CN113589686B - GSA-IFCM-based unit cycle time sequence self-adaptive extraction method - Google Patents
GSA-IFCM-based unit cycle time sequence self-adaptive extraction method Download PDFInfo
- Publication number
- CN113589686B CN113589686B CN202110715227.4A CN202110715227A CN113589686B CN 113589686 B CN113589686 B CN 113589686B CN 202110715227 A CN202110715227 A CN 202110715227A CN 113589686 B CN113589686 B CN 113589686B
- Authority
- CN
- China
- Prior art keywords
- time sequence
- value
- signal
- gsa
- ifcm
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
- G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention discloses a unit cycle time sequence self-adaptive extraction method based on GSA-IFCM, comprising the following steps that S1, in the running process of a gun bullet supply and delivery mechanism, a sensor is utilized to firstly obtain an original vibration signal of the gun bullet supply and delivery mechanism; s2, clustering the maximum value of the envelope curve of the original vibration signal by using a GSA-IFCM algorithm, and extracting the central point of the unit cycle time sequence in the original vibration signal; s3, after a central point of a unit period time sequence in an original vibration signal is obtained, the central point is processed by using a t-MSV, the length of the unit period time sequence is obtained in the front-back direction of the central point, a final wave bottom is found through setting a threshold value to determine the start-stop position of the unit period time sequence, and then the unit period time sequence is extracted; through verification, the method can effectively overcome the influence of data noise, improve the accuracy of clustering results and accurately extract the unit period time sequence.
Description
Technical Field
The invention relates to the technical field of a detection method of the health state of mechanical equipment, in particular to a unit cycle time sequence self-adaptive extraction method based on GSA-IFCM.
Background
The middle-large caliber cannon bullet feeding mechanism usually performs cyclic reciprocating motion in the operation process, the process is complex and is often accompanied by severe impact, friction and vibration, the working health state of the cannon bullet feeding mechanism is always a barrier for limiting the performance and the practicability of the cannon bullet feeding mechanism, and the cannon bullet feeding mechanism also becomes a focus in the field of monitoring the state of mechanical equipment and fault diagnosis; the precondition for evaluating the health state is that the state monitoring data are collected, the monitoring data are non-periodic and non-stable random signals in a strict mathematical sense, but the working principle and the process are considered from a macroscopic view, and the monitoring data show obvious approximate periodicity when continuously running, so that the vibration signal of the bullet feeding mechanism is a typical approximate periodic signal, and the approximate periodic signal is a special approximate periodic time sequence;
at present, for the problems of period estimation and extraction of the approximate period time sequence, wu Shujin and the like analyze the approximate period time sequence for the first time, give detailed mathematical definition and provide a moment estimation method for estimating the period of the time sequence; hong Shuhui and the like extract two-dimensional data capable of reflecting time transformation from the existing time series data by adopting a fitting estimation method, so that the periodic variation of the approximate periodic time series is reflected more truly; the method has good effect on the approximate periodic time sequence with small variation range, and the elastic feeding mechanism can be used for circularly reciprocating, the working beats are different and the instantaneous impact is large, and the signal amplitude is transient and the variation range is large, so that the method can not effectively extract the unit periodic time sequence; chen Renxiang and the like, a unit period data sample is constructed for dividing an approximate period vibration signal of the harmonic speed reducer by using a phase difference frequency spectrum correction-cross correlation method so as to accurately describe the running state information of the speed reducer, but the method only aims at the vibration signal with an envelope curve approximate to a sine curve, and the envelope curve cannot be effectively constructed by calculating the approximate period impact signal;
the current adaptive extraction method for extracting the approximate periodic time sequence has little research, most of the methods also need to manually intercept the unit periodic time sequence, rely on expert experience and have low efficiency.
Disclosure of Invention
Aiming at the problems, the invention aims to provide a unit period time sequence self-adaptive extraction method based on GSA-IFCM, and in the operation process of a medium-and-large-caliber cannon bullet feeding mechanism, the method firstly combines an improved fuzzy C-means clustering algorithm with a genetic simulated annealing algorithm to extract a clustering center approximate to the maximum value of a periodic signal; and then extracting the unit period time sequence by using a time window energy method, wherein the method can effectively overcome the influence of data noise, improve the accuracy of a clustering result and accurately extract the unit period time sequence.
In order to achieve the above purpose, the technical scheme adopted by the invention is as follows:
the unit period time sequence self-adaptive extraction method based on GSA-IFCM comprises the following steps of
S1, in the running process of a gun bullet supply and delivery mechanism, a sensor is utilized to firstly obtain an original vibration signal of the gun bullet supply and delivery mechanism;
s2, clustering the maximum value of the envelope curve of the original vibration signal by using a GSA-IFCM algorithm, and extracting the central point of the unit cycle time sequence in the original vibration signal;
s3, after the central point of the unit cycle time sequence in the original vibration signal is obtained, the central point is processed by using the t-MSV, the length of the unit cycle time sequence is obtained in the front-back direction of the central point, the final wave bottom is found through setting a threshold value to determine the start-stop position of the unit cycle time sequence, and then the unit cycle time sequence is extracted.
Preferably, the process of acquiring the unit cycle time series center point in step S2 includes:
s201, calculating an effective envelope curve of an original vibration signal by using an envelope function in MATLAB software;
s202, extracting a maximum value of an envelope curve of the vibration signal, and further obtaining an effective maximum value point of an original vibration signal;
s203, taking the abscissa and the ordinate of the maximum point as input vectors, setting the number of categories at the same time, and carrying out iterative computation through a GSA-IFCM algorithm to continuously correct the clustering center until an optimal clustering center is obtained, thereby obtaining a unit period time sequence center point.
Preferably, the obtaining the unit cycle time series center point in step S203 includes:
s2031, setting the number c of clusters and time sample data X= { X with the length n 1 ,x 2 ,…,x n }，{A 1 ,A 2 ,…,A c -c categories representing this time sequence;
s2032, initializing control parameters of a genetic annealing algorithm: population individual size sizepop, maximum evolution number MAXGEN, crossover probability P, mutation probability P m Initial annealing temperature T 0 Temperature cooling coefficient k, end temperature T end ；
S2033, initializing a weighting parameter b, and setting a weighting parameter step length b s ；
S2034, randomly initializing c cluster centers, generating an initial population Po, calculating membership degree of each sample for each cluster center by using a formula (3), and calculating fitness value f of each individual i Where i=1, 2, …, sizepop;
s2035. setting a cycle count variable gen=0;
s2036 performing selection, crossover and mutation genetic operation on the group Po, calculating c clustering centers, membership degree of each sample and fitness value f of each individual by using the formula (3) and the formula (4) for newly generated individuals i ' and variance value Var of each cluster center abscissa; if f i '＞f i The old individual is replaced with the new individual, otherwise the probability p=exp ((f) i -f i ') T) accepting new individuals, discarding old individuals;
s2037, detecting whether the gen reaches the maximum evolution times MAXGEN; if gen < MAXGEN, gen=gen+1, go to S2038; otherwise, go to S2036;
s2038 detecting whether T reaches the end temperature T end The method comprises the steps of carrying out a first treatment on the surface of the If T i ＜T end Then go to S2039; otherwise, executing the cooling operation T i+1 ＝kT i Go to S2035;
s2039 detecting whether the variance value Var reaches the termination variance Var end The method comprises the steps of carrying out a first treatment on the surface of the If Var < Var end The algorithm is successfully ended, and a global optimal solution is returned; otherwise, the weighting parameter is increased by a step b=b+b s Go to S2034.
Preferably, the calculation process of the cluster center in step S2034 includes:
(1) The fitness value is obtained by adopting a sequenced fitness distribution function: fint v=ranking (J b )，J b The smaller the individual, the higher the fitness of the individual, the objective function J b The method comprises the following steps:
wherein, in formula (1):d ik is Euclidean distance used to measure the ith sample x i Distance from the k-th class center point; m is the characteristic number of the sample; b is a weighting parameter, and the value range is 1-b-infinity; u (u) k (x i ) Is sample x i For class A k Membership degree of (abbreviated as u) ik ) The calculation formula is as follows:
wherein in formula (2), the sum of membership values of one sample for each cluster is 1, i.e
(2) Set I k ＝{i|2≤c≤n；d ik =0 }, I e I for all I classes k ,u ik =0; then update cluster center v ij Is:
preferably, the specific process of extracting the unit cycle time sequence by processing the center point with the t-MSV in step S3 comprises
S301, setting a time window with smaller width, and calculating the energy value of a time sequence in the window in a forward circulation mode by taking the time window width as a step length and taking a clustering center as a starting point, wherein a parameter mean square value calculation formula for representing the signal energy value is as follows:
wherein, in formula (5): x is X i (t) is the energy value of the time series with the length N, Y x 2 Is root mean square; when the mean square value monotonically increases, the energy of the signal is increased, the signal is in an impact oscillation stage, when the mean square value monotonically decreases, the energy of the signal is decreased, and when the mean square value is stabilized at a certain value, the energy of the signal is stabilized, so that the wave bottom corresponding to each wave crest is the starting position and the ending position of the impact signal;
s302, calculating the time corresponding to the wave bottom of the time window energy curve graph, setting a corresponding threshold value, judging the wave bottom as the final wave bottom if the energy of the time window fails to detect the energy rising in the threshold value, wherein the time corresponding to the wave bottom is the optimal initial position of the periodic time sequence, and the calculation formula is as follows:
x f ＝max(x m )+x h (6)
s.t.f(x+ε)′≤0 (7)
wherein in the formula (6) and the formula (7), x m For the corresponding time of the signal peak x h For the wave width of the wave, x f The time corresponding to the final wave bottom in the front direction;
s303, acquiring the sample length backwards and obtaining the time corresponding to the final wave bottom in the rear direction, determining the start and stop positions of the unit cycle sequence, and further extracting the unit cycle start and stop sequence.
The beneficial effects of the invention are as follows: the invention discloses a unit period time sequence self-adaptive extraction method based on GSA-IFCM, which is improved compared with the prior art in that:
(1) Aiming at the problem that the unit periodic time sequence of the approximate periodic signal is difficult to extract in the reciprocating motion process of the elastic feeding mechanism, the invention provides a GSA-IFCM-based unit periodic time sequence self-adaptive extraction method, which comprises the steps of firstly, combining an improved fuzzy C-means clustering algorithm with a genetic simulated annealing algorithm to extract a clustering center of the approximate periodic signal maximum value; then extracting a unit period time sequence by using a time window energy method;
(2) The method utilizes the combination method of GSA and IFCM, can effectively overcome the influence of data noise, and the clustering result has stronger stability and better objective function value, and has obvious superiority for data clustering which is randomly distributed and contains noise;
(3) According to the method, the time domain characteristic parameter mean square value of the signal is obtained through the time window, the energy trend of the signal can be represented, compared with other time domain characteristic parameters, the vibration rule of the original signal can be effectively displayed, the unit period time sequence can be accurately extracted, the influence of data noise can be effectively overcome, the accuracy of a clustering result is improved, and the unit period time sequence can be accurately extracted.
Drawings
Fig. 1 is an algorithm flow chart of the unit cycle time sequence adaptive extraction method based on the GSA-IFCM.
FIG. 2 is an algorithm flow chart of the GSA-IFCM algorithm of the invention.
FIG. 3 is an algorithm flow chart of the t-MSV algorithm of the present invention.
FIG. 4 is a diagram of a test bench for a loading mechanism according to embodiment 1 of the invention.
Fig. 5 is a time domain waveform diagram of an acceleration signal according to embodiment 1 of the present invention.
Fig. 6 is a diagram of the original signal and envelope of embodiment 1 of the present invention.
Fig. 7 is a plot of the original signal and maxima for example 1 of the present invention.
FIG. 8 is a graph of maxima points and cluster centers according to example 1 of the present invention.
Fig. 9 is a windowing diagram of an original signal in embodiment 1 of the present invention.
FIG. 10 is an energy graph of example 1 of the present invention.
Fig. 11 is a unit cycle time series extracted in example 1 of the present invention.
Fig. 12 is a graph of time distance in the state of the FCM algorithm, GSA-based FCM algorithm, and GSA-IFCM-based algorithm according to the standard of embodiment 1 of the present invention.
Fig. 13 is a graph of a time domain signature time window according to embodiment 1 of the present invention.
Wherein: in fig. 12, fig. (a) shows a standard FCM algorithm, a GSA-based FCM algorithm and a GSA-IFCM algorithm-based time distance graph in a normal state of a cannon bullet feeding mechanism; FIG. (b) shows a graph of time distance under the conditions of a standard FCM algorithm, a GSA-based FCM algorithm and a GSA-IFCM algorithm under the condition that a ship gun bullet supply mechanism generates roller cracks; and (c) shows a FCM algorithm, a GSA-based FCM algorithm and a GSA-IFCM algorithm-based time distance graph which are used for representing the standard when the ship gun bullet supply mechanism generates the abrasion of the skateboard.
Detailed Description
In order to enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention is further described below with reference to the accompanying drawings and examples.
The unit period time sequence self-adaptive extraction method based on the GSA-IFCM is shown in the accompanying figures 1-3, and is characterized in that: comprising the steps of
S1, installing a sensor on a gun feeding and conveying mechanism in the operation process of the gun feeding and conveying mechanism, and firstly obtaining an original vibration signal of the gun feeding and conveying mechanism by using the sensor;
s2, clustering the maximum value of the envelope curve of the original vibration signal by using a GSA-IFCM algorithm, and extracting the central point of the unit cycle time sequence in the original vibration signal, wherein the method specifically comprises the following steps:
s201, calculating an effective envelope curve of an original vibration signal in MATLAB software by using an envelope function in the MATLAB software;
s202, extracting a maximum value of an envelope curve of the vibration signal, and further obtaining an effective maximum value point of an original vibration signal;
s203, taking the abscissa and the ordinate of the maximum point as input vectors, setting the number of categories at the same time, and iteratively calculating and continuously correcting the clustering center through a GSA-IFCM algorithm until the optimal clustering center is obtained, thereby obtaining a unit period time sequence center point, wherein the GSA-IFCM algorithm is a key for obtaining the clustering center point, and the flow chart is shown in figure 2 and comprises the following steps:
s2031, setting the number c of clusters and time sample data X= { X with the length n 1 ,x 2 ,…,x n }，{A 1 ,A 2 ,…,A c -c categories representing this time sequence;
s2032, initializing control parameters of a genetic annealing algorithm: population individual size sizepop, maximum evolution number MAXGEN, crossover probability P, mutation probability P m Initial annealing temperature T 0 Temperature cooling coefficient k, end temperature T end ；
S2033, initializing a weighting parameter b, and setting a weighting parameter step length b s ；
S2034, randomly initializing c cluster centers, generating an initial population Po, calculating membership degree of each sample for each cluster center by using a formula (3), and calculating fitness value f of each individual i Where i=1, 2, …, sizepop; the specific process comprises the following steps:
(1) The fitness value is obtained by adopting a sequenced fitness distribution function: fintv=ranking (J b )，J b The smaller the individual, the higher the fitness of the individual, the objective function J b The method comprises the following steps:
wherein, in formula (1):d ik is Euclidean distance used to measure the ith sample x i Distance from the k-th class center point; m is the characteristic number of the sample; b is a weighting parameter, and the value range is 1-b-infinity; u (u) k (x i ) Is sample x i For class A k Membership degree of (abbreviated as u) ik ) The calculation formula is as follows:
wherein in formula (2), the sum of membership values of one sample for each cluster is 1, i.e
Set I k ＝{i|2≤c≤n；d ik =0 }, I e I for all I classes k ,u ik =0; then update cluster center v ij Is:
s2035. setting a cycle count variable gen=0;
s2036 performing genetic operations such as selection, crossover and mutation on the group Po, and calculating c clustering centers, membership degree of each sample and fitness value f of each individual for newly generated individuals by using the formula (3) and the formula (4) i ' and variance value Var of each cluster center abscissa; if f i '＞f i The old individual is replaced with the new individual, otherwise the probability p=exp ((f) i -f i ') T) accepting new individuals, discarding old individuals;
s2037, detecting whether the gen reaches the maximum evolution times MAXGEN; if gen < MAXGEN, gen=gen+1, go to S2038; otherwise, go to S2036;
s2038 detecting whether T reaches the end temperature T end The method comprises the steps of carrying out a first treatment on the surface of the If T i ＜T end Then go to S2039; otherwise, executing the cooling operation T i+1 ＝kT i Go to S2035;
s2039 detecting whether the variance value Var reaches the termination variance Var end The method comprises the steps of carrying out a first treatment on the surface of the If Var < Var end The algorithm is successfully ended, and a global optimal solution is returned; otherwise, the weighting parameter is increased by a step b=b+b s Go to S2034;
s3, after a clustering center of an approximate periodic signal maximum value in an original vibration signal is obtained, a t-MSV is utilized to process a central point, the length of a unit periodic time sequence is obtained in the front-back direction of the central point, a final wave bottom is found through setting a threshold value to determine the start-stop position of the unit periodic time sequence, and further the unit periodic time sequence is extracted, and the specific process comprises the following steps:
s301, setting a time window with smaller width, and calculating the energy value of a time sequence in the window in a forward circulation mode by taking the time window width as a step length and taking a clustering center as a starting point, wherein a parameter mean square value calculation formula for representing the signal energy value is as follows:
wherein, in formula (5): x is X i (t) is the energy value of the time series with the length N, Y x 2 Is root mean square; when the mean square value monotonically increases, the energy of the signal is increased, the signal is in an impact oscillation stage, when the mean square value monotonically decreases, the energy of the signal is decreased, and when the mean square value is stabilized at a certain value, the energy of the signal is stabilized, so that the wave bottom corresponding to each wave crest is the starting position and the ending position of the impact signal;
s302, calculating the time corresponding to the wave bottom of the time window energy curve graph, setting a corresponding threshold value, judging the wave bottom as the final wave bottom if the energy of the time window fails to detect the energy rising in the threshold value, wherein the time corresponding to the wave bottom is the optimal initial position of the periodic time sequence, and the calculation formula is as follows:
x f ＝max(x m )+x h (6)
s.t.f(x+ε)′≤0 (7)
wherein in the formula (6) and the formula (7), x m For the corresponding time of the signal peak x h For the wave width of the wave, x f The time corresponding to the final wave bottom in the front direction;
s303, acquiring a sample length backwards and synchronizing with the step S302 to acquire the sample length in the front-back direction of the clustering center, namely accurately extracting a unit period time sequence, wherein the calculation formula is as follows:
x＝x f +x b (8)
wherein in formula (8), x b And x is a unit period time sequence for the time corresponding to the final wave bottom in the rear direction.
Example 1: as shown in fig. 4-13: the algorithm of the unit period time sequence self-adaptive extraction method based on the GSA-IFCM also comprises the following steps of
S4, test and result analysis process
S401 description of the test
The test data are collected from a test bed of a certain type of bullet feeding mechanism, as shown in figure 4;
6 vibration acceleration sensors are arranged near a swinging machine sliding plate of the bench device and a pressing plate machine above the roller, the sensors are ICP acceleration sensors, the sampling frequency is 10kHz, and a 32-channel LMS signal acquisition system is adopted; respectively collecting 24 groups of vibration acceleration signals under three states of normal operation, roller crack and skateboard abrasion, wherein each group of signals comprises 20 cyclic actions, and the time domain waveforms of the vibration acceleration signals under the three states are shown in figure 5;
as can be seen from fig. 5, the vibration acceleration signal measured by the bullet feeding mechanism has a larger impact vibration, wherein the impact is most remarkable in the state of roller crack; meanwhile, from the original vibration acceleration signal, the cycle length and the maximum amplitude of each unit cycle time sequence are not fixed values, and the intervals of adjacent cycles are not equal, so that the unit cycle time sequence cannot be circularly extracted by using one fixed cycle;
s402 self-adaptive extraction method for vibration signals of test bed
The extraction method provided by the invention is applied to the test data, and vibration signals in a normal working state are selected, and the specific steps are as follows:
(1) Calculating an effective envelope curve of the original vibration signal as shown in fig. 6, wherein a dark black curve is the original vibration signal, and light black is the envelope curve;
(2) Extracting the maximum value of the envelope curve, further obtaining an effective maximum value point of the original signal, wherein the maximum value point of the envelope curve is obviously smaller than the maximum value of the original signal in the graph as shown in fig. 7, so that the calculation speed is improved;
(3) The GSA-IFCM algorithm is utilized, the abscissa and the ordinate of the maximum point are used as input vectors, the number of categories is set at the same time, and the optimal clustering center of the maximum point is obtained through continuous iterative calculation, so that the central point of the unit period time sequence is obtained, as shown in figure 8;
(4) Through continuous experiments, a time window with the window width of 0.01s is selected, and the time windows are accumulated forward circularly by a clustering center, as shown in fig. 9;
(5) Selecting a mean square value to represent the energy of the time sequence in the time window by comparing and adopting other time domain characteristic variances and peak values, and obtaining an energy graph, wherein the graph comprises 3 wave peaks corresponding to three impacts on the left side of a central point as shown in fig. 10;
(6) Setting a time window threshold k=100, and if the energy of the time window fails to detect the energy rise within the threshold, calculating the time corresponding to the bottom of the last peak of the time window energy graph (as shown in the circle of fig. 10), namely, the starting position of the unit cycle time sequence;
calculating the end position of the unit cycle time sequence to the right by taking the cluster center as a starting point according to the steps, namely extracting the whole unit cycle time sequence, as shown in fig. 11;
in the test, the complete reciprocating time is about 6.8s, and each peak corresponds to: the method can effectively realize accurate reproduction of the motion state of the mechanism.
S403, analysis of test results
(1) The approximate periodic vibration signals collected under each state can be used for effectively extracting a unit periodic time sequence by the method; in the test, the SA algorithm sets the cooling coefficient q=0.8 and the initial temperature T 0 Stop temperature t=100 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, the 1 group of approximate periodic vibration signal maxima under three health states are clustered respectively by using a standard FCM algorithm, a GSA-based FCM algorithm and a GSA-based IFCM algorithm, and a time distance graph between a vibration signal clustering center and a unit time sequence center point under each state is shown in FIG. 12; if the distance is less than half of the unit time series (light black line in fig. 12), the clustering result is considered to be correct;
as can be seen from fig. 12, the standard FCM algorithm is adopted to perform clustering processing on the large-scale data, and because the method is sensitive to initial center selection, the method is easier to converge to a local optimal solution, and can not effectively perform uniform clustering on the maximum value of the original signal; clustering three vibration signal maxima by using a GSA-based FCM algorithm, wherein the clustering centers are approximately uniformly distributed, but the probability of sinking into local minimum points is still caused by fuzzy value of the weighting parameters, for example, 2 clustering center points are abnormal in FIG. 11; the GSA-IFCM algorithm can effectively overcome the influence of data noise, the central point of each periodic time sequence is accurately extracted, the possibility of early convergence is extremely low, the obtained clustering result has stronger stability and better objective function value, and the clustering method has obvious superiority for randomly distributed data clustering;
the test collects 8 groups of vibration acceleration signals under three states in total, and each group of vibration acceleration signals is subjected to 20 cyclic actions, namely 160 cyclic actions under the three states; the correct number and accuracy of the clustering results of each group of original signals are shown in table 2:
table 2: accuracy of clustering results
As can be seen from Table 2, the clustering accuracy in the normal state is higher than that of the roller crack and the sliding plate abrasion by adopting 3 different methods, the reason is that the noise of the vibration signal in the normal state is smaller, the influence on the algorithm is smaller, the accuracy in the roller crack state is not 100%, the vibration is most intense in the state, the peak-peak value can reach 900g, but the GSA-IFCM algorithm can still reach higher accuracy;
(2) To verify the advantage of using the mean square value of the time domain feature to represent the energy of the time window, the vibration signals in the normal state are used to respectively compare the time window graphs calculated by using the variance and the peak value of other normalized time domain features, as shown in fig. 13; when the peak characteristic parameter is selected, as shown in a peak curve of fig. 13, it is found that a fluctuation occurs in the second peak, the fluctuation process of the original signal cannot be accurately described, and the peak only describes the variation range of the signal value, so that it is difficult to determine the start-stop range of the signal; the variance characteristic parameters are selected, as shown in the variance curve of fig. 13, the wave peak accords with the vibration rule of the signal, but the first wave peak and the third wave peak are not obviously distinguished, and the description of the section with small signal discrete degree is weaker; selecting a mean square value characteristic parameter, as shown in a mean square value curve in fig. 13, the characteristic can accurately describe the vibration process of the signal, effectively represents the energy change rule of the signal, and corresponds to three impacts on the left side of a central point; therefore, compared with the extraction of signals by adopting other time domain characteristic parameters such as peak value, variance and the like, the unit periodic time sequence can be accurately extracted by utilizing the mean square value to represent the time window energy.
Conclusion: aiming at the problem that the unit cycle time sequence of the approximate periodic signal is difficult to extract in the reciprocating motion process of the elastic feeding mechanism, the invention provides a GSA-IFCM-based unit cycle time sequence self-adaptive extraction method, and experiments show that the method can effectively overcome the influence of data noise, improve the accuracy of a clustering result and accurately extract the unit cycle time sequence;
1) By using the method of combining GSA and IFCM, the influence of data noise can be effectively overcome, the clustering result has stronger stability and better objective function value, and has obvious superiority for data clustering which is randomly distributed and contains noise;
2) The energy trend of the signal can be represented by obtaining the mean square value of the time domain characteristic parameter of the signal through the time window, compared with other time domain characteristic parameters, the vibration rule of the original signal can be effectively displayed, and the unit cycle time sequence can be accurately extracted.
The foregoing has shown and described the basic principles, principal features and advantages of the invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, and that the above embodiments and descriptions are merely illustrative of the principles of the present invention, and various changes and modifications may be made without departing from the spirit and scope of the invention, which is defined in the appended claims. The scope of the invention is defined by the appended claims and equivalents thereof.
Claims (4)
1. The unit period time sequence self-adaptive extraction method based on the GSA-IFCM is characterized in that: comprising the steps of
S1, in the running process of a gun bullet supply and delivery mechanism, a sensor is utilized to firstly obtain an original vibration signal of the gun bullet supply and delivery mechanism;
s2, clustering the maximum value of the envelope curve of the original vibration signal by using a GSA-IFCM algorithm, and extracting the central point of the unit cycle time sequence in the original vibration signal;
s3, after a central point of a unit period time sequence in an original vibration signal is obtained, the central point is processed by using a t-MSV, the length of the unit period time sequence is obtained in the front-back direction of the central point, a final wave bottom is found through setting a threshold value to determine the start-stop position of the unit period time sequence, and then the unit period time sequence is extracted;
the specific process of extracting the unit cycle time sequence by processing the center point by using the t-MSV in the step S3 comprises
S301, setting a time window with smaller width, and calculating the energy value of a time sequence in the window in a forward circulation mode by taking the time window width as a step length and taking a clustering center as a starting point, wherein a parameter mean square value calculation formula for representing the signal energy value is as follows:
wherein, in formula (5): x is X i (t) is the energy value of the time series with the length N, Y x 2 Is root mean square; when the mean square value monotonically increases, the energy of the signal is increased, the signal is in an impact oscillation stage, when the mean square value monotonically decreases, the energy of the signal is decreased, and when the mean square value is stabilized at a certain value, the energy of the signal is stabilized, so that the wave bottom corresponding to each wave crest is the starting position and the ending position of the impact signal;
s302, calculating the time corresponding to the wave bottom of the time window energy curve graph, setting a corresponding threshold value, judging the wave bottom as the final wave bottom if the energy of the time window fails to detect the energy rising in the threshold value, wherein the time corresponding to the wave bottom is the optimal initial position of the periodic time sequence, and the calculation formula is as follows:
x f ＝max(x m )+x h (6)
s.t.f(x+ε)′≤0 (7)
wherein in the formula (6) and the formula (7), x m For the corresponding time of the signal peak x h For the wave width of the wave, x f The time corresponding to the final wave bottom in the front direction;
s303, acquiring the sample length backwards and obtaining the time corresponding to the final wave bottom in the rear direction, determining the start and stop positions of the unit cycle sequence, and further extracting the unit cycle start and stop sequence.
2. The GSA-IFCM-based unit cycle time series adaptive extraction method of claim 1, wherein: the process for obtaining the center point of the unit cycle time series in step S2 includes:
s201, calculating an effective envelope curve of an original vibration signal by using an envelope function in MATLAB software;
s202, extracting a maximum value of an envelope curve of the vibration signal, and further obtaining an effective maximum value point of an original vibration signal;
s203, taking the abscissa and the ordinate of the maximum point as input vectors, setting the number of categories at the same time, and carrying out iterative computation through a GSA-IFCM algorithm to continuously correct the clustering center until an optimal clustering center is obtained, thereby obtaining a unit period time sequence center point.
3. The GSA-IFCM-based unit cycle time series adaptive extraction method of claim 2, wherein: the process of obtaining the center point of the unit cycle time series in step S203 includes:
s2031, setting the number c of clusters and time sample data X= { X with the length n 1 ,x 2 ,…,x n }，{A 1 ,A 2 ,…,A c -c categories representing this time sequence;
s2032, initializing control parameters of a genetic annealing algorithm: population individual size sizepop, maximum evolution number MAXGEN, crossover probability P, mutation probability P m Initial annealing temperature T 0 Temperature cooling coefficient k, end temperature T end ；
S2033, initializing a weighting parameter b, and setting a weighting parameter step length b s ；
S2034, randomly initializing c cluster centers, generating an initial population Po, calculating membership degree of each sample for each cluster center by using a formula (3), and calculating fitness value f of each individual i Where i=1, 2, …, sizepop;
s2035. setting a cycle count variable gen=0;
s2036 performing selection, crossover and mutation genetic operation on the group Po, calculating c clustering centers, membership degree of each sample and fitness value f of each individual by using the formula (3) and the formula (4) for newly generated individuals i ' and variance value Var of each cluster center abscissa; if f i '＞f i The old individual is replaced with the new individual, otherwise the probability p=exp ((f) i -f i ') T) accepting new individuals, discarding old individuals;
s2037, detecting whether the gen reaches the maximum evolution times MAXGEN; if gen < MAXGEN, gen=gen+1, go to S2038; otherwise, go to S2036;
s2038 detecting whether T reaches the end temperature T end The method comprises the steps of carrying out a first treatment on the surface of the If T i ＜T end Then go toS2039; otherwise, executing the cooling operation T i+1 ＝kT i Go to S2035;
s2039, detecting whether the variance value Var reaches a termination variance Varend; if Var < Var end The algorithm is successfully ended, and a global optimal solution is returned; otherwise, the weighting parameter is increased by a step b=b+b s Go to S2034.
4. A GSA-IFCM-based unit cycle time series adaptive extraction method of claim 3, wherein: the process of calculating the cluster center in step S2034 includes:
(1) The fitness value is obtained by adopting a sequenced fitness distribution function: fintv=ranking (J b )，J b The smaller the individual, the higher the fitness of the individual, the objective function J b The method comprises the following steps:
wherein, in formula (1):d ik is Euclidean distance used to measure the ith sample x i Distance from the k-th class center point; m is the characteristic number of the sample; b is a weighting parameter, and the value range is 1-b-infinity; u (u) k (x i ) Is sample x i For class A k Membership degree of (abbreviated as u) ik ) The calculation formula is as follows:
wherein in formula (2), the sum of membership values of one sample for each cluster is 1, i.e
(2) Set I k ＝{i|2≤c≤n；d ik =0 }, I e I for all I classes k ,u ik ＝0；
Then update cluster center v ij Is:
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110715227.4A CN113589686B (en) | 2021-06-26 | 2021-06-26 | GSA-IFCM-based unit cycle time sequence self-adaptive extraction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110715227.4A CN113589686B (en) | 2021-06-26 | 2021-06-26 | GSA-IFCM-based unit cycle time sequence self-adaptive extraction method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113589686A CN113589686A (en) | 2021-11-02 |
| CN113589686B true CN113589686B (en) | 2023-09-29 |
Family
ID=78244770
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110715227.4A Active CN113589686B (en) | 2021-06-26 | 2021-06-26 | GSA-IFCM-based unit cycle time sequence self-adaptive extraction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113589686B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115460304B (en) * | 2022-11-10 | 2023-01-31 | 广州铭创通讯科技有限公司 | Protocol layer data analysis method and system for intercepting wireless communication |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH01109401A (en) * | 1987-10-22 | 1989-04-26 | Konica Corp | Process control method using plant model |
| US5201292A (en) * | 1991-08-30 | 1993-04-13 | Loral Aerospace Corp. | Apparatus and method for detecting vibration patterns |
| AU2009213031A1 (en) * | 2009-09-10 | 2011-03-24 | Kingston, William Mr | Tideway anchor |
| CN102435554A (en) * | 2011-09-08 | 2012-05-02 | 北京师范大学 | Method for acquiring farmland multiple-cropping index |
| CN106295888A (en) * | 2016-08-12 | 2017-01-04 | 东南大学 | A kind of public building based on measured data is joined and is built parking position and share time window and determine method |
| CN106527135A (en) * | 2016-11-21 | 2017-03-22 | 云南电网有限责任公司电力科学研究院 | Building energy management system optimization control method and device |
| CN107727399A (en) * | 2017-10-27 | 2018-02-23 | 中国人民解放军海军工程大学 | Fault Diagnosis of Roller Bearings based on dual-tree complex wavelet packet stream shape domain noise reduction |
| CN110457867A (en) * | 2019-10-08 | 2019-11-15 | 杭州知衣科技有限公司 | A kind of time series data based on machine learning is filled up and restoring method |
| CN112347571A (en) * | 2020-09-18 | 2021-02-09 | 中国人民解放军海军工程大学 | Rolling bearing residual life prediction method considering model and data uncertainty |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8185252B2 (en) * | 2010-11-22 | 2012-05-22 | General Electric Company | DSM enabling of electro mechanically controlled refrigeration systems |
| EP3032367A1 (en) * | 2014-12-08 | 2016-06-15 | Kabushiki Kaisha Toshiba | Plant monitoring system |
-
2021
- 2021-06-26 CN CN202110715227.4A patent/CN113589686B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH01109401A (en) * | 1987-10-22 | 1989-04-26 | Konica Corp | Process control method using plant model |
| US5201292A (en) * | 1991-08-30 | 1993-04-13 | Loral Aerospace Corp. | Apparatus and method for detecting vibration patterns |
| AU2009213031A1 (en) * | 2009-09-10 | 2011-03-24 | Kingston, William Mr | Tideway anchor |
| CN102435554A (en) * | 2011-09-08 | 2012-05-02 | 北京师范大学 | Method for acquiring farmland multiple-cropping index |
| CN106295888A (en) * | 2016-08-12 | 2017-01-04 | 东南大学 | A kind of public building based on measured data is joined and is built parking position and share time window and determine method |
| CN106527135A (en) * | 2016-11-21 | 2017-03-22 | 云南电网有限责任公司电力科学研究院 | Building energy management system optimization control method and device |
| CN107727399A (en) * | 2017-10-27 | 2018-02-23 | 中国人民解放军海军工程大学 | Fault Diagnosis of Roller Bearings based on dual-tree complex wavelet packet stream shape domain noise reduction |
| CN110457867A (en) * | 2019-10-08 | 2019-11-15 | 杭州知衣科技有限公司 | A kind of time series data based on machine learning is filled up and restoring method |
| CN112347571A (en) * | 2020-09-18 | 2021-02-09 | 中国人民解放军海军工程大学 | Rolling bearing residual life prediction method considering model and data uncertainty |
Non-Patent Citations (5)
| Title |
|---|
| Extracting Fundamental Periods to Segment Biomedical Signals;Anastasia Motrenko ， Vadim Strijov;IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS;第20卷(第6期);1466-1476 * |
| 基于机械振动的旋转供弹机构故障诊断综述;汤建，梁伟阁 等;控制工程;第25卷(第10期);1777-1781 * |
| 基于遗传模拟退火FCM算法的BDS-3三频组合优化选取;蔡聪聪 等;测绘工程;第29卷(第5期);60-65 * |
| 多输入混合深度学习网络的健康因子构建方法;张钢 等;系统工程与电子技术;第42卷(第10期);2390-2397 * |
| 脉搏信号中有效信号识别与特征提取方法研究;洋洋;陈小惠;王保强;姜吉荣;;电子测量与仪器学报(第01期);126-132 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113589686A (en) | 2021-11-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Pomponi et al. | A real-time approach to acoustic emission clustering | |
| CN105975443B (en) | Network anomalous behaviors detection method and system based on Lasso | |
| WO2016049983A1 (en) | User keyboard key-pressing behavior mode modeling and analysis system, and identity recognition method thereof | |
| Yan et al. | A novel intelligent detection method for rolling bearing based on IVMD and instantaneous energy distribution-permutation entropy | |
| US20220367057A1 (en) | Missing medical diagnosis data imputation method and apparatus, electronic device and medium | |
| CN115901249B (en) | Rolling bearing performance degradation evaluation method combining feature optimization and multi-strategy optimization SVDD | |
| CN114662386A (en) | Bearing fault diagnosis method and system | |
| CN114036610A (en) | Penetration depth prediction method based on data enhancement | |
| CN113589686B (en) | GSA-IFCM-based unit cycle time sequence self-adaptive extraction method | |
| CN111368892A (en) | Generalized S transformation and SVM electric energy quality disturbance efficient identification method | |
| CN111956210A (en) | Electrocardiosignal R wave identification method based on EEMD and signal structure analysis | |
| Yolacan et al. | System call anomaly detection using multi-hmms | |
| CN106228026A (en) | A kind of predicting residual useful life algorithm based on optimum degenerative character amount | |
| CN117579343A (en) | Network intrusion detection method oriented to reinforcement learning near-end policy optimization | |
| CN116595465A (en) | High-dimensional sparse data outlier detection method and system based on self-encoder and data enhancement | |
| CN112115184A (en) | Time series data detection method and device, computer equipment and storage medium | |
| CN113705107A (en) | Power consumption analysis method based on mean ridge regression | |
| CN112988186B (en) | Updating method and device of abnormality detection system | |
| CN117048524A (en) | Method and device for detecting vehicle faults, vehicle and storage medium | |
| Mufazzal et al. | Enhancing the classification performance of machine learning techniques by using hjorth’s and other statistical parameters for precise tracking of naturally evolving faults in ball bearings | |
| CN115242457A (en) | Method and device for detecting log data, electronic equipment and storage medium | |
| CN113889147A (en) | Mechanical system voiceprint signal matching method based on dynamic time warping algorithm | |
| CN115860708A (en) | Ship equipment state monitoring and fault diagnosis method | |
| Wen et al. | Rolling bearing fault diagnosis method based on improved variational mode decomposition and information entropy | |
| CN115438696A (en) | Gas turbine sensor fault diagnosis method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |