CN116595324A - Method for extracting signal transient impact starting point - Google Patents
Method for extracting signal transient impact starting point Download PDFInfo
- Publication number
- CN116595324A CN116595324A CN202310882321.8A CN202310882321A CN116595324A CN 116595324 A CN116595324 A CN 116595324A CN 202310882321 A CN202310882321 A CN 202310882321A CN 116595324 A CN116595324 A CN 116595324A
- Authority
- CN
- China
- Prior art keywords
- signal
- section
- segment
- transient impact
- point
- 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.)
- Granted
Links
- 230000001052 transient effect Effects 0.000 title claims abstract description 98
- 238000000034 method Methods 0.000 title claims abstract description 30
- 230000001133 acceleration Effects 0.000 claims abstract description 32
- 230000009466 transformation Effects 0.000 claims abstract description 8
- 238000000354 decomposition reaction Methods 0.000 claims description 17
- 238000005070 sampling Methods 0.000 claims description 9
- 230000011218 segmentation Effects 0.000 claims description 8
- 230000001351 cycling effect Effects 0.000 claims 1
- 238000001914 filtration Methods 0.000 abstract description 18
- 230000008569 process Effects 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000003672 processing method Methods 0.000 description 3
- 238000003745 diagnosis Methods 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 235000009413 Ratibida columnifera Nutrition 0.000 description 1
- 241000510442 Ratibida peduncularis Species 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000003116 impacting effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
- G06F2218/06—Denoising by applying a scale-space analysis, e.g. using wavelet analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Abstract
The invention discloses a method for extracting a signal transient impact starting point, which belongs to the technical field of signal processing of sensors and comprises the following steps: acquiring an electric signal containing a transient impact starting point, converting the electric signal into an acceleration signal, denoising the acceleration signal by wavelet packet transformation to obtain a denoised signalXSignal is sent toXAs a target signal, dividing the target signal into a plurality of signal segments arranged in time sequence, and judging whether the variance of the first signal segment is smaller than a first preset value. Denoising signal noise to be processed by adopting wavelet packet filtering to adapt to the condition of complex signal noise components, segmenting the signal, comparing the characteristic matching degree between the current signal segment and the preceding signal segment to find the signal segment containing transient impact starting point, and obtaining the signal segment containing transient impact starting pointThe transient impact starting point is searched in the signal segment of the transient impact starting point, so that the search of the transient impact starting point is quickened, and the efficiency is improved.
Description
Technical Field
The invention relates to the technical field of signal processing of sensors, in particular to a method for extracting a signal transient impact starting point.
Background
Impact refers to a very short duration, non-periodic load action relative to the natural period of the system, typically resulting from abrupt changes in mechanical component forces, manifested as an instantaneous increase in signal and then decay in mechanical vibration. The key to the state of motion and fault diagnosis of most mechanical devices is how to extract a local impulse response of a definite physical meaning from the motion state and fault diagnosis.
The vibration signal is used as an external dynamic expression form of mechanical equipment, is widely studied at present and comprises an acceleration signal, a displacement signal and the like, and compared with other vibration signals, the acceleration signal can more directly and rapidly embody the motion state of the machinery according to Newton's second law.
Since the acceleration signal detected in actual engineering is often accompanied by many other noise components, the signal detected by the acceleration sensor needs to be denoised to obtain a more reliable acceleration signal. The general filtering denoising processing method is difficult to cope with the situation of acceleration signals containing unknown effective components, the set filtering cut-off frequency is often unreasonable, and the denoising effect is poor or the effective components are filtered easily. And the current method for processing the impact response is low in speed and efficiency, and is not suitable for processing large-batch data in actual engineering.
Disclosure of Invention
The invention aims to provide a method for extracting transient impact starting points of signals, which aims to solve the problems of low speed and low efficiency of batch extraction of the transient impact starting points under the condition that the effective components of the signals are unknown.
In order to solve the technical problems, the invention provides a method for extracting a signal transient impact starting point, which comprises the following steps:
s1: acquiring an electric signal containing a transient impact starting point, converting the electric signal into an acceleration signal, and denoising the acceleration signal by wavelet packet transformation to obtain a denoised signalXThe signal is processedXAs a target signal;
s2: dividing the target signal into a plurality of signal segments arranged according to a time sequence, marking the N-th signal segment arranged according to the time sequence as the N-th signal segment, and judging whether the variance of the first signal segment is smaller than a first preset value or not;
s3: if the variance of the first signal segment is smaller than a first preset value, executing steps S4-S6, and if the variance of the first signal segment is not smaller than the first preset value, taking the first signal segment as a target signal, and repeatedly executing step S2;
s4: extracting feature matching degree between an M-th section signal section to be executed and a preamble signal section, wherein if the feature matching degree is larger than a second preset value, the M-th section signal section is a data section containing a transient impact starting point, and the data section containing the transient impact starting point is recorded as an impact starting section, wherein the preamble signal section comprises a first signal section to an M-1-th section signal section;
s5: extracting a characteristic value of the preamble signal section as an upper threshold value and a lower threshold value;
s6: and acquiring a threshold range based on the upper threshold value and the lower threshold value, searching a first data point which exceeds the threshold value range in the impact initial section and is arranged in time sequence, marking the data point as a transient impact point, and taking a first zero crossing point positioned before the transient impact point as a transient impact point.
Preferably, if the variance of the first signal segment is not smaller than the first preset value, that is, a transient impact starting point exists in the first signal segment, the step S2 is repeatedly executed until the window length of the first signal segment is smaller than a third preset value, and if the variance of the first signal segment is not smaller than the first preset value, the step S2 is stopped, the first signal segment is directly recorded as an impact starting point, and the signal acquisition starting point of the target signal is recorded as a transient impact starting point.
Preferably, the step of denoising the acceleration signal using wavelet packet transformation comprises:
setting a plurality of wavelet bases, and respectively carrying out wavelet packet decomposition and reconstruction on the acceleration signals;
based on the same decomposition layer number, calculating the maximum error and the average error of the reconstructed signals under different wavelet bases, selecting the wavelet base with the minimum maximum error and the minimum average error as the optimal wavelet base, and carrying out wavelet packet decomposition to obtain the decomposition signal.
Preferably, after the decomposition signal is acquired, the following steps are further performed:
setting a plurality of threshold rules, and respectively carrying out data processing and reconstruction on the decomposed signals;
calculating the signal-to-noise ratio and root mean square error of the reconstructed signal under different threshold rules, and selecting the threshold rule with the highest signal-to-noise ratio and the smallest root mean square error as the optimal threshold rule;
denoising the acceleration signal by adopting the optimal threshold rule and the optimal wavelet base to obtain the denoised signalX。
Preferably, the target signal is processed for a first window time length T 1 Segmenting to form the signal segments, wherein the segmentation number N S =[L1]/(f s T 1 +1), L1 is the number of sampling points,f s for picking upSample frequency.
Preferably, if the variance of the first signal segment is smaller than the first preset value, the following steps are cycled K times from the second signal segment, wherein K is less than or equal to N S -1：
Acquiring the variance of the M-th section signal section as a first signal characteristic value, and acquiring the mean value of the M-th section signal section as a second signal characteristic value;
acquiring a variance of the preamble signal section as a third signal characteristic value, and acquiring a mean value of the preamble signal section as a fourth signal characteristic value;
matching the M-th section signal section with the preamble data section to obtain the characteristic matching degree of the M-th section signal section and the preamble signal section;
if the feature matching degree is not greater than a second preset value, the M-th signal section is an non-impact section;
and if the characteristic matching degree is larger than a second preset value, the M-th signal section is the impact starting section.
Preferably, the ratio of the first signal characteristic value to the third signal characteristic value is the characteristic matching degree of the mth signal section and the preamble signal section.
Preferably, the difference between the second signal characteristic value and the fourth signal characteristic value is the characteristic matching degree of the mth signal section and the preamble signal section.
Preferably, a point from the starting point of signal acquisition to the first data point of the impact starting section is used as an impact non-starting signal section, the data values in the impact non-starting signal section are arranged in sequence from big to small, the average value of the data values of the fourth preset value before being sequentially selected as the upper threshold value, and the average value of the data values of the third preset value before being reversely selected as the lower threshold value.
According to the method for extracting the transient impact starting point of the signal, the transient impact starting point is extracted under the condition that the effective components of the signal are unknown, the sampling signal is subjected to denoising treatment by adopting wavelet packet filtering, the condition that the noise component of the signal is complex in actual engineering is more met, and the wavelet packet filtering parameters can be adjusted according to the source signal, so that a better filtering effect is achieved.
In addition, the time domain signals are segmented after wavelet packet filtering, and feature matching degree is extracted according to a time window segmentation sequence, so that transient impact starting points are extracted from time domain waveforms of acceleration signals, the problems of large calculated amount, low efficiency and the like in practical application of some processing methods are solved, and the practicability of the transient impact starting points is effectively improved.
Drawings
FIG. 1 is a flow chart of a method for extracting a signal transient impact starting point provided by the invention;
FIG. 2 is a flow chart of a method for extracting transient impact starting points from signal segmentation features provided by the invention;
fig. 3 is a schematic diagram of finding a transient starting point according to an embodiment of the present invention.
Detailed Description
The method for extracting the signal transient impact point according to the present invention is described in further detail below with reference to the accompanying drawings and specific embodiments. The advantages and features of the present invention will become more apparent from the following description. It should be noted that the drawings are in a very simplified form and are all to a non-precise scale, merely for convenience and clarity in aiding in the description of embodiments of the invention.
The inventor researches and discovers that some methods for searching the transient impact starting point of the signal are low in processing speed and low in efficiency.
Based on the above, the real core idea of the invention is that the signal noise to be processed is denoised by adopting wavelet packet filtering to adapt to the condition of complex signal noise components, then the signal is segmented, the signal segment containing the transient impact starting point is searched by comparing the characteristic matching degree between the current signal segment and the preamble signal segment, and then the transient impact starting point is searched in the signal segment containing the transient impact starting point, thereby accelerating the search of the transient impact starting point and improving the efficiency.
Specifically, please refer to fig. 1, which is a schematic diagram of an embodiment of the present invention. As shown in fig. 1, a method for extracting a signal transient impact starting point includes the following steps:
s1: acquiring an electric signal containing a transient impact starting point, converting the electric signal into an acceleration signal, and denoising the acceleration signal by wavelet packet transformation to obtain a denoised signalXThe signal is processedXAs a target signal.
S2: dividing the target signal into a plurality of signal segments arranged according to a time sequence, marking the N-th signal segment arranged according to the time sequence as the N-th signal segment, and judging whether the variance of the first signal segment is smaller than a first preset value. Wherein N is more than or equal to 2.
S3: if the variance of the first signal segment is smaller than the first preset value, executing steps S4-S6, and if the variance of the first signal segment is not smaller than the first preset value, taking the first signal segment as a target signal, and repeatedly executing step S2. And if the variance of the first signal segment is not smaller than the first preset value, that is, a transient impact starting point exists in the first signal segment, repeating the step S2 until the window length of the first signal segment is smaller than a third preset value and the variance of the first signal segment is not smaller than the first preset value, stopping executing the step S2, directly marking the first signal segment as an impact starting point, and taking the signal acquisition starting point of the target signal as a transient impact starting point.
S4: and extracting the feature matching degree between the M-th section signal section to be executed and the preamble signal section, wherein if the feature matching degree is larger than a second preset value, the M-th section signal section is a data section containing a transient impact starting point, and the data section containing the transient impact starting point is recorded as an impact starting section, wherein the preamble signal section comprises a first signal section to an M-1-th section signal section. Wherein M is more than or equal to 2.
S5: and extracting the characteristic value of the preamble signal segment as an upper threshold value and a lower threshold value.
S6: and acquiring a threshold range based on the upper threshold value and the lower threshold value, searching a first data point which exceeds the threshold value range in the impact initial section and is arranged in time sequence, marking the data point as a transient impact point, and taking a first zero crossing point positioned before the transient impact point as a transient impact point.
The sampling signal can be subjected to denoising processing under the condition that the effective components of the signal are unknown through wavelet packet filtering, and a proper wavelet base and a threshold rule can be selected according to a source signal so as to achieve a better denoising effect. The time domain signal is segmented after the wavelet packet is filtered, the feature matching degree is extracted according to the time window segmentation sequence, and then the transient impact starting point is extracted from the time domain waveform of the acceleration signal, so that the problems of large calculated amount, low efficiency and the like in the practical application of some processing methods are solved, and the practicability of the transient impact starting point is effectively improved.
Based on the above steps, it can be known that, in the case that the transient impact starting point exists in the first signal segment, the step S2 is repeatedly executed until the window length of the first signal segment is smaller than a third preset value and the variance of the first signal segment is not smaller than the first preset value, then the step S2 is stopped, the first signal segment is directly recorded as the impact starting point, the signal acquisition starting point of the target signal is directly recorded as the transient impact starting point, so as to adapt to acceleration signals with different lengths, when the window length of the signal segment divided by the target signal is sufficiently small, the first signal segment is cut to obtain the situation that the transient impact starting point cannot be processed in the signal acquisition starting point due to the fact that the transient impact starting point exists in the first signal segment, so that special processing is performed for the special situation, namely, the step S2 is repeatedly executed until the window length of the first signal segment is smaller than the third preset value and the variance of the first signal segment is not smaller than the first preset value, then the step S2 is stopped, the first signal segment is directly recorded as the transient impact starting point, and the transient impact starting point is directly recorded as the target impact starting point. The length of time of the third preset value may be specifically set according to the length of time of the obtained acceleration signal, even in the case of an actual mechanical device, or the number of times of performing segmentation on the target signal may be defined.
Taking the step S2 of limiting the execution of the step S2 twice as an example, when the window length of the first signal segment is smaller than the third preset value, if a transient impact starting point exists in the first signal segment after the target signal is divided into a plurality of signal segments, the first signal segment is segmented again, the first signal segment is divided into a plurality of sub-signal segments, and the following specific execution flow is provided:
s1: directly acquiring an electric signal containing a transient impact starting point from a sensor, converting the electric signal into an acceleration signal through the rule of correspondence between the sensitivity of the sensor and the voltage-acceleration range, denoising the acceleration signal by wavelet packet transformation to obtain a denoised signalX. According to the actually measured acceleration signal or the source signal, the wavelet packet filtering parameters are adjusted, the denoising processing of the wavelet packet filtering is carried out based on the adjusted wavelet packet filtering parameters, and proper wavelet packet transformation and threshold rules are required to be selected for filtering.
The step of denoising the acceleration signal by adopting wavelet packet transformation comprises the following steps:
and a plurality of wavelet bases and fixed decomposition layer numbers are arranged, wavelet packet decomposition and reconstruction are respectively carried out on the acceleration signals, and common wavelet bases can be selected for data processing, such as Symlets, daubechies, haar, biorthogonal, coiflets.
And then, calculating the maximum error and the average error of the reconstructed signals under different wavelet bases based on the same decomposition layer number, and selecting the wavelet base with the minimum maximum error and the minimum average error as the optimal wavelet base to decompose the wavelet packet to obtain a decomposition signal.
After the decomposed signal is obtained, the following steps are also performed to obtain the optimal threshold rule for the denoising process:
setting a plurality of threshold rules, respectively carrying out data processing and reconstruction on the decomposed signals, for example, four threshold rules of a maximum and minimum threshold, a heuristic threshold, a fixed threshold and an unbiased likelihood estimation threshold, respectively carrying out data processing and reconstruction on the decomposed signals based on the same decomposed layer number;
and calculating the signal-to-noise ratio and the root mean square error of the reconstructed signal under different threshold rules, and selecting the threshold rule with the highest signal-to-noise ratio and the smallest root mean square error as the optimal threshold rule.
Based on the steps, obtaining a corresponding optimal threshold rule and an optimal wavelet base, and denoising the acceleration signal by adopting the optimal threshold rule and the optimal wavelet baseObtaining the denoised signalX。
S2: dividing the target signal into a plurality of signal segments arranged in time sequence, and recording the N-th signal segment arranged in time sequence as the N-th signal segment so as to perform segment comparison and obtain the signal segment containing the transient impact starting point.
Specifically, the target signal is processed in a first window time length T 1 Segmenting to form a plurality of signal segments, wherein the segmentation number N S1 =[L1]/(f s1 T 1 +1), L1 is the number of sampling points,f s1 is the sampling frequency.
S3: if the variance of the first signal segment is not smaller than the first preset value, a transient impact starting point exists in the first signal segment, namely the first signal segment is an impact starting segment, and the following steps are executed:
s3.1: the first signal segment is segmented into a plurality of sub-signal segments which are arranged according to time sequence, wherein the N-th sub-signal segment which is arranged according to time sequence is marked as the N-th sub-signal segment.
S3.2: if the variance of the first sub-signal segment is larger than the first preset value, the first sub-signal segment is the impact initial segment, and the signalXThe signal acquisition starting point of (2) is the transient impact starting point.
If the first signal section is the impact starting section, the first signal section is divided into a first window time length T 2 Forming a plurality of time sequence arranged sub-signal segments in a segmenting way, wherein the number of the segments N S2 =[L2]/（f s2 T 2 +1), L2 is the number of sampling points,f s2 is the sampling frequency.
S3.3: if the variance of the first sub-signal segment is smaller than the first preset value, the transient impact starting point does not exist in the first sub-signal segment, and the transient impact starting point is not the impact starting segment. Then a comparison of the current signal segment (or mth segment signal segment) and the preamble signal segment is performed as in step S5.
S4: if the variance of the first signal segment is not smaller than the first preset value, the transient impact starting point does not exist in the first signal segment, and the following step is executed, namely, step S5. I.e. step S5 is performed when no transient impact start point is present in the first sub-signal segment or the first signal segment is not present.
S5: firstly, starting to circularly execute the following steps from the second signal segment or the second sub-signal segment until an impact initial segment containing a transient impact starting point is found:
s5.1: extracting the feature matching degree between the current signal section and the preamble signal section to be executed, if the feature matching degree is larger than a second preset value, the current signal section is a data section containing a transient impact starting point, and the data section containing the transient impact starting point is marked as an impact starting section, wherein a point from the signal acquisition starting point to the first data point of the current signal section or the sub-signal section is the preamble signal section; if the feature matching degree is not greater than the second preset value, the data segment of the next window time of the current signal segment is set as the current signal segment of the next cycle.
Specifically, if the variance of the first signal segment is smaller than the first preset value, or the variance of the first sub-signal segment is smaller than the first preset value, the following steps are repeated K times from the second signal segment or the second sub-signal segment, wherein K is smaller than or equal to N S -1：
Acquiring a variance of a current signal section as a first signal characteristic value, and acquiring a mean value of the current signal section as a second signal characteristic value;
acquiring a variance of the preamble signal section as a third signal characteristic value, and taking a mean value of the preamble signal section as a fourth signal characteristic value;
matching the current signal segment with the preamble data segment to obtain the characteristic matching degree of the current signal segment and the preamble signal segment;
if the feature matching degree is not greater than the second preset value, the current signal section is an non-impact section;
if the feature matching degree is larger than a second preset value, the current signal section is an impact initial section;
and setting the data segment of the next window time of the current signal segment as the current signal segment of the next cycle.
The ratio of the first signal characteristic value to the third signal characteristic value is the characteristic matching degree of the current signal section and the preamble signal section. Or the difference between the second signal characteristic value and the fourth signal characteristic value is the characteristic matching degree of the current signal section and the preamble signal section.
When the ratio of the first signal characteristic value to the third signal characteristic value is the characteristic matching degree of the current signal section and the preamble signal section, the characteristic matching degree is larger than a second preset value, and the current signal section is an impact initial section; when the difference between the second signal characteristic value and the fourth signal characteristic value is the characteristic matching degree of the current signal section and the preamble signal section, the characteristic matching degree is larger than a second preset value, and the current signal section is the impact starting section.
Or in another embodiment, the ratio of the first signal characteristic value to the third signal characteristic value is recorded as a first characteristic matching degree of the current signal section and the preamble signal section, the difference between the second signal characteristic value and the fourth signal characteristic value is recorded as a second characteristic matching degree of the current signal section and the preamble signal section, and when the first characteristic matching degree is greater than a preset value and the second characteristic matching degree is greater than another preset value, the current signal section is an impact starting section.
The comparison is sequentially performed from the second signal segment or the second sub-signal segment according to the time sequence, wherein the current signal segment, namely the Mth signal segment (also can be the Mth sub-signal segment), and after the comparison of the Mth signal segment is performed, the comparison of the M+1th signal segment (also can be the Mth+1th sub-signal segment) is sequentially performed.
S5.2: and secondly, extracting characteristic values of the data segments from the signal acquisition starting point to the front of the impact starting point according to the searched impact starting point as an upper threshold value and a lower threshold value.
Specifically, a point from a signal acquisition starting point to a first data point of an impact starting section is taken as an impact non-starting signal section, data values in the impact non-starting signal section are arranged in a sequence from big to small, the average value of the data values of the fourth preset value before being sequentially selected as an upper threshold value, and the average value of the data values of the third preset value before being sequentially selected as a lower threshold value. For example, the average value of the first 5 data values is selected to avoid data errors.
S5.3: and acquiring a threshold range based on the upper threshold value and the lower threshold value, searching data points exceeding the threshold range in the impact initial section, recording the data points as transient impact points, and taking the first zero crossing point positioned before the transient impact points as the transient impact point.
Specifically, a threshold range is formed based on an upper threshold and a lower threshold, the threshold formed by the upper threshold and the lower threshold is used as the threshold range, the first data point which exceeds the threshold range and is arranged in time sequence in the impact starting section is searched, the data point is recorded as a transient impact point, and the first zero crossing point positioned before the transient impact point is the transient impact point. Wherein the first data point here refers to the zero crossing point located before the transient impact start point (time sequence), and the same first data point is also the first data point in time sequence out of the threshold range.
Based on the above method, the present invention also provides an embodiment of data processing, and the implementation flow can be seen from fig. 2 to fig. 3, and the specific implementation steps are as follows:
step S1: firstly, an acceleration signal, namely a source signal S, is obtained according to an actually measured electric signal (comprising a transient impact starting point), and a plurality of common wavelet bases are selected for carrying out wavelet packet filtering to carry out digital signal processing by adopting Symlets or Daubechies quadrature bases: haar wavelets, daubechies (dbN) wavelets Symlets (symN), mexicanHat (mexh) wavelets, morlet wavelets, meyerhaar functions, and setting the number of decomposition layers to be 4, and respectively performing wavelet packet decomposition and reconstruction on the source signal by adopting the different wavelet bases to obtain an optimal wavelet base and an optimal threshold rule.
Wherein, the supporting length of Daubechies is 2N-1, the supporting length of symlets (n=2, 3, …, 8) is 2N-1, N is the number of the selected wavelet base, the supporting lengths of other wavelet bases are not described in detail, it can be understood that if the signal with higher amplitude value needs to be processed in practice, the wavelet base with the supporting length less than 5 is removed; if it is actually necessary to process a higher frequency resolution signal, the wavelet basis with a support length greater than 17 is removed.
And then calculating the maximum error and the average error of the reconstructed signals under the residual different wavelet bases, and selecting the wavelet base with the minimum maximum error and the minimum average error as the optimal wavelet base to carry out wavelet packet decomposition to obtain a decomposed signal.
And respectively carrying out data processing and reconstruction on the decomposition signals obtained based on the four threshold rules of the maximum and minimum threshold, the heuristic threshold, the fixed threshold and the unbiased likelihood estimation threshold, then calculating the signal-to-noise ratio and the root mean square error of the reconstruction signals under different threshold rules, and selecting the threshold rule with the highest signal-to-noise ratio and the minimum root mean square error as the optimal threshold rule.
Denoising the source signal according to the obtained optimal threshold rule and the optimal wavelet packet filtering parameters to obtain a target denoised signalX。
Step S2: a flow chart of a method for extracting transient impact starting points by signal segmentation features is shown in fig. 2, and signals are processedXSegmenting in a time window of 0.5 seconds, e.g. signalsXThe time length is 3.2 seconds, the number of segments is 7, and the number of points N of 1 to 6 single-segment data p1 =0.5f s Section 7 data Point number 0.2f s Then calculating the variance of the first signal segment, if the variance of the first signal segment is larger than 1.1, indicating that a transient impact starting point exists in the first signal segment, segmenting the first signal segment by a time window of 0.1 second, and counting N single-segment data points p2 =0.1f s And calculate T 2 And if the variance of the first sub-signal segment is larger than 1.1, a transient impact starting point exists in the first sub-signal segment, and the signal acquisition starting point is the transient impact starting point.
If the variance of the first sub-signal segment is not greater than 1.1 or the variance of the first signal segment is not greater than 1.1, that is, the transient impact starting point does not exist in the first sub-signal segment or the first signal segment, calculating the mean matching degree or variance matching degree of the current signal segment and the preamble signal segment from the second sub-signal segment or the second signal segment.
Section i (i=2, 3, …, K) mean and data point 1 to data point i×n p The average value and the average value matching degree calculating process are as follows:
calculating an i-th segment mean value:
calculate data points 1 through i p Average value:
calculating the average matching degree:
wherein M is p For the second preset value, the second signal characteristic value and the fourth signal characteristic value are average values, and the acceleration deviation degree is represented by the difference of the average values, wherein the second preset value is generally not more than 1.5.
Section i (i=2, 3, …, K) variance and data point 1 to data point i×n p Variance, variance matching degree calculation process:
calculating the ith section variance:
calculate data points 1 through i p Variance:
calculating the variance matching degree:
wherein V is p For the second preset value, the value generally does not exceed 0.06, the first signal characteristic value and the third signal characteristic value are both variances, and the amplitude variation degree of the signal is represented by the ratio of the variances.
The matching calculation process is circulated by K (K is less than or equal to N) s -1) at the end of each cycle, setting the next window time period as the current signal period until the degree of matching is greater than a second preset value, the current signal period being the impact start period.
Step S3: according to the impact initial section obtained in the step S2, a point from the signal acquisition initial point to the first data point of the impact initial section is taken as an impact non-initial signalAnd the sections are used for arranging the data values in the impacting non-starting signal section according to the sequence from large to small, sequentially selecting the average value of the data values of the fourth preset value before as an upper threshold value, and reversely selecting the average value of the data values of the third preset value before as a lower threshold value. To avoid data errors, the data values of the signals before the start are arranged in order from large to small, for example, the average value of the first 5 data values is sequentially selected and used as the upper threshold value for judging the impact signal, and is set as T up The average value of the first 5 data values is selected in the reverse order and is used as the lower threshold value for judging the impact signal, and is set as T low 。
Step S4: the specific schematic diagram of the transient impact starting point is shown in fig. 3, the first data point of the impact starting section is searched, the first data point which is larger than the upper threshold value or smaller than the lower threshold value in the impact starting section data is the transient impact point, and the first zero crossing point at the left side of the transient impact point is the transient starting point.
In summary, the method for extracting the transient impact starting point of the signal provided by the embodiment of the invention extracts the transient impact starting point under the condition that the effective components of the signal are unknown, adopts wavelet packet filtering to carry out denoising treatment on the sampled signal, is more in line with the condition that the noise components of the signal are complex in actual engineering, and can also adjust the wavelet packet filtering parameters according to the source signal so as to achieve better filtering effect.
The above description is only illustrative of the preferred embodiments of the present invention and is not intended to limit the scope of the present invention, and any alterations and modifications made by those skilled in the art based on the above disclosure shall fall within the scope of the appended claims.
Claims (9)
1. The method for extracting the signal transient impact starting point is characterized by comprising the following steps of:
s1: acquiring an electric signal containing a transient impact starting point, converting the electric signal into an acceleration signal, and denoising the acceleration signal by wavelet packet transformation to obtain a denoised signalXThe signal is processedXAs a target signal;
s2: dividing the target signal into a plurality of signal segments arranged according to a time sequence, marking the N-th signal segment arranged according to the time sequence as the N-th signal segment, and judging whether the variance of the first signal segment is smaller than a first preset value or not;
s3: if the variance of the first signal segment is smaller than a first preset value, executing steps S4-S6, and if the variance of the first signal segment is not smaller than the first preset value, taking the first signal segment as a target signal, and repeatedly executing step S2;
s4: extracting feature matching degree between an M-th section signal section to be executed and a preamble signal section, wherein if the feature matching degree is larger than a second preset value, the M-th section signal section is a data section containing a transient impact starting point, and the data section containing the transient impact starting point is recorded as an impact starting section, wherein the preamble signal section comprises a first signal section to an M-1-th section signal section;
s5: extracting a characteristic value of the preamble signal section as an upper threshold value and a lower threshold value;
s6: and acquiring a threshold range based on the upper threshold value and the lower threshold value, searching a first data point which exceeds the threshold value range in the impact initial section and is arranged in time sequence, marking the data point as a transient impact point, and taking a first zero crossing point positioned before the transient impact point as a transient impact point.
2. The method of claim 1, wherein if the variance of the first signal segment is not less than the first preset value, that is, the first signal segment has a transient impact starting point, repeating step S2 until the window length of the first signal segment is less than a third preset value and the variance of the first signal segment is not less than the first preset value, stopping executing step S2, directly marking the first signal segment as an impact starting segment, and taking the signal acquisition starting point of the target signal as a transient impact starting point.
3. The method of extracting a signal transient impact point of claim 1, wherein the step of denoising the acceleration signal using wavelet packet transform comprises:
setting a plurality of wavelet bases, and respectively carrying out wavelet packet decomposition and reconstruction on the acceleration signals;
based on the same decomposition layer number, calculating the maximum error and the average error of the reconstructed signals under different wavelet bases, selecting the wavelet base with the minimum maximum error and the minimum average error as the optimal wavelet base, and carrying out wavelet packet decomposition to obtain the decomposition signal.
4. A method of extracting a signal transient impact point as defined in claim 3, further comprising, after the acquisition of the decomposed signal, the steps of:
setting a plurality of threshold rules, and respectively carrying out data processing and reconstruction on the decomposed signals;
calculating the signal-to-noise ratio and root mean square error of the reconstructed signal under different threshold rules, and selecting the threshold rule with the highest signal-to-noise ratio and the smallest root mean square error as the optimal threshold rule;
denoising the acceleration signal by adopting the optimal threshold rule and the optimal wavelet base to obtain the denoised signalX。
5. The method for extracting a signal transient impact point according to claim 1, wherein said target signal is represented by a first window time length T 1 Segmenting to form the plurality of signal segments, wherein the segmentation number N S =[L1]/(f s T 1 +1), L1 is the number of sampling points,f s is the sampling frequency.
6. The method of extracting a signal transient impact point of claim 5, wherein if the variance of said first signal segment is less than said first preset value, cycling K times from said second signal segment, wherein k.ltoreq.n S -1：
Acquiring the variance of the M-th section signal section as a first signal characteristic value, and acquiring the mean value of the M-th section signal section as a second signal characteristic value;
acquiring a variance of the preamble signal section as a third signal characteristic value, and acquiring a mean value of the preamble signal section as a fourth signal characteristic value;
matching the M-th section signal section with the preamble signal section to obtain the characteristic matching degree of the M-th section signal section and the preamble signal section;
if the feature matching degree is not greater than a second preset value, the M-th signal section is an non-impact section;
and if the characteristic matching degree is larger than a second preset value, the M-th signal section is the impact starting section.
7. The method of extracting a signal transient impact point of claim 6, wherein a ratio of said first signal characteristic value to said third signal characteristic value is a characteristic matching degree of said mth signal segment and said preamble signal segment.
8. The method of extracting a signal transient impact point of claim 7, wherein a difference between said second signal characteristic value and said fourth signal characteristic value is a characteristic matching degree of said mth signal segment and said preamble signal segment.
9. The method for extracting a signal transient impact starting point according to claim 7, wherein a point from the signal acquisition starting point to a point before a first data point of the impact starting section is used as an impact non-starting signal section, data values in the impact non-starting signal section are arranged in a sequence from large to small, a mean value of a plurality of data values of a fourth preset value before being sequentially selected as the upper threshold value, and a mean value of a plurality of data values of a third preset value before being reversely selected as the lower threshold value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310882321.8A CN116595324B (en) | 2023-07-19 | 2023-07-19 | Method for extracting signal transient impact starting point |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310882321.8A CN116595324B (en) | 2023-07-19 | 2023-07-19 | Method for extracting signal transient impact starting point |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116595324A true CN116595324A (en) | 2023-08-15 |
| CN116595324B CN116595324B (en) | 2023-11-10 |
Family
ID=87606617
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310882321.8A Active CN116595324B (en) | 2023-07-19 | 2023-07-19 | Method for extracting signal transient impact starting point |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116595324B (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105241666A (en) * | 2015-09-21 | 2016-01-13 | 华南理工大学 | Rolling bearing fault feature extraction method based on signal sparse representation theory |
| WO2016206056A1 (en) * | 2015-06-25 | 2016-12-29 | 国家电网公司 | Circuit breaker detection method, device and system |
| US9692529B1 (en) * | 2016-08-16 | 2017-06-27 | Southwest Research Institute | System and method for signal emitter identification using higher-order cumulants |
| US20180231406A1 (en) * | 2016-02-23 | 2018-08-16 | Hefei University Of Technology | Kalman filter based anti-transient-impact-vibration-interference signal processing method and system for vortex flowmeter |
| CN109001703A (en) * | 2018-08-10 | 2018-12-14 | 南京信息工程大学 | A kind of sea clutter denoising method based on the processing of wavelet packet multi-threshold |
| CN109598255A (en) * | 2018-12-19 | 2019-04-09 | 北京化工大学 | A kind of reciprocating mechanical vibration signal impact initial point self-adaptation extraction method based on energy operator k- gradient |
| CN114186577A (en) * | 2020-08-27 | 2022-03-15 | 江苏金恒信息科技股份有限公司 | Vibration acceleration signal denoising method and device based on wavelet transformation |
| CN115493836A (en) * | 2022-11-15 | 2022-12-20 | 北谷电子有限公司 | Mechanical fault diagnosis method and system |
| CN116451027A (en) * | 2023-04-25 | 2023-07-18 | 桂林电子科技大学 | Unmanned aerial vehicle identification method based on remote control signal two-stage empirical mode decomposition |
-
2023
- 2023-07-19 CN CN202310882321.8A patent/CN116595324B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016206056A1 (en) * | 2015-06-25 | 2016-12-29 | 国家电网公司 | Circuit breaker detection method, device and system |
| CN105241666A (en) * | 2015-09-21 | 2016-01-13 | 华南理工大学 | Rolling bearing fault feature extraction method based on signal sparse representation theory |
| US20180231406A1 (en) * | 2016-02-23 | 2018-08-16 | Hefei University Of Technology | Kalman filter based anti-transient-impact-vibration-interference signal processing method and system for vortex flowmeter |
| US9692529B1 (en) * | 2016-08-16 | 2017-06-27 | Southwest Research Institute | System and method for signal emitter identification using higher-order cumulants |
| CN109001703A (en) * | 2018-08-10 | 2018-12-14 | 南京信息工程大学 | A kind of sea clutter denoising method based on the processing of wavelet packet multi-threshold |
| CN109598255A (en) * | 2018-12-19 | 2019-04-09 | 北京化工大学 | A kind of reciprocating mechanical vibration signal impact initial point self-adaptation extraction method based on energy operator k- gradient |
| CN114186577A (en) * | 2020-08-27 | 2022-03-15 | 江苏金恒信息科技股份有限公司 | Vibration acceleration signal denoising method and device based on wavelet transformation |
| CN115493836A (en) * | 2022-11-15 | 2022-12-20 | 北谷电子有限公司 | Mechanical fault diagnosis method and system |
| CN116451027A (en) * | 2023-04-25 | 2023-07-18 | 桂林电子科技大学 | Unmanned aerial vehicle identification method based on remote control signal two-stage empirical mode decomposition |
Non-Patent Citations (3)
| Title |
|---|
| JIANG JINSHUI等: "Research on the Method of Features Extraction for Non-stationary Transient Signal Based on EMD method", 《ICCSN\'09: PROCEEDING OF THE 2009 INTERNATIONAL CONFERENCE ON COMMUNICATION SOFTWARE AND NETWORKS》, pages 637 - 641 * |
| 康玉梅等: "基于小波包分数幂阈值的声发射去噪研究", 《有色矿冶》, vol. 32, no. 05, pages 16 - 20 * |
| 秦毅等: "改进的Hilbert-Huang变换在信号瞬态特征提取中的应用", 《振动与冲击》, vol. 27, no. 11, pages 129 - 133 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116595324B (en) | 2023-11-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107451557B (en) | Power transmission line short-circuit fault diagnosis method based on empirical wavelet transform and local energy | |
| CN110836775B (en) | Rolling bearing fault identification method and system based on multi-dimensional entropy space distance | |
| KR20090078075A (en) | Fault diagnosis of inductirn motors by dft and wavelet | |
| CN110702411A (en) | Residual error network rolling bearing fault diagnosis method based on time-frequency analysis | |
| CN112857804B (en) | Rolling bearing fault diagnosis method, device, medium and computer equipment | |
| CN108549078B (en) | Cross-channel combination and detection method for radar pulse signals | |
| CN112284727B (en) | Rotary machine fault diagnosis method based on convolution maximum and minimum penalty algorithm | |
| CN113468760B (en) | Motor weak fault detection method and system based on dictionary learning | |
| CN113537102B (en) | Feature extraction method of microseismic signals | |
| CN113177537A (en) | Fault diagnosis method and system for rotary mechanical equipment | |
| Lu et al. | Adaptive online dictionary learning for bearing fault diagnosis | |
| CN107966287B (en) | Weak fault feature extraction method for self-adaptive electromechanical equipment | |
| CN114486263B (en) | Noise reduction demodulation method for vibration signal of rolling bearing of rotary machine | |
| CN111189624B (en) | Method for identifying loosening state of bolt connection structure based on vibration signal time-frequency characteristics | |
| CN115730199A (en) | Method and system for noise reduction and fault feature extraction of vibration signal of rolling bearing | |
| CN102184451A (en) | Multi-wavelet fusion feature based neuron action feature extraction method | |
| CN116595324B (en) | Method for extracting signal transient impact starting point | |
| CN106980722B (en) | Method for detecting and removing harmonic component in impulse response | |
| Li et al. | A Reliable Strategy for Improving Automatic First‐Arrival Picking of High‐Noise Three‐Component Microseismic Data | |
| CN114061746B (en) | Repeated transient signal extraction method in rotary machinery fault diagnosis | |
| CN115017964B (en) | Magnetotelluric signal denoising method and system based on attention mechanism sparse representation | |
| CN116127297A (en) | Electromechanical device fault diagnosis method and system | |
| CN115563480A (en) | Gear fault identification method for screening octave geometric modal decomposition based on kurtosis ratio coefficient | |
| CN115329820A (en) | Power grid disturbance identification method and system based on improved FastICA | |
| CN112116917B (en) | Phase jump degree-based method for separating acoustic signals of reactor body and fan |
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 |