CN112014107A - Improved empirical mode decomposition bearing vibration analysis method and system - Google Patents
Improved empirical mode decomposition bearing vibration analysis method and system Download PDFInfo
- Publication number
- CN112014107A CN112014107A CN202010169517.9A CN202010169517A CN112014107A CN 112014107 A CN112014107 A CN 112014107A CN 202010169517 A CN202010169517 A CN 202010169517A CN 112014107 A CN112014107 A CN 112014107A
- Authority
- CN
- China
- Prior art keywords
- energy signal
- low
- decomposition
- signal
- bearing vibration
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/04—Bearings
- G01M13/045—Acoustic or vibration analysis
Abstract
The invention relates to an improved empirical mode decomposition bearing vibration analysis method and system. Firstly, carrying out energy separation on signals to process non-event mode aliasing, then respectively carrying out event mode aliasing on the separated signals to obtain signals to be decomposed, and eliminating an end effect in empirical mode decomposition by utilizing a monotone consistency principle so as to obtain final bearing vibration decomposition signals. The invention can improve the bearing vibration signal decomposition precision, increase the analysis efficiency and be more convenient for use in engineering practice.
Description
Technical Field
The invention relates to the field of bearing vibration analysis, in particular to a vibration analysis method and system for an improved empirical mode decomposition bearing.
Background
In 1998, the Norden E Huang of the national aerospace agency proposes a new method for non-stationary and non-linear signals, and the basic idea is to decompose signals into a series of Intrinsic Mode Functions (IMF) by performing Empirical Mode Decomposition (EMD) on the non-stationary signals, and to use the methods of EMD Decomposition and continuous mean screening to prove the completeness and orthogonality of the EMD Decomposition. Along with the continuous and deep research, the development of EMD gradually forms a theoretical system of the EMD, a new time-frequency analysis method for analyzing nonlinear and non-stationary signals is formed, and the EMD is widely applied to the field of analysis of vibration signals of rotary machines.
However, bearing vibration signals have strong non-linearity and non-stationary characteristics, and from the research situation of the empirical mode decomposition method at home and abroad, modal aliasing and end effect phenomena exist in the empirical mode decomposition, so that the research on condition monitoring, fault diagnosis and trend prediction based on the bearing vibration signals has many limitations. Modal aliasing is the separation of modal components that cannot be effectively measured according to a time characteristic scale, so that more than one modal component is contained in a natural modal function, and thus the inherent properties of a signal cannot be clearly reflected. The mode aliasing is divided into an event mode aliasing and a non-event mode, and the event mode easily causes a local extreme value phenomenon, so that the EMD decomposition precision is reduced, and the EMD decomposition is invalid when the EMD decomposition is serious. On the other hand, non-event mode aliasing is a phenomenon in which some frequency signals cannot be separated correctly in a mixed signal in which several frequencies coexist. The end-point effect is generated when the EMD decomposition is used, because whether the end-point is an extreme value or not is difficult to determine when the envelope of the signal is calculated by the maximum value and the minimum value of the signal, and the end-point effect causes the problem that the EMD decomposition is difficult to converge. Due to the defects, the empirical mode method has certain limitation in bearing vibration signal analysis, and therefore, the improvement of the EMD algorithm is of great significance.
Disclosure of Invention
In view of the above problems, the present invention provides an improved empirical mode decomposition bearing vibration analysis method and system.
In order to achieve the purpose, the invention provides the following scheme:
an improved empirical mode decomposition bearing vibration analysis method comprises the following steps:
decomposing the bearing vibration signal to obtain a first high energy signal and a first low energy signal;
determining whether an incident modality exists in the first high energy signal and the first low energy signal;
if yes, processing the first high-energy signal and the first low-energy signal according to an event mode processing method;
if not, directly decomposing the first high-energy signal and the first low-energy signal by adopting an empirical mode decomposition method;
obtaining a second high energy signal and a second low energy signal by suppressing an endpoint effect when performing empirical mode decomposition on the first high energy signal and the first low energy signal;
judging whether the second low-energy signal has obvious waveform characteristics;
if so, performing Hilbert-Huang conversion on the second high-energy signal and the second low-energy signal to obtain a conversion map, and ending the whole decomposition process;
and if not, assigning the second low-energy signal to the bearing vibration signal for decomposing again until obvious characteristics exist after the low-energy signal is decomposed or the signal to be decomposed does not meet the condition of empirical mode decomposition, performing Hilbert-Huang transformation on the obtained low-energy signal to obtain a transformation map, and finishing the whole decomposition process.
Optionally, the decomposing the bearing vibration signal to obtain a first high energy signal and a first low energy signal specifically includes:
and decomposing the bearing vibration signal by using a singular value to obtain a first high-energy signal and a first low-energy signal.
Optionally, the processing the first high energy signal and the first low energy signal according to an event modality processing method specifically includes:
determining the time domain of the generation of the event mode based on the time characteristic scale stability and the combined stability method based on the extreme value difference stability;
and removing the incident modes of the first high-energy signal and the first low-energy signal by adopting a singular value denoising method according to a time domain.
Optionally, when performing empirical mode decomposition on the first high energy signal and the first low energy signal, obtaining a second high energy signal and a second low energy signal by suppressing an endpoint effect, specifically including:
when empirical mode decomposition is carried out on the first high-energy signal and the first low-energy signal, the endpoint effect is restrained by utilizing the principle of monotonicity consistency of the signals during each decomposition, and a second high-energy signal and a second low-energy signal are obtained.
An improved empirical mode decomposition bearing vibration analysis system comprising:
the first decomposition module is used for decomposing the bearing vibration signal to obtain a first high-energy signal and a first low-energy signal;
the first judging module is used for judging whether an event modality exists in the first high-energy signal and the first low-energy signal or not;
an event modality processing module, configured to process the first high energy signal and the first low energy signal according to an event modality processing method when an event modality exists in the first high energy signal and the first low energy signal;
a second decomposition module, configured to directly decompose the first high energy signal and the first low energy signal by using an empirical mode decomposition method when there is no event mode in the first high energy signal and the first low energy signal;
an end effect suppression module, configured to obtain a second high energy signal and a second low energy signal by suppressing an end effect when performing empirical mode decomposition on the first high energy signal and the first low energy signal;
the second judging module is used for judging whether the second low-energy signal has obvious waveform characteristics;
the Hilbert-Huang transform module is used for performing Hilbert-Huang transform on the second high-energy signal and the second low-energy signal to obtain a transform map and ending the whole decomposition process when the second low-energy signal has obvious waveform characteristics;
and the third decomposition module is used for assigning the second low-energy signal to the bearing vibration signal to decompose again when the second low-energy signal has no obvious waveform characteristics until obvious characteristics exist after the low-energy signal is decomposed or the signal to be decomposed does not meet the condition of empirical mode decomposition, then performing Hilbert-Huang transform on the obtained low-energy signal to obtain a transform map, and finishing the whole decomposition process.
Optionally, the first decomposition module specifically includes:
and the first decomposition unit is used for decomposing the bearing vibration signal by using a singular value to obtain a first high-energy signal and a first low-energy signal.
Optionally, the event modality processing module specifically includes:
the time domain determining unit is used for determining the time domain of the generation of the event mode based on the time characteristic scale smoothness and a combined smoothness method based on extreme value difference smoothness;
and the removing unit is used for removing the incident modes of the first high-energy signal and the first low-energy signal by adopting a singular value denoising method according to a time domain.
Optionally, the end-point effect suppression module specifically includes:
and the endpoint effect suppression unit is used for suppressing the endpoint effect by utilizing the principle of monotonicity consistency of signals in each decomposition when empirical mode decomposition is carried out on the first high-energy signal and the first low-energy signal so as to obtain a second high-energy signal and a second low-energy signal.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
according to the method, firstly, according to the fact that bearing vibration signals with different amplitudes or different frequencies have different energies in the same time, energy separation is carried out on the bearing vibration signals to process non-event mode aliasing, then event mode aliasing processing is carried out on the separated bearing vibration signals respectively to obtain signals to be decomposed, and then an end point effect in empirical mode decomposition is eliminated, so that final decomposed signals are obtained. The present invention has the following significant advantages over conventional methods:
(1) aiming at the non-event mode processing method, the obstruction that the frequency cannot be covered under the condition that the target frequency cannot be known is overcome.
(2) The method has higher precision and requires less time than the conventional EEMD method for processing the event mode.
(3) The processing method aiming at the end point effect has higher precision and requires less time compared with the traditional method.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without inventive labor.
FIG. 1 is a flow chart of an improved empirical mode decomposition bearing vibration analysis method in accordance with an embodiment 1 of the present invention;
FIG. 2 is a flow chart of an improved empirical mode decomposition bearing vibration analysis method in accordance with an embodiment 2 of the present invention;
FIG. 3 is a diagram of the end-point effect monotonicity consistent data processing process of the present invention (monotonicity algorithm flow);
FIG. 4 is a graph of comparison of extreme predicted end points;
FIG. 5 is a comparison of results of empirical mode decomposition based on an endpoint effect improvement method (comparison of results of decomposition of three methods);
FIG. 6 is a non-event signal diagram;
FIG. 7 is a signal diagram after removing aliased signals using the EEMD algorithm;
FIG. 8 is a graph of signals obtained after singular value decomposition;
FIG. 9 is a diagram of non-event modality aliased signals;
FIG. 10 is a graph showing the decomposition result of a conventional EMD;
FIG. 11 is a graph showing the result of decomposition of the improved method of the present invention;
FIG. 12 is a plot of data signal frequency for the bearing data center of Kaiser university of Tibet;
FIG. 13 is a Hilbert spectrum of an EMD decomposition;
FIG. 14 is a decomposed Hilbert spectrum of the present invention;
FIG. 15 is a block diagram of an improved EMD bearing vibration analysis system in accordance with the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Example 1:
FIG. 1 is a flow chart of an improved empirical mode decomposition bearing vibration analysis method of the present invention. As shown in fig. 1, an improved empirical mode decomposition bearing vibration analysis method includes:
step 101: decomposing the bearing vibration signal to obtain a first high energy signal and a first low energy signal, and specifically comprising:
and decomposing the bearing vibration signal by using a singular value to obtain a first high-energy signal and a first low-energy signal.
Step 102: and judging whether the first high-energy signal and the first low-energy signal have the event mode or not.
Step 103: if yes, processing the first high-energy signal and the first low-energy signal according to an event mode processing method, which specifically comprises:
a joint stationarity method based on temporal feature scale stationarity and on extreme difference stationarity determines a time domain of generation of the event modality.
And removing the incident modes of the first high-energy signal and the first low-energy signal by adopting a singular value denoising method according to the time domain. This step is mainly to separate high and low energy signals and remove abnormal signals, i.e. event signals.
Step 104: and if not, decomposing the first high-energy signal and the first low-energy signal by adopting an empirical mode decomposition method.
Step 105: when empirical mode decomposition is performed on the first high-energy signal and the first low-energy signal, the second high-energy signal and the second low-energy signal are obtained by suppressing an endpoint effect, and the method specifically includes:
when empirical mode decomposition is carried out on the first high-energy signal and the first low-energy signal, the endpoint effect is restrained by utilizing the principle of monotonicity consistency of the signals during each decomposition, and a second high-energy signal and a second low-energy signal are obtained.
For each decomposition, the principle of monotonicity consistency of signals is utilized to suppress the end effect, and the method specifically comprises the following steps:
step1. extracting maximum value p of vibration signal of signal bearingmaxAnd a minimum value pminThe nearest maximum p at the right endpointmax(end) and minimum value pmin(end); adding the absolute value of the difference between the corresponding maximum value and minimum value to obtain an error set, namely:
err(i)＝|pmax(end)-pmax(i)|+|pmin(end)-pmin(i) and rank the error set from small to large.
step2, solving the monotonicity conditions of the nearest maximum value and the left side of the minimum value of the right end point; namely:
pmax(end)pminand (end) {1 increasing, 2 decreasing, 3 decreasing, 4 decreasing }, comparing the monotonicity of the left side of each point (i) in the required error set, if the monotonicity is consistent, keeping the monotonicity, and if not, deleting the monotonicity.
step3. calculate the monotonicity of the point (i) retained in step2 with the first extreme point to the right, i.e.:
pmax(i+1)pmin(i +1) {1 increase/increase, 2 increase/decrease, 3 decrease/increase, and 4 decrease/decrease }, and comparing the appearance frequencies of the four cases 1 to 4, points with higher appearance frequencies are selected with little difference (1: (4) }<3) Considered equivalent.
step4. select the higher frequency point in step3, construct an equation or system of equations, i.e.:
Ampmax(end)+Am-1pmax(end-1)+....+A0pmax(end-m)
＝pmax(end+1) ....(2)
equation (1), namely: substituting the coefficient A solved by AX ═ B into equation (2) to solve pmax(end +1) whose coordinate is the difference between the minimum error point coordinate in the error set and the latter coordinate plus the end point pmax(end) coordinate determination. The right-end minimum value can be obtained by the same method; after the signal is inverted, the maximum value and the minimum value of the left end can be obtained by the same method (denoising process).
Because the method is obtained by searching for approximate points based on the extreme value error sum method, and part of screening conditions are too strict, no solution can be caused when the data is few, and the following processing is carried out: extracting a maximum value p in step (1)maxAnd a minimum value pminAt the right end, the first maximum pmax(end) and a first minimum value pmin(end) subtracting the remaining maximum and minimum values to obtain difference values, respectivelyThe absolute values are added to obtain an error set, as shown in the formula: err _ max (i) ═ pmax(end)-pmax(i) | and err _ min (i) ═ pmin(end)-pmin(i) And I, arranging the error sets from small to large, and then solving.
Step 106: and judging whether the second low-energy signal has obvious waveform characteristics.
Step 107: if so, performing Hilbert-Huang transformation on the second high-energy signal and the second low-energy signal to obtain a transformation map, and ending the whole decomposition process.
Step 108: and if not, assigning the second low-energy signal to the bearing vibration signal for decomposing again until obvious characteristics exist after the low-energy signal is decomposed or the signal to be decomposed does not meet the conditions of empirical mode decomposition, performing Hilbert-Huang transformation on the obtained low-energy signal to obtain a transformation map, and finishing the whole decomposition process.
Example 2:
fig. 2 is a flowchart of an improved empirical mode decomposition bearing vibration analysis method according to embodiment 2 of the present invention. As shown in fig. 2, an improved empirical mode decomposition bearing vibration analysis method according to embodiment 2 of the present invention specifically includes:
1) separating bearing vibration signal x (t) according to energy by singular value decomposition to obtain high-energy signal xH(t) and a low energy signal xL(t)。
2) Determining a high energy signal xH(t) and a low energy signal xL(t) whether an event mode exists or not, if so, removing according to an event mode processing method to obtain a high-energy signal xH' (t) and a low energy signal xL' (t), if not present, proceed to the next step.
3) During empirical mode decomposition, the endpoint effect is suppressed by using the principle of monotonicity consistency of signals for each decomposition; the method comprises the following specific steps:
step1. extraction of the maximum p of the signal x (t)maxAnd a minimum value pminThe nearest maximum value p at the right endpointmax(end) and minimum value pmin(end); the corresponding maximum value is compared withThe absolute values of the difference between the minima are summed to obtain a set of errors, namely:
err(i)＝|pmax(end)-pmax(i)|+|pmin(end)-pmin(i) and rank the error set from small to large.
step2, solving the monotonicity conditions of the nearest maximum value and the left side of the minimum value of the right end point; namely:
pmax(end)pminand (end) {1 increasing, 2 decreasing, 3 decreasing, 4 decreasing }, comparing the monotonicity of the left side of each point (i) in the required error set, if the monotonicity is consistent, keeping the monotonicity, and if not, deleting the monotonicity.
step3. calculate the monotonicity of the point (i) retained in step2 with the first extreme point to the right, i.e.:
pmax(i+1)pmin(i +1) {1 increase/increase, 2 increase/decrease, 3 decrease/increase, and 4 decrease/decrease }, and comparing the appearance frequencies of the four cases 1 to 4, points with higher appearance frequencies are selected with little difference (1: (4) }<3) Considered equivalent.
step4. select the higher frequency point in step3, construct an equation or system of equations, i.e.:
Ampmax(end)+Am-1pmax(end-1)+....+A0pmax(end-m)
＝pmax(end+1) ....(2)
equation (1), namely: substituting the coefficient A solved by AX ═ B into equation (2) to solve pmax(end +1) whose coordinate is the difference between the minimum error point coordinate in the error set and the latter coordinate plus the end point pmax(end) coordinate determination. The right-end minimum value can be obtained by the same method; after the signal is inverted, the maximum value and the minimum value of the left end can be obtained by the same method (denoising process).
Because the method is obtained by searching approximate points based on the extreme value error sum method, and part of screening conditions are too strict, the method can be used for searching dataRarely may result in no solution, when the following is done: extracting a maximum value p in step (1)maxAnd a minimum value pminAt the right end, the first maximum pmax(end) and a first minimum value pmin(end) subtracting the residual maximum value and the residual minimum value respectively to obtain difference values, and adding the absolute values of the two difference values to obtain an error set, wherein the formula is as follows: err _ max (i) ═ pmax(end)-pmax(i) | and err _ min (i) ═ pmin(end)-pmin(i) And (4) arranging the error sets from small to large, and solving according to the steps (2), (3) and (4). FIG. 3 is a diagram of the endpoint effect monotonicity consistency data processing process of the present invention.
After the endpoint effect is inhibited through the principle of monotonicity consistency, the decomposed high-energy signal IMF is obtainedH(t) and the decomposed low energy signal IMFL(t)。
4) For decomposed signal IMFL(t) judging whether obvious waveform characteristics exist or not, if so, ending the whole decomposition process, and if not, outputting the signal xL(t) assigning to x (t) to decompose again until obvious characteristics exist after the low-energy signal decomposition or the signal to be decomposed does not meet the condition of EMD decomposition, performing Hilbert-Huang transformation, and finishing the whole decomposition process.
Fig. 4 is a comparison diagram of endpoint prediction extrema. FIG. 5 is a comparison of results of empirical mode decomposition based on an endpoint effect refinement method. In fig. 4,' denotes an actual value envelope point; ' the point is an extreme value obtained by an end point extreme value method, and then a point is obtained by adopting cubic spline interpolation, so that the point obtained by fitting the left end maximum value and the right end minimum value is closer to a real value, the point obtained by fitting the left end minimum value and the right end maximum value is farther from the real value, and the end point effect is stronger; the point of '+' refers to a radial neural network RBF three-time spline interpolation curve, the interpolation points of the predicted values of the minimum values at the left end and the right end of the radial neural network RBF three-time spline interpolation curve have smaller deviation from the actual point, the interpolation points of the predicted values of the maximum values at the two ends are very close to the actual result, and the end point effect is weaker; '-' is the result obtained by the monotone consistency algorithm provided by the invention, the interpolation points of the predicted values of the maximum values at the left and right ends have smaller deviation from the actual point, and the interpolation points of the predicted values of the minimum values at the two ends are basically consistent with the actual result, so the end point effect is weaker. On the whole, the monotonicity prediction extreme value and the radial neural network prediction extreme value are better than those of the extreme point which is the data end point, and the extreme value predicted by the monotonicity algorithm and the extreme value predicted by the radial neural network have equal precision. As shown in fig. 5 and table 1, in the comparison of predicting the additional extreme value by using three methods, namely, the method of using the end point data as the additional extreme value, the neural network RBF and the monotonicity, it can be obtained that the method of using the data end point as the additional extreme value has a very good effect of suppressing the end point effect, but when the data amount is small, the error of the additional extreme value increases the number of decomposition layers, and generates a non-existent IMF component; the neural network continuation algorithm has good end effect inhibition result, does not generate nonexistent IMF components, but has obvious defects and slow operation speed; the monotonicity method can inhibit the end effect, no IMF component exists, the operation time is less than one percent of RBF of the neural network, and the decomposition precision is very different from the neural network. Therefore, whether theoretical analysis or experimental analysis shows that the improved algorithm is a method capable of better processing the end point effect. Table 1 shows the time comparison of the improved method of the present invention to the conventional method for treating the endpoint effect. Table 1 is as follows:
TABLE 1 comparison of the time to process endpoint effects for the improved method of the present invention versus the conventional method
Fig. 6 is a non-event signal diagram, as shown in fig. 6, 7, 8 and table 2. FIG. 7 is a signal diagram after removing aliased signals using the EEMD algorithm. Fig. 8 is a diagram of a signal obtained after singular value decomposition, i.e., after the present invention is applied. Table 2 is a run-time comparison of handling event modalities. The EEMD algorithm and the signals obtained after the processing of the invention substantially meet the requirements of eliminating aliasing signals without intermittent signals and abnormal event signals, and the EEMD algorithm has maximum value deviation and minimum value deviation after the processing; the signals after singular value decomposition not only eliminate aliasing signals, but also are almost consistent. And the EEMD run time is more than 1000 times the singular value decomposition program run time. The singular value decomposition removes aliasing signals better than EEMD removes aliasing signals, both from the precision of removing aliasing signals and from a runtime perspective. FIG. 9 is a diagram of non-event modality aliased signals. Table 2 is as follows:
TABLE 2 handling event modality runtime comparison
Fig. 10 and 11 show the same endpoint processing method. Fig. 10 is a decomposition result diagram of the conventional EMD. FIG. 11 is a decomposition result diagram of the improved method of the present invention. Specifically, fig. 10 shows that the EMD decomposes 8 IMF eigenmode functions and one residual component RES, and the IMFs 1-3 have no obvious frequency characteristic due to frequency dispersion, mainly caused by the influence of noise signals; the IMF 4-8 frequencies are more concentrated than the IMF 1-3 frequencies, but target frequency components have no obvious characteristics, and other frequency components appear in the IMF7 and the IMF8, so that the EMD decomposition can be known to have a non-event modal aliasing phenomenon. As can be seen from fig. 11, the improved method of the present invention decomposes 4 IMF eigenmode functions and 1 residual component RES, and two residual components appear due to the energy separation and decomposition method; as can be seen from FIG. 11, the IMFs 1-4 have the same component number as the actual signal, and have prominent frequency characteristics, in the IMF1, the frequency is about 20Hz, the IMF2 frequency is about 10Hz, the IMF3 frequency is about 18Hz, and the IMF4 frequency is about 8 Hz. Comparing and analyzing the EMD and the improved method of the invention, the improved method of the invention can eliminate the non-event mode aliasing phenomenon, and the EMD can not eliminate the non-event mode aliasing phenomenon. The improved method of the present invention is thus superior to EMD decomposition.
Vibration data of bearing data centers of the university of Kaiser storage are selected for analysis. FIG. 12 shows data signal frequencies for the bearing data center of Kaiser university of storageAnd (4) rate graph. As shown in fig. 12, the bearing vibration test platform comprised a 2 hp motor (left), a torque sensor (center), a dynamometer (right) and electronic control equipment (not shown). The bearing under test supports the motor shaft. The bearing model used in the experimental test is 6205-2RSJEM SKF, and the failure diameter is 0.3556 mm. The diameter of the bearing pitch circle is 39.04mm, the diameter D of the steel ball is 7.94mm, the number of rolling elements n is 9, and the pressure angle alpha is 0 deg. The damage of the inner ring is single-point damage of electric spark machining, the diameter of a fault point is 0.3556mm, the depth is 0.2794mm, and the rotation frequency f of the rotating shaftr29.57Hz, the sampling frequency is 12kHz, and the inner ring fault characteristic frequency f is calculatedi＝160.11Hz。
Fig. 13 is a Hilbert spectrum of EMD decomposition. As can be seen from fig. 13, EMD is directly decomposed, the frequency of the sub-map IMF1 is approximately concentrated around 2000Hz, the sub-map IMF2 is approximately concentrated around 1000Hz, and the frequencies of IMF3 and IMF4 are relatively scattered, so that the frequencies from IMF5 to IMF11 are relatively concentrated, but no waveform characteristic is presented, and no characteristic frequency of 160.11Hz is displayed, so that the fault frequency cannot be detected by directly adopting EMD. FIG. 14 is a Hilbert spectrum decomposed by the present invention. As can be seen from the IMFs 1 to 4 first decomposed in fig. 14, the frequencies are relatively concentrated and can be shown to float up and down around a certain frequency, the frequency of the IMF1 floats around 1836Hz, the frequency of the IMF2 is approximately 950Hz, and the frequencies of the IMFs 3 and 4 are relatively dispersed; the IMF obtained by the second decomposition is not much different from the IMF obtained by the first decomposition in a Hilbert spectrum; from the IMF diagram obtained by the third decomposition, it can be known that the IMF1 and the IMF2 are frequency-concentrated and exhibit de-waveform characteristics, the frequency of the IMF1 is around 360Hz, and the frequency of the IMF2 is around 161.2Hz, which is almost consistent with the fault frequency, so that the improved EMD decomposition method provided by the invention is superior to the traditional EMD method.
Example 3:
FIG. 15 is a block diagram of an improved EMD bearing vibration analysis system in accordance with the present invention. As shown in fig. 15, an improved empirical mode decomposition bearing vibration analysis system includes:
the first decomposition module 201 is configured to decompose the bearing vibration signal to obtain a first high energy signal and a first low energy signal.
The first determining module 202 is configured to determine whether an event modality exists in the first high-energy signal and the first low-energy signal.
And the event modality processing module 203 is used for processing the first high-energy signal and the first low-energy signal according to the event modality processing method when an event modality exists in the first high-energy signal and the first low-energy signal.
And a second decomposition module 204, configured to decompose the first high energy signal and the first low energy signal directly by using an empirical mode decomposition method when there is no event mode in the first high energy signal and the first low energy signal.
And the end-point effect suppression module 205 is configured to obtain a second high-energy signal and a second low-energy signal by suppressing an end-point effect when performing empirical mode decomposition on the first high-energy signal and the first low-energy signal.
And a second determining module 206, configured to determine whether the second low energy signal has a significant waveform characteristic.
And the hilbert-yellow transformation module 207 is configured to perform hilbert-yellow transformation on the second high-energy signal and the second low-energy signal to obtain a transformation map when the second low-energy signal has obvious waveform characteristics, and end the whole decomposition process.
And the third decomposition module 208 is configured to assign the second low-energy signal to the bearing vibration signal to decompose again when no obvious waveform feature exists in the second low-energy signal until an obvious feature exists after the low-energy signal is decomposed or the signal to be decomposed does not meet the condition of empirical mode decomposition, perform hilbert-yellow conversion on the obtained low-energy signal to obtain a conversion map, and end the whole decomposition process.
The first decomposition module 201 specifically includes:
and the first decomposition unit is used for decomposing the bearing vibration signal by using a singular value to obtain a first high-energy signal and a first low-energy signal.
The event modality processing module 203 specifically includes:
and the time domain determining unit is used for determining the time domain of the generation of the event mode based on the time characteristic scale smoothness and a combined smoothness method based on extreme value difference smoothness.
And the removing unit is used for removing the incident modes of the first high-energy signal and the first low-energy signal by adopting a singular value denoising method according to the time domain.
The end-point effect suppression module 205 specifically includes:
and the end effect suppression unit is used for suppressing the end effect by utilizing the monotonicity consistency principle of the signals during each decomposition when empirical mode decomposition is carried out on the first high-energy signal and the first low-energy signal so as to obtain a second high-energy signal and a second low-energy signal.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In summary, this summary should not be construed to limit the present invention.
Claims (8)
1. An improved empirical mode decomposition bearing vibration analysis method is characterized by comprising the following steps:
decomposing the bearing vibration signal to obtain a first high energy signal and a first low energy signal;
determining whether an incident modality exists in the first high energy signal and the first low energy signal;
if yes, processing the first high-energy signal and the first low-energy signal according to an event mode processing method;
if not, directly decomposing the first high-energy signal and the first low-energy signal by adopting an empirical mode decomposition method;
obtaining a second high energy signal and a second low energy signal by suppressing an endpoint effect when performing empirical mode decomposition on the first high energy signal and the first low energy signal;
judging whether the second low-energy signal has obvious waveform characteristics;
if so, performing Hilbert-Huang transformation on the second high-energy signal and the second low-energy signal to obtain a transformation map, and ending the whole decomposition process;
and if not, assigning the second low-energy signal to the bearing vibration signal for decomposing again until obvious characteristics exist after the low-energy signal is decomposed or the signal to be decomposed does not meet the condition of empirical mode decomposition, performing Hilbert-Huang transformation on the obtained low-energy signal to obtain a transformation map, and finishing the whole decomposition process.
2. The improved empirical mode decomposition bearing vibration analysis method of claim 1, wherein decomposing the bearing vibration signal to obtain a first high energy signal and a first low energy signal comprises:
and decomposing the bearing vibration signal by using a singular value to obtain a first high-energy signal and a first low-energy signal.
3. The improved EMD bearing vibration analysis method of claim 1, wherein said processing said first high energy signal and said first low energy signal according to an event mode processing method, comprises:
determining a time domain of the generation of the incident mode based on a combined stationarity method of the time characteristic scale stationarity and the extreme value difference stationarity;
and removing the incident modes of the first high-energy signal and the first low-energy signal by adopting a singular value denoising method according to a time domain.
4. The improved method for analyzing vibration of an EMD bearing of claim 1, wherein the obtaining of the second high energy signal and the second low energy signal by suppressing an end-point effect when performing EMD on the first high energy signal and the first low energy signal comprises:
when empirical mode decomposition is carried out on the first high-energy signal and the first low-energy signal, the endpoint effect is restrained by utilizing the principle of monotonicity consistency of the signals during each decomposition, and a second high-energy signal and a second low-energy signal are obtained.
5. An improved empirical mode decomposition bearing vibration analysis system, comprising:
the first decomposition module is used for decomposing the bearing vibration signal to obtain a first high-energy signal and a first low-energy signal;
the first judging module is used for judging whether an event modality exists in the first high-energy signal and the first low-energy signal or not;
an event modality processing module, configured to process the first high energy signal and the first low energy signal according to an event modality processing method when an event modality exists in the first high energy signal and the first low energy signal;
a second decomposition module, configured to directly decompose the first high energy signal and the first low energy signal by using an empirical mode decomposition method when there is no event mode in the first high energy signal and the first low energy signal;
an end effect suppression module, configured to obtain a second high energy signal and a second low energy signal by suppressing an end effect when performing empirical mode decomposition on the first high energy signal and the first low energy signal;
the second judging module is used for judging whether the second low-energy signal has obvious waveform characteristics;
the Hilbert-Huang transform module is used for performing Hilbert-Huang transform on the second high-energy signal and the second low-energy signal to obtain a transform map and ending the whole decomposition process when the second low-energy signal has obvious waveform characteristics;
and the third decomposition module is used for assigning the second low-energy signal to the bearing vibration signal to decompose again when the second low-energy signal has no obvious waveform characteristics until obvious characteristics exist after the low-energy signal is decomposed or the signal to be decomposed does not meet the condition of empirical mode decomposition, then performing Hilbert-Huang transform on the obtained low-energy signal to obtain a transform map, and finishing the whole decomposition process.
6. The improved EMD bearing vibration analysis system of claim 5, wherein the first decomposition module specifically comprises:
and the first decomposition unit is used for decomposing the bearing vibration signal by using a singular value to obtain a first high-energy signal and a first low-energy signal.
7. The improved EMD bearing vibration analysis system of claim 5, wherein the event modality processing module specifically comprises:
the time domain determining unit is used for determining the time domain of the generation of the event mode based on the time characteristic scale smoothness and a combined smoothness method based on extreme value difference smoothness;
and the removing unit is used for removing the incident modes of the first high-energy signal and the first low-energy signal by adopting a singular value denoising method according to a time domain.
8. The improved EMD bearing vibration analysis system of claim 5, wherein the end-point effect suppression module specifically comprises:
and the end effect suppression unit is used for suppressing the end effect by utilizing the principle of monotonicity consistency of signals for each decomposition when empirical mode decomposition is carried out on the first high-energy signal and the first low-energy signal so as to obtain a second high-energy signal and a second low-energy signal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010169517.9A CN112014107A (en) | 2020-03-12 | 2020-03-12 | Improved empirical mode decomposition bearing vibration analysis method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010169517.9A CN112014107A (en) | 2020-03-12 | 2020-03-12 | Improved empirical mode decomposition bearing vibration analysis method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112014107A true CN112014107A (en) | 2020-12-01 |
Family
ID=73506778
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010169517.9A Pending CN112014107A (en) | 2020-03-12 | 2020-03-12 | Improved empirical mode decomposition bearing vibration analysis method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112014107A (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104007315A (en) * | 2014-04-29 | 2014-08-27 | 昆明理工大学 | Improved empirical mode decomposition processing method |
| CN104156623A (en) * | 2014-08-27 | 2014-11-19 | 湖南科技大学 | Diagnosis method of mechanical composite faults |
| CN108229382A (en) * | 2017-12-29 | 2018-06-29 | 广州供电局有限公司 | Vibration signal characteristics extracting method, device, storage medium and computer equipment |
-
2020
- 2020-03-12 CN CN202010169517.9A patent/CN112014107A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104007315A (en) * | 2014-04-29 | 2014-08-27 | 昆明理工大学 | Improved empirical mode decomposition processing method |
| CN104156623A (en) * | 2014-08-27 | 2014-11-19 | 湖南科技大学 | Diagnosis method of mechanical composite faults |
| CN108229382A (en) * | 2017-12-29 | 2018-06-29 | 广州供电局有限公司 | Vibration signal characteristics extracting method, device, storage medium and computer equipment |
Non-Patent Citations (1)
| Title |
|---|
| 李曜洲: "经验模态分解的方法改进研究", 《中国优秀博硕士学位论文全文数据库(硕士)工程科技Ⅱ辑》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Meng et al. | Vibration response and fault characteristics analysis of gear based on time-varying mesh stiffness | |
| Wang et al. | Autoregressive model-based gear fault diagnosis | |
| CN110779724B (en) | Bearing fault diagnosis method based on frequency domain group sparse noise reduction | |
| Jiang et al. | A novel rolling bearing defect detection method based on bispectrum analysis and cloud model-improved EEMD | |
| Li et al. | An EEMD-based denoising method for seismic signal of high arch dam combining wavelet with singular spectrum analysis | |
| Sharma | A review on vibration-based fault diagnosis techniques for wind turbine gearboxes operating under nonstationary conditions | |
| Gelman et al. | Adaptive vibration condition monitoring technology for local tooth damage in gearboxes. | |
| CN113361331B (en) | Power Frequency Interference Elimination Method, System and Medium Based on Windowed Interpolation FFT | |
| CN111259323A (en) | Method for accurately positioning faults of mechanical transmission system of rocker arm of coal mining machine | |
| CN112014107A (en) | Improved empirical mode decomposition bearing vibration analysis method and system | |
| CN111060317A (en) | Method for judging fault signal of rolling bearing of mining fan motor | |
| JP5958932B2 (en) | Dynamic equipment condition monitoring system, method and program | |
| JPH11124096A (en) | Monitoring method for vehicle with acceleration sensor, in particular, power transmission assembly of helicopter | |
| Zhao et al. | Peak envelope spectrum Fourier decomposition method and its application in fault diagnosis of rolling bearings | |
| CN112067297B (en) | Bearing fault feature extraction method | |
| Pereira et al. | Detecting punctual damage to gears through the continuous Morlet wavelet transform | |
| Gravel | Analysis of ripple on noisy gears | |
| Wen et al. | A composite method of marine Shafting’s fault diagnosis by ship hull vibrations based on EEMD | |
| Ma et al. | Envelope demodulation method based on SET for fault diagnosis of rolling bearings under variable speed | |
| Mohamad et al. | Effect of variations in tooth flank form among teeth on gear vibration and an sensory evaluation method using potential gear noise | |
| Damine et al. | Enhancement of the AR-MED Deconvolution Process using Ensemble Empirical Mode Decomposition in Bearing Fault Detection | |
| Wang et al. | Kernel Regression Residual Decomposition Method to Detect Rolling Element Bearing Faults | |
| Liu et al. | Research of Crack Detection and Delay Crack Propagation on a Nonlinear Rotor. | |
| D’Elia et al. | On the use of Fourier-Bessel series expansion for gear diagnostics | |
| CN110646200B (en) | Method for diagnosing weak local fault of gear of fixed shaft gear box |
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 |