The dark black curve in the figure is the original vibration signal and the light black curve is the envelope;

(2) Extract the maximum value of the envelope, and then find the effective maximum point of the original signal. As shown in Figure 7, the maximum value point of the envelope in the figure is obviously smaller than the maximum value of the original signal. In this way to improve calculation speed;

(3) Use the GSA-IFCM algorithm, take the horizontal and vertical coordinates of the maximum value points as the input vector, and set the number of categories at the same time, and obtain the optimal clustering center of the maximum value points through continuous iterative calculations to obtain the unit The center point of the periodic time series, as shown in Figure 8;

(4) Through continuous testing, select a time window with a window width of 0.01s, and cycle forward from the clustering center to accumulate time windows, as shown in Figure 9;

(5) By comparing the variance and peak value of other time domain features, the mean square value is selected to characterize the energy of the time series within the time window, and the energy curve is obtained, as shown in Figure 10. The graph contains 3 peaks, which are to the left of the center point. There are three impacts corresponding to the side;

(6) Set the time window threshold k = 100. If the time window energy cannot detect an increase in energy within the threshold, calculate the time corresponding to the last peak and bottom of the time window energy curve (as shown in the 10th circle in the figure). That is, the starting position of the unit period time series;

Follow the above steps to calculate the end position of the unit period time series to the right starting from the cluster center, and then extract the entire unit period time series, as shown in Figure 11;

In the test, a complete reciprocating motion takes about 6.8 seconds, and each wave peak corresponds to the following actions: re-advance - starting swing - back swing - closing latch. This method can effectively achieve accurate reproduction of the mechanism's motion state.

S403. Test result analysis

(1) The unit period time series can be effectively extracted from the approximately periodic vibration signals collected in each state using the method of the present invention; in the experiment, the SA algorithm set the cooling coefficient q = 0.8, the initial temperature T ₀ = 100, and the end temperature T _end =1; The relevant initial parameters of the GA algorithm are shown in Table 1;

Table 1: GA algorithm related parameter table

In order to verify the effectiveness of the method described in the present invention, a group of approximate periodic vibration signal maximum values in three health states were clustered by using the standard FCM algorithm, the GSA-based FCM algorithm and the GSA-based IFCM algorithm respectively. , the time distance curve between the vibration signal cluster center and the center point of the unit time series in each state is shown in Figure 12; if the distance is less than half of the unit time series (light black straight line in Figure 12), the clustering result is considered correct ;

Observing Figure 12, we can see that the standard FCM algorithm is used to cluster large-scale data. Since it is more sensitive to the initial center selection, it is easier to converge to the local optimal solution and cannot effectively uniformly cluster the maximum values of the original signal. Class; the FCM algorithm based on GSA is used to cluster the maximum values of the three vibration signals. Although the cluster centers are approximately uniformly distributed, there is still the possibility of falling into local minimum points due to the fuzzy values of the weighting parameters, such as Figure 11 There are 2 clustering center points that are abnormal; the use of GSA-IFCM algorithm can effectively overcome the impact of data noise and accurately extract the center points of each periodic time series. The possibility of premature convergence is extremely small. The obtained clustering results It has stronger stability and better objective function value, and has obvious advantages for randomly distributed data clustering;

The test collected a total of 8 groups of vibration acceleration signals in each of three states, with 20 cyclic actions in each group, that is, 160 cyclic actions in each of the three states. The correct number and accuracy of the clustering results of each group of original signals are shown in Table 2 Show:

Table 2: Accuracy of clustering results

It can be seen from Table 2 that using three different methods, the clustering accuracy under normal conditions is higher than that for roller cracks and skateboard wear. The reason is that the noise of the vibration signal is smaller under normal conditions and has less impact on the algorithm. Roller cracks The reason why the accuracy does not reach 100% in this state is that the vibration is the most violent in this state, and the peak-to-peak value can reach 900g. However, high accuracy can still be achieved using the GSA-IFCM algorithm;

(2) In order to verify the advantage of using the mean square value of time domain characteristics to characterize the time window energy, the vibration signal in the normal state is used to compare the time window curves calculated by using other normalized time domain characteristic variances and peak values, such as As shown in Figure 13; when selecting the peak characteristic parameters, as shown in the peak curve in Figure 13, it is found that a fluctuation appears in the second peak, which cannot accurately describe the fluctuation process of the original signal, and the peak value only describes the changing range of the signal value. It is difficult to determine the starting and ending range of the signal; and by selecting the variance characteristic parameters, as shown in the variance curve in Figure 13, it can be seen that the wave peaks conform to the vibration pattern of the signal, but the distinction between the first and third wave peaks is not obvious, and the degree of discreteness of the signal is small The section description is weak; select the mean square value characteristic parameter, as shown in the mean square value curve in Figure 13. It can be seen that this feature can accurately describe the vibration process of the signal, effectively characterize its energy change pattern, and exists three times to the left of the center point. Corresponding to the impact; therefore, compared with extracting signals using other time domain characteristic parameters such as peak value and variance, using the mean square value to characterize the time window energy can accurately extract the unit period time series.

Conclusion: In order to solve the problem of difficulty in extracting the unit period time series representing the approximate periodic signal during the reciprocating motion of the bomb feeding mechanism, the present invention proposes an adaptive extraction method of unit period time series based on GSA-IFCM. Experiments show that this method can be effective. Overcome the impact of data noise, improve the accuracy of clustering results, and accurately extract unit period time series;

1) The method of combining GSA and IFCM can effectively overcome the influence of data noise. The clustering results have stronger stability and better objective function values. It is obviously superior to clustering of randomly distributed data containing noise. sex;

2) Obtaining the mean square value of the time domain characteristic parameters of the signal through the time window can represent the energy trend of the signal. Compared with other time domain characteristic parameters, it can effectively show the vibration pattern of the original signal and accurately extract the unit period time series.

The basic principles, main features and advantages of the present invention have been shown and described above. Those skilled in the industry should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions only illustrate the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will also have other aspects. Various changes and modifications are possible, which fall within the scope of the claimed invention. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims (4)

1. The adaptive extraction method of unit period time series based on GSA-IFCM is characterized by: including steps S1. During the operation of the naval gun's bomb supply and transport mechanism, the sensor is first used to obtain the original vibration signal of the naval gun's bomb supply and transport mechanism; S2. Use the GSA-IFCM algorithm to perform clustering processing on the maximum values of the envelope of the original vibration signal, and extract the center point of the unit period time series in the original vibration signal; S3. After obtaining the center point of the unit period time series in the original vibration signal, use t-MSV to process the center point, obtain the length of the unit period time series in the forward and backward directions of the center point, and find the final wave bottom by setting the threshold to determine the unit. The starting and ending positions of the periodic time series, and then extract the unit period time series; The specific process of using t-MSV to process the center point and extract the unit period time series described in step S3 includes S301. Set a time window with a smaller width, use the time window width as the step size, and use the cluster center as the starting point to circularly calculate the energy value of the time series in the window. The calculation formula of the parameter mean square value that represents the signal energy value is: ^Among them, in equation (5): X _i (t) is the energy value of a time series of length N, and _Y In the shock and oscillation stage, when the mean square value decreases monotonically, it means that the energy of the signal is declining. When the mean square value stabilizes at a certain value, it means that the energy of the signal tends to be stable. Therefore, the wave bottom corresponding to each wave peak is the starting point of the impact signal. start and end positions; S302. Calculate the time corresponding to the wave bottom of the time window energy curve, and set the corresponding threshold. If the energy in the time window fails to detect an increase in energy within the threshold, it is determined that this wave bottom is the final wave bottom, and the time corresponding to this wave bottom is is the optimal starting position of the periodic time series, and its calculation formula is: x _f =max(x _m )+x _h (6) s.t.f(x+ε)′≤0 (7) Among them, in equations (6) and (7), x _m is the time corresponding to the signal peak, x _h is the wave width of the wave, and x _f is the time corresponding to the final wave bottom in the forward direction; S303. Obtaining the sample length backward is the same as step S302, obtaining the time corresponding to the final wave bottom in the backward direction, determining the start and end positions of the unit period sequence, and then extracting the start and end sequence of the unit period. 2. The adaptive extraction method of unit period time series based on GSA-IFCM according to claim 1, characterized in that: the acquisition process of unit period time series center point in step S2 includes: S201. Use the envelope function in MATLAB software to calculate the effective envelope of the original vibration signal; S202. Extract the maximum value of the vibration signal envelope, and then obtain the effective maximum value point of the original vibration signal; S203. Use the horizontal and vertical coordinates of the maximum value point as the input vector, and set the number of categories at the same time. Iteratively calculate and continuously correct the clustering center through the GSA-IFCM algorithm until the optimal clustering center is obtained, thereby obtaining the unit period time series. center point. 3. The adaptive extraction method of unit period time series based on GSA-IFCM according to claim 2, characterized in that: the obtaining process of unit period time series center point in step S203 includes: S2031. _{Set the} number of clusters _c and _the time _sample data of _length n category; S2032. Initialize the genetic annealing algorithm control parameters: population size sizepop, maximum number of evolutions MAXGEN, crossover probability P, mutation probability P _m , annealing initial temperature T ₀ , temperature cooling coefficient k, termination temperature T _end ; S2033. Initialize the weighting parameter b and set the weighting parameter step size b _s ; S2034. Randomly initialize c clustering centers and generate the initial population Po. For each clustering center, use equation (3) to calculate the membership degree of each sample and the fitness value _fi of each individual, where i=1, 2,…,sizepop; S2035. Set the loop count variable gen=0; S2036. Implement selection, crossover and mutation genetic operations on the population Po, and use formula (3) and formula (4) to calculate c cluster centers, the membership degree of each sample, and the fitness value f of each individual for the newly generated individuals. _i ', and the variance value Var of the abscissa of each cluster center; if f _i '>f _i , replace the old individual with the new individual, otherwise, use probability P=exp(( _{fi -fi} ')T) Accept the new individual and discard the old; S2037. Check whether gen reaches the maximum number of evolutions MAXGEN; if gen<MAXGEN, then gen=gen+1, go to S2038; otherwise, go to S2036; S2038. Detect whether T reaches the end temperature T _end ; if _Ti <T _end , go to S2039; otherwise, perform the cooling operation Ti ₊₁ = kT _i , go to S2035; S2039. Check whether the variance value Var reaches the end variance Varend; if Var<Var _end , the algorithm ends successfully and returns the global optimal solution; otherwise, the weighted parameters increase the step size b=b+b _s and go to S2034. 4. The adaptive extraction method of unit period time series based on GSA-IFCM according to claim 3, characterized in that: the calculation process of the cluster center in step S2034 includes: (1) The fitness value is obtained by using the ranking fitness distribution function: FintV=ranking(J _b ). The smaller J _b is, the higher the individual fitness is. The objective function J _b is: Among them, in formula (1): d _ik is the Euclidean distance, used to measure the distance between the i-th sample x _i and the k-th category center point; m is the feature number of the sample; b is a weighting parameter, the value range is 1≤b≤∞; u _k ( _xi ) is the membership degree of sample x _i to class A _k (abbreviated as u _ik ), and its calculation formula is: Among them, in formula (2), the sum of the membership values of a sample for each cluster is 1, that is (2) Suppose I _k ={i|2≤c≤n;d _ik =0}, for all i categories, i∈I _k , u _ik =0; Then the iterative equation for updating the cluster center v _ij is: