CN214036158U - Water pump fault early warning device based on entropy weight method - Google Patents
Water pump fault early warning device based on entropy weight method Download PDFInfo
- Publication number
- CN214036158U CN214036158U CN202022410657.9U CN202022410657U CN214036158U CN 214036158 U CN214036158 U CN 214036158U CN 202022410657 U CN202022410657 U CN 202022410657U CN 214036158 U CN214036158 U CN 214036158U
- Authority
- CN
- China
- Prior art keywords
- water pump
- signal
- early warning
- detected
- characteristic
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Abstract
The utility model relates to a water supply equipment failure diagnosis technical field, along with artificial intelligence in the infiltration of industry each side, the failure diagnosis to water pump package has emerged multiple intelligent algorithm, mainly uses neural network etc. to establish the fault identification model of water pump package as leading, but this kind of algorithm need be equipped with huge water pump package control parameter signal storehouse, and this just causes that the in-process difficulty of establishing complete database is great, and this method is in the experimental phase. The utility model discloses a water pump trouble early warning device based on entropy weight method, include: the method comprises the following steps that a data storage unit, a feature matrix unit electrically connected with the data storage unit and a weight unit electrically connected with the feature matrix unit extract a signal X to be detected and a reference signal Y, and establish a signal feature matrix A to be detected and a reference signal feature matrix B for the signal X to be detected and the reference signal Y respectively; and constructing an early warning index a and a threshold value matrix alfa, judging whether the water pump set is normal or not, and establishing a state table Tab.
Description
Technical Field
The utility model relates to a water supply equipment failure diagnosis technical field specifically is a water pump fault early warning device based on entropy weight method. The core idea is that an early warning index is calculated through a signal to be detected and a reference signal, a threshold value matrix of the early warning index is constructed, and finally whether the water pump set fails or not is comprehensively judged through comparing and analyzing the early warning index and the threshold value of the early warning index.
Background
The water pump plays an important role in a water supply system as an important water supply device, and the dynamic balance state of the whole water pump is one of important indexes for checking whether the running state of the water pump is good or not. In the process of using and managing actual equipment, measures such as oil filling lubrication, cleaning and the like are mostly adopted for post management or preventive maintenance of the equipment, and the whole dynamic balance condition is in an out-of-control state. Once the dynamic balance of the water pump is affected, the long-term operation can deepen the fault degree of components such as an impeller and the like, even affect the service life of the whole water pump,
with the penetration of artificial intelligence in various industrial aspects, various intelligent algorithms are developed for fault diagnosis of the water pump, mainly a fault identification model of the whole water pump set is established by a neural network and the like, but the algorithm needs to be complete enough in acquisition signals and parameter base of the water pump, enough in fault types, and in consideration of actual conditions, the establishment process of a complete data set is difficult, and the method is mostly in a laboratory stage, and few in actual application cases.
Compared with the establishment of a fault identification model, the device provided by the patent is simple and convenient to operate, small in calculated amount, free of requirement on a complete fault blade database, and more suitable for application in practical engineering.
Disclosure of Invention
An object of the utility model is to provide a water pump trouble early warning device based on entropy weight method considers a series of actual problems, adopts the signal when the water pump group is normal, but establishes the threshold value matrix of the biggest early warning index when different rotational speeds, different water pumps are normal. When the water pump is diagnosed and pre-warned, the pre-warning index value of the water pump is calculated, and whether the water pump is pre-warned or not is indicated by comparing and analyzing the threshold value of the water pump at the same rotating speed. Further, an energy characteristic matrix is extracted from normal dynamic balance signals of the water pump set by utilizing wavelet transformation, a threshold value matrix of a fault early warning index is established based on an entropy weight method, the early warning index of the water pump to be detected obtained through calculation is compared and analyzed with a corresponding threshold value, the number of layers of wavelet decomposition is changed, and the above steps are repeated, and finally, the task of whether the water pump to be detected carries out fault early warning or not is achieved through comprehensive analysis.
The technical proposal provided by the utility model is that the proposal comprises,
the system comprises a data storage unit, a feature matrix unit and a weight unit, wherein the data storage unit is electrically connected with the feature matrix unit, and the feature matrix unit is electrically connected with the weight unit;
the data storage unit is used for periodically storing and updating dynamic balance signals of all water pumps of a water pump set, which are extracted by the water pump set under the fault-free condition; the dynamic balance signal comprises the dynamic balance of all water pumps in the water pump set at different rotating speeds; extracting the dynamic balance signal as a standby reference signal for early warning of faults of all water pumps in the water pump group at different rotating speeds;
the characteristic matrix unit extracts a signal X to be detected and a reference signal Y from the water pump set, respectively performs wavelet decomposition on the signal X to be detected and the reference signal Y, extracts respective energy coefficient characteristic attribute values, and each energy coefficient characteristic attribute value forms an energy coefficient characteristic attribute characteristic dimension so as to respectively construct a signal characteristic matrix A to be detected and a reference signal characteristic matrix B;
the weighting unit respectively extracts the energy coefficient characteristic attribute values of the same position of the nth column of the signal characteristic matrix A to be measured and the reference signal characteristic matrix B to form an m × 2 characteristic matrix Q, wherein the total number of the characteristic matrices is K, n is {1,2,3, … … K }, K is the number of layers of wavelet decomposition, and m is the number of segments of signal division; respectively calculating the characteristic weight values of the first dimension and the second dimension of the energy coefficient characteristic dimension in the K characteristic matrixes Q by using an entropy weight method, wherein the characteristic weight values are respectively bn1And bn2；
Constructing an index a of the water pump group fault early warning for quantitatively describing the dynamic balance of the water pump, wherein the formula is as follows:
and constructing a threshold value matrix alfa by the fault early warning index a of the water pump group, traversing the threshold value matrix alfa according to the serial number and the dynamic balance of the water pump to be detected, and determining the early warning threshold value a of the water pump to be detectedijAnd the water pump set fault early warning index a, and establishing a water pump set state information table Tab;
and adjusting the wavelet decomposition layer number K value and the change times thereof, recording the calculation result of each time in a water pump dynamic balance information table Tab in the water pump set, and judging the running conditions of all water pumps in the water pump set at different rotating speeds.
Further, the signal to be measured X and the reference signal Y are set with a certain number of sampling points as sample lengths, defined as wlen, and then the signal to be measured X and the reference signal Y are divided into m segments of signals, that is, the signals are
Wherein m is an integer part of the result in the formula, the number of sampling points of wlen at least comprises dynamic balance sampling points of each rotating speed section of the water pump set, X, Y represents the signal to be detected and the reference signal, wlen satisfies the condition
Wherein w is a water pumpThe angular velocity f is the working frequency of the water pump.
Further, the wavelet decomposition is to calculate the energy coefficient after the decomposition of the reference signal Y and the signal X to be measured, i.e. E, by using a wavelet energy formulan＝∑xn 2Wherein xnEach decomposed signal segment corresponds to a reference value, n is {1,2,3.. K }, and K is the number of layers of wavelet decomposition of the signal; the signal X to be detected and the reference signal Y are respectively divided into m sections of signals, each section of signal is respectively subjected to K-layer wavelet decomposition, the signal energy coefficient after each section of wavelet decomposition is an energy coefficient characteristic attribute value, the signal energy coefficient characteristic attribute after each section of wavelet decomposition is composed of energy coefficient characteristic dimensions, and the m sections of energy coefficient characteristic dimensions are composed of a signal characteristic matrix A to be detected and a reference signal characteristic matrix B.
Further, an entropy weight method is adopted for respectively calculating the feature weight values of the feature attributes of the first dimension and the second dimension in the K feature matrices Q, namely, normalization is firstly carried out according to each feature attribute value in the feature matrices Q, namely
Where i ═ 1,2,3.. ·. m }, j ═ 1,2}, min (X)i) Is the minimum value, max (X), of the energy coefficient characteristic attribute value corresponding to the dynamic balance signal of the ith sectioni) Is the maximum value, X, of the energy coefficient characteristic attribute value corresponding to the dynamic balance signal of the ith sectionijAnd obtaining a normalized matrix Q' of the characteristic matrix Q for the energy coefficient characteristic attribute value corresponding to the dynamic balance signal of the ith dimension.
Further, the construction of the water pump fault early warning index a is that
Wherein b isn1And bn2And respectively representing the characteristic weight values of the energy coefficient characteristic attributes in the first dimension and the second dimension in the characteristic matrix Q.
Further, the threshold value matrix alfa is respectively stored in the dynamic balance signal storage units of three time periodsExtracting dynamic balance signals of the same rotating speed and the same water pump, calculating the fault early warning indexes a of the water pump groups according to 2 to 4, and respectively using a1、a2、a3If the rotation speed of the water pump is not within the preset rotation speed range, the fault early warning index a of the water pump at the rotation speed is the maximum value of the three values; and (3) performing the same treatment on other water pumps in the water pump group to obtain a threshold value matrix alfa of the early warning index a:
wherein u is the total number of the water pumps in the water pump set, v is the number of the water pumps in the water pump set under different rotating speeds, and the value range of i is [1, u]Is an integer between, j has a value in the range of [1, v]Is an integer of (1).
Further, the early warning threshold value a of the water pump to be detectedijComparing the water pump set with an index a of fault early warning of the water pump set, and when a is less than aijAnd if so, the dynamic balance of the water pump to be detected is normal, otherwise, the early warning index of the water pump group to be detected exceeds the early warning threshold value of the water pump group, the corresponding water pump group state information table Tab is established, and the diagnosis result is recorded.
Further, according to the energy coefficient in the normalization matrix Q', a calculation formula is used
Wherein
If p isijIs defined when 0
And obtaining the characteristic weight value in the normalization matrix Q'.
Further, the signal X to be detected is a section of dynamic balance signal sent by a certain water pump in the water pump set at a certain fixed rotating speed; the reference signal Y is a dynamic balance signal of the water pump corresponding to the signal to be detected X at the same rotating speed, wherein the dynamic balance signal traverses the standby reference signal in the data storage unit, and the holding time and the sample length of the two signals of the signal to be detected X and the reference signal Y are consistent.
Further, performing K-layer wavelet decomposition on the same section of dynamic balance signal in the two signals of the signal to be detected X and the reference signal Y respectively, wherein K is a constant set manually, and extracting respective energy coefficients after decomposition by using a wavelet energy formula to generate a corresponding signal to be detected feature dimension X1 and a corresponding reference signal feature dimension Y1 respectively; and performing the same processing on the rest sections of the dynamic balance signals, and respectively generating a characteristic matrix A of the signal to be detected and a characteristic matrix B of the reference signal, wherein the sizes of the characteristic matrix A and the characteristic matrix B are both mxK, m is the number of rows of the characteristic matrix, namely the number of sections of the water pump, and K is the number of columns of the characteristic matrix, namely the number of layers of wavelet decomposition.
The utility model discloses an aspect of the technical effect that technical scheme brought lies in, has avoidd and has adopted the difficulty degree of establishing complete data set with neural network etc. in the whole electric wire netting trouble recognition model in-process, has improved the accuracy and the stability of water pump group trouble early warning, more presses close to practical application.
Drawings
Fig. 1 is a schematic diagram of the relationship between control modules of a water pump fault early warning device based on an entropy weight method according to the present invention;
fig. 2 is a flow chart of the water pump fault early warning device based on the entropy weight method of the present invention;
the method comprises the following steps of 100, a data storage unit 101, a feature matrix unit 102 and a weight unit.
Detailed Description
Example 1
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 some embodiments of the present invention, not all embodiments. Based on the embodiments in the present invention, all other embodiments obtained by a person skilled in the art without creative work belong to the protection scope of the present invention.
The specific implementation method is as follows:
as shown in fig. 1 and fig. 2, it is that the utility model relates to a flow chart of water pump fault early warning device based on entropy weight method, it includes:
a data storage unit 100, a feature matrix unit 101 and a weight unit 102, wherein the data storage unit 100 is electrically connected with the feature matrix unit 101, and the feature matrix unit 101 is electrically connected with the weight unit 102;
the data storage unit 100 periodically stores and updates dynamic balance signals of all water pumps of a water pump group, which are extracted by the water pump group under the fault-free condition; the dynamic balance signal comprises the dynamic balance of all water pumps in the water pump set at different rotating speeds; extracting the dynamic balance signal as a standby reference signal for early warning of faults of all water pumps in the water pump group at different rotating speeds;
aiming at a power grid, establishing a data storage unit of a dynamic balance signal of a water pump set, wherein the collected water pumps comprise dynamic balance signals of all water pumps of the power grid under different rotating speed scales, the dynamic balance signals are used as standby reference signals when the water pump set is subjected to fault early warning, the storage unit needs to be completely updated at regular intervals, and the time period is measured by taking months as a unit;
at a fixed moment, when fault diagnosis is carried out on a water pump group, a section of dynamic balance signal sent by the dynamic balance of the water pump at the moment is collected as a signal X to be measured, a certain number of sampling points are set as sample lengths, wherein the sampling points are defined as wlen, then the signal to be measured can be divided into m sections of dynamic balance signals, and a specific calculation formula is as follows:
wherein m is an integer part of the result in the formula, the number of sampling points of wlen at least comprises dynamic balance sampling points of each rotating speed section of the water pump set, X, Y represents the signal to be detected and the reference signal, wlen satisfies the condition
Searching a dynamic balance signal of the same water pump and the signal to be detected at the same rotating speed in a reference signal storage unit, wherein the time and the sample length of the reference signal are consistent with those of the signal to be detected;
performing K-layer wavelet decomposition on each section of dynamic balance signals of the two signals, wherein K is a constant set artificially, and extracting an energy coefficient after decomposition by using a wavelet energy formula, wherein the calculation formula is as follows:
En＝∑xn 2 (3)
wherein xnFor each decomposed signal, n ═ 1,2,3.. K }, where K is the number of layers in the wavelet decomposition and K ═ 1,2,3, … … K }.
Therefore, each section of dynamic balance signal can be represented by a vector generated by an energy coefficient, and energy characteristic matrixes A and B capable of representing the essence of two signals can be extracted from the signal to be measured and the reference signal through the transformation, wherein the large size of the two matrixes is m multiplied by K, namely the number of rows of the matrixes is the number of sections of the water pump, and the columns of the matrixes are the number of layers of wavelet decomposition;
respectively extracting the energy coefficient characteristic attribute values of the same position of the nth column of the signal characteristic matrix A to be detected and the reference signal characteristic matrix B to form an m × 2 characteristic matrix Q, wherein the total number of the characteristic matrices is K, n is {1,2,3, … … K }, K is the number of layers of wavelet decomposition, and m is the number of segments of signal division; respectively calculating the characteristic weight values of the first dimension and the second dimension of the energy coefficient characteristic dimension in the K characteristic matrixes Q by using an entropy weight method, wherein the characteristic weight values are respectively bn1And bn2。
The entropy weight method used for calculating the characteristic weight values of the first dimension and the second dimension in the K matrixes Q respectively is to normalize according to each energy coefficient value in the characteristic matrix Q, namely to normalize
Where i ═ 1,2,3... m }, j ═ 1,2}, m ═ jin(Xi) Is the minimum value, max (X), of the energy coefficient corresponding to the dynamic balance signal in the ith sectioni) Is the maximum value, X, of the energy coefficient corresponding to the dynamic balance signal in the ith sectionijAnd normalizing the energy coefficient corresponding to the dynamic balance signal of the ith dimension to obtain a data normalization table so as to obtain a normalization matrix Q' of the characteristic matrix Q. According to the energy coefficient in the normalization matrix Q', a calculation formula is used
Wherein
If p isijIs defined when 0
And obtaining the characteristic weight value in the normalization matrix Q'.
Constructing an index a of the water pump set fault early warning of the water pump set for quantitatively describing the dynamic balance of the water pump, wherein a formula is as follows:
and constructing a threshold value matrix alfa by the water pump set fault early warning index a.
A threshold value matrix alfa of the early warning index a is constructed by utilizing a storage unit of the dynamic balance signal,
and establishing a storage unit for dynamic balance signals in three time periods, respectively taking out the dynamic balance signals of the same water pump at the same rotating speed from the three storage units, calculating the a indexes of the dynamic balance signals according to 2 to 4, and respectively representing the values of the three indexes which can be calculated under the fixed working condition of the water pump by a1, a2 and a3, wherein the fault early warning index of the water pump at the rotating speed is the maximum value of the three indexes. The threshold value matrix alfa of the early warning index a can be obtained by performing the same processing on different water pumps:
in the formula (5), u is the total number of the water pumps of a certain power grid, v is the number of different rotating speed scales of the power grid, the value range of i is an integer between [1 and u ], the value range of j is an integer of [1 and v ],
determining the early warning threshold value of the water pump from alfa according to the serial number and the dynamic balance condition of the water pump to be detected, and when the calculated a is less than aijIf so, the dynamic balance of the water pump is normal, otherwise, early warning is carried out, a water pump equipment manager is reminded that a water pump group possibly has faults, and the obtained result is recorded in a water pump dynamic balance state information table Tab;
changing the value of the wavelet decomposition layer number K, the wavelet decomposition layer numbers involved in 2 to 6 are required to be kept consistent, the number of times of circulation of the whole process can be realized by setting the change number of times of the K value, each result is recorded in a water pump dynamic balance state information table Tab, if the early warning number of times in the table is larger than the normal number of times, the water pump gives an early warning to the water pump fault, otherwise, the water pump dynamic balance is displayed to be normal, and therefore the fault early warning of the water pump set is realized.
Example 2
The embodiment further describes an implementation process and cautions of the water pump fault early warning device based on the entropy weight method by combining with an example of a field.
In the field of a certain water pump, 5 water pump groups are counted, and a water pump group fault early warning system needs to be established for all water pumps in the 5 water pump groups by using an entropy weight method and is used for managing field water pump equipment. The specific method comprises the following steps:
1. extracting a dynamic balance signal Y to be detected and a reference dynamic balance signal X at a certain rotating speed, and dividing the dynamic balance signal Y to be detected and the reference dynamic balance signal X into 4 sections for 3-layer wavelet decomposition, so that the detected reference dynamic balance signal Y and the detected dynamic balance signal X are as follows:
the reference dynamic balance signal to be detected
The reference dynamic balance signal
2. Respectively extracting the energy coefficients of the nth columns of the signal characteristic matrix A to be detected and the reference signal characteristic matrix B, and combining the energy coefficients into an m × 2 matrix Q, wherein n is {1,2,3, … … K }, and K matrices Q are counted, where m is 4 and K is 3, so that the characteristic matrices Q are respectively:
Constructing an index a of the water pump fault early warning for quantitatively describing the dynamic balance of the water pump, wherein a calculation formula is as follows:
and a | + | b11-b12| + | b21-b22| + | b31-b32| + | 0.012+0.278+0.16 | -0.45, and the fault early warning index of the water pump group is 0.45.
3. And constructing a threshold value matrix alfa of the early warning index a. And establishing a storage unit of the dynamic balance signals in three time periods, and adopting the dynamic balance signals of the same water pump at the same rotating speed, calculating the early warning index a according to the above, and calculating the values of the three indexes under the fixed working condition of the water pump, wherein the values are respectively represented by a1, a2 and a3, and the fault early warning index of the water pump at the rotating speed is the maximum value of the three indexes. The specific process is as follows, assuming that the water pump is No. 1, in three time periods, namely time period 1, time period 2 and time period 3, respectively, under the condition of no fault of the water pump and under the actual operation state, a section of dynamic balance signal is extracted, and the section of dynamic balance signal is divided into 4 sections to be subjected to 3-layer wavelet decomposition, and a series of operations are carried out to obtain corresponding early warning indexes a1, a2 and a 3. The measured reference dynamic balance signal Y and the measured dynamic balance signal X are as follows:
the Q matrix corresponding to each group of dynamic balance signals extracts the same column energy coefficient of the reference dynamic balance signal X and the dynamic balance signal Y to be measured, and combines the Q matrix as follows:
carrying out entropy weight normalization processing on each Q matrix to obtain each normalization matrix, wherein the normalization matrixes are as follows:
according to a calculation formula
Wherein
If p isijIs defined when 0
The characteristic weight values in the normalization matrix Q' are derived, i.e. bi1, bi 2. The details are set forth in the following table:
the index a of the fault early warning is used for quantitatively describing the dynamic balance of the water pump, and the calculation formula is as follows:
The operation is also carried out on other water pumps, and the entropy weight method is set for establishing a water pump group fault early warning system for the field 5 water pumps for the management of field water pump equipment. Extracting a reference dynamic balance signal Y and a dynamic balance signal X to be detected at a certain rotating speed, dividing the reference dynamic balance signal Y and the dynamic balance signal X to be detected into 4 sections for 3-layer wavelet decomposition, and sampling a water pump at three speed sections, wherein v is 3, u is 5, and the method specifically comprises the following steps:
The data are presented in the following table, as follows:
| water pump set | Rotational speed 1 | Rotational speed 2 | Rotational speed 3 |
| No. 1 water pump | 0.66 | 0.45 | 0.42 |
| No. 2 water pump | 0.64 | 0.42 | 0.27 |
| No. 3 water pump | 0.78 | 0.38 | 0.56 |
| No. 4 water pump | 0.45 | 0.31 | 0.69 |
| No. 5 water pump | 0.38 | 0.39 | 0.36 |
Calculating a fault early warning index 0.45 of the water pump group according to the step 2 in the embodiment, comparing the fault early warning actual values of the water pumps of the water pump group in the upper table at different rotating speeds, namely determining an early warning threshold value of each water pump from alfa according to the serial number and the dynamic balance condition of the water pump to be detected, and when the calculated a is less than aijAnd if not, early warning is carried out to remind a water pump equipment manager that a water pump group possibly has a fault, and the obtained result is recorded in a water pump dynamic balance state information table Tab as follows:
finally, the adjustment of the K value and the number of times, namely the number of layers of wavelet decomposition, is tried, and each result is recorded in a water pump dynamic balance state information table Tab, as shown in the following table, for early warning of the water pump fault of the water pump set.
Example 3
The difference between this embodiment and embodiment 2 is that the characteristic attribute weight in the normalization matrix Q' is subjected to secondary entropy weighting by using an entropy weighting method, and the difference and the relation between the final result and embodiment 1 are compared. The following table shows the characteristic weights in the normalized matrix Q' obtained in example 2:
according to weight calculation formula of each feature dimension of characteristic attributes of water pump sample
The weights of the feature dimensions in the feature matrix Y can be obtained as shown in the following table:
formula is used for index a of fault early warning at the moment
The calculated a1, a2 and a3 are respectively: 0.05, 0.08 and 0.03, and the fault early warning index of the No. 1 water pump at the rotating speed is the maximum value of the three indexes, namely 0.08. The operation processing is also carried out on the water pumps, and the entropy weight method is set for establishing a water pump group fault early warning system for the field 5 water pumps for the management of field water pump equipment. Extracting a reference dynamic balance signal Y and a dynamic balance signal X to be detected at a certain rotating speed, dividing the reference dynamic balance signal Y and the dynamic balance signal X to be detected into 4 sections for 3-layer wavelet decomposition, and sampling a water pump at three speed sections, wherein v is 3, u is 5, and the method specifically comprises the following steps:
As embodied in the following table:
| water pump set | Rotational speed 1 | Rotational speed 2 | Rotational speed 3 |
| No. 1 water pump | 0.35 | 0.02 | 0.06 |
| No. 2 water pump | 0.78 | 0.05 | 0.03 |
| No. 3 water pump | 0.09 | 0.02 | 0.08 |
| No. 4 water pump | 0.02 | 0.03 | 0.35 |
| No. 5 water pump | 0.01 | 0.06 | 0.08 |
According to the fault early warning index 0.08 of the water pump group of the water pump set in the embodiment, the fault early warning actual values of the water pumps of the water pump group in the reference table under different rotating speeds are compared, namely the early warning threshold value of each water pump is determined from alfa according to the serial number and the dynamic balance condition of the water pump to be detected, and when a is calculated to be less than aijAnd if not, early warning is carried out to remind a water pump equipment manager that a water pump group possibly has a fault, and the obtained result is recorded in a water pump dynamic balance state information table Tab as follows:
finally, the adjustment of the K value and the number of times, namely the number of layers of wavelet decomposition, is tried, and each result is recorded in a water pump dynamic balance state information table Tab, as shown in the following table, for early warning of the water pump fault of the water pump set.
According to the calculation result of the embodiment, the results of extracting the secondary weights of the energy coefficient matrix characteristic dimensions of all the water pump groups based on the entropy weight method are consistent with those of the embodiment 2, that is, the water pump faults of the water pump groups can be early-warned and judged by solving the characteristic attribute weight of the normalization matrix or the secondary weight of the characteristic dimension of the normalization matrix by the entropy weight method.
Example 4
In this embodiment, whether the sequence of the reference dynamic balance signal Y or the dynamic balance signal X to be detected corresponding to the first feature dimension and the second feature dimension affects the final determination result will be further described. Similarly, assuming that all water pumps of a certain water pump group in a certain water pump field are 5, performing 3-layer wavelet decomposition processing on the feature matrix extracted by all the water pump groups of the 5 water pumps, and obtaining the following results:
the dynamic balance signal to be measured
The reference dynamic balance signal
Respectively extracting the energy coefficients of the nth columns of the signal characteristic matrix to be detected and the reference signal characteristic matrix, and combining the energy coefficients into an m × 2 matrix Q, wherein n is {1,2,3, … … K }, K matrices Q are counted, where m is 4 and K is 3, so that the characteristic matrices Q are respectively:
at this time, the characteristic dimension of the characteristic matrix of the signal to be detected is a first characteristic dimension of the characteristic matrix Q, and the characteristic dimension of the characteristic matrix of the reference signal is a second characteristic dimension of the characteristic matrix Q. Actually, an index a of fault early warning is obtained to quantitatively describe the dynamic balance of the water pump, and a calculation formula is as follows:
wherein b isn1And bn2And respectively representing the characteristic weight values of the energy coefficient characteristic attributes of the first dimension and the second dimension in the characteristic matrix Q, and knowing that the sequence of the first characteristic dimension and the second characteristic dimension of the characteristic matrix Q respectively corresponding to the reference dynamic balance signal Y or the dynamic balance signal X to be detected has no influence on the water pump set fault early warning index a.
Claims (2)
1. The utility model provides a water pump trouble early warning device based on entropy weight method which characterized in that includes:
the system comprises a data storage unit, a feature matrix unit and a weight unit, wherein the data storage unit is electrically connected with the feature matrix unit, and the feature matrix unit is electrically connected with the weight unit;
the data storage unit receives and stores dynamic balance signals of all water pumps of a water pump set extracted by the water pump set under the condition of no fault so as to output standby reference signals when all the water pumps perform fault early warning;
the characteristic matrix unit receives a signal X to be detected and a reference signal Y in the water pump set, respectively carries out wavelet decomposition on the signal X to be detected and the reference signal Y, extracts characteristic attribute values of respective energy coefficients, and further respectively constructs and outputs a characteristic matrix A of the signal to be detected and a characteristic matrix B of the reference signal;
and the weight unit receives a characteristic matrix Q consisting of the characteristic matrix A of the signal to be detected and the characteristic matrix B of the reference signal, and respectively calculates and obtains the characteristic weight values of a first dimension and a second dimension in the K layers of characteristic matrix Q of wavelet decomposition by using an entropy weight method so as to judge the running conditions of all water pumps in the water pump group at different rotating speeds.
2. The entropy weight method-based water pump fault early warning device according to claim 1, wherein: the signal X to be detected is a section of dynamic balance signal sent by a certain water pump in the water pump set at a certain fixed rotating speed; the reference signal Y is the spare reference signal in the traversal data storage unit.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202022410657.9U CN214036158U (en) | 2020-10-27 | 2020-10-27 | Water pump fault early warning device based on entropy weight method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202022410657.9U CN214036158U (en) | 2020-10-27 | 2020-10-27 | Water pump fault early warning device based on entropy weight method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN214036158U true CN214036158U (en) | 2021-08-24 |
Family
ID=77354651
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202022410657.9U Active CN214036158U (en) | 2020-10-27 | 2020-10-27 | Water pump fault early warning device based on entropy weight method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN214036158U (en) |
-
2020
- 2020-10-27 CN CN202022410657.9U patent/CN214036158U/en active Active
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107169426B (en) | Crowd emotion abnormality detection and positioning method based on deep neural network | |
| CN115578015B (en) | Sewage treatment whole process supervision method, system and storage medium based on Internet of things | |
| CN109063308B (en) | Health assessment method based on deep quantum learning | |
| CN111680820B (en) | Distributed photovoltaic power station fault diagnosis method and device | |
| CN109829236A (en) | A kind of Compressor Fault Diagnosis method based on XGBoost feature extraction | |
| CN111340063B (en) | Data anomaly detection method for coal mill | |
| CN111753893A (en) | Wind turbine generator power cluster prediction method based on clustering and deep learning | |
| CN111950585A (en) | XGboost-based underground comprehensive pipe gallery safety condition assessment method | |
| CN112287018B (en) | 10kV pole tower damage risk assessment method and system under typhoon disaster | |
| CN112818604A (en) | Wind turbine generator risk degree assessment method based on wind power prediction | |
| CN111080982A (en) | Dam safety intelligent monitoring and early warning system and method based on multiple sensors | |
| CN106875037A (en) | Wind-force Forecasting Methodology and device | |
| CN114252103B (en) | Fusion power station operation fault prediction method | |
| CN111723839A (en) | Method for predicting line loss rate of distribution room based on edge calculation | |
| CN109299208B (en) | Intelligent visual risk assessment method for transmission tower in typhoon disaster | |
| CN116050599A (en) | Line icing fault prediction method, system, storage medium and equipment | |
| CN112363012A (en) | Power grid fault early warning device and method | |
| CN214036158U (en) | Water pump fault early warning device based on entropy weight method | |
| CN117235934A (en) | Degradation monitoring and degradation grade identification method for centrifugal pump | |
| CN117421994A (en) | Edge application health monitoring method and system | |
| CN116030955B (en) | Medical equipment state monitoring method and related device based on Internet of things | |
| CN115935285A (en) | Multi-element time series anomaly detection method and system based on mask map neural network model | |
| CN107977727B (en) | Method for predicting blocking probability of optical cable network based on social development and climate factors | |
| CN113268552A (en) | Generator equipment hidden danger early warning method based on locality sensitive hashing | |
| CN112302976B (en) | Fan blade fault early warning method based on entropy weight method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| GR01 | Patent grant | ||
| GR01 | Patent grant |