CN112051836A - Underwater robot propeller state monitoring method based on multi-core model - Google Patents
Underwater robot propeller state monitoring method based on multi-core model Download PDFInfo
- Publication number
- CN112051836A CN112051836A CN202010951862.8A CN202010951862A CN112051836A CN 112051836 A CN112051836 A CN 112051836A CN 202010951862 A CN202010951862 A CN 202010951862A CN 112051836 A CN112051836 A CN 112051836A
- Authority
- CN
- China
- Prior art keywords
- fault
- sample
- monitoring
- training
- target
- 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
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0259—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterized by the response to fault detection
- G05B23/0275—Fault isolation and identification, e.g. classify fault; estimate cause or root of failure
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Devices For Executing Special Programs (AREA)
- Monitoring And Testing Of Nuclear Reactors (AREA)
Abstract
The invention discloses a multinuclear model-based state monitoring method for a propeller of an underwater robot, which comprises the steps of respectively taking training samples corresponding to different fault grades as target samples, respectively training a plurality of hypersphere models, then taking training samples corresponding to other fault grades except the target grades as non-target samples, calculating monitoring coefficients, establishing a plurality of single-core identification models, and finally integrating the plurality of single-core identification models into a multinuclear identification model. In the invention, the training samples corresponding to a certain fault grade are respectively used as target samples or non-target samples at different stages, so that the fault sample information is fully utilized, and the identification precision is improved.
Description
Technical Field
The invention relates to a method for monitoring the state of a propeller of an underwater robot, in particular to a method for monitoring the state of the propeller of the underwater robot based on a multi-core model.
Background
With the increase of the development and utilization of marine resources, the related technical field of the underwater robot is rapidly developed, and the propeller is used as a core component of the underwater robot and plays a key role in the normal work of the underwater robot, so that the related research on the propeller of the underwater robot is carried out. Of these, the more important is the underwater robot propeller state monitoring technology. By monitoring the underwater robot propeller, the working state of each propeller is reflected in real time, and the underwater robot can be ensured to complete operation and recovery smoothly. Therefore, it is important to monitor the state of the underwater robot propeller.
Chinese patent 201710185935.5 discloses a method for monitoring the state of an underwater robot based on fuzzy support vector field description, which comprises the steps of taking a fault sample when a propeller normally works as a target sample, introducing a fuzzy membership coefficient of the fault sample into the target sample, training a hypersphere, establishing a state monitoring model, taking the fault sample when the running state of the propeller is to be measured as a sample to be measured, introducing the fuzzy membership coefficient into the sample to be measured, bringing the sample to be measured into the monitoring model, and monitoring the running state of the propeller of the underwater robot. This patent not only can judge whether the propeller breaks down, can judge propeller trouble severity moreover. In addition, the monitoring coefficient of this patent increases along with the increase of trouble degree, and the monitoring coefficient is monotonous mapping relation with the trouble degree promptly for a monitoring coefficient only corresponds a trouble degree, is favorable to carrying out propeller trouble degree and discerns, but this patent only regards as the target sample with the trouble sample when the propeller normally works, and the trouble sample that corresponds to other trouble degrees does not make full use of, leads to the trouble to discern the precision limited.
Disclosure of Invention
The purpose of the invention is as follows: the invention aims to provide a multinuclear model-based method for monitoring the state of a propeller of an underwater robot, which solves the problem of low identification precision of the existing method.
The technical scheme is as follows: the method for monitoring the state of the underwater robot propeller based on the multi-core model comprises the following steps:
(1) method for extracting propeller fault feature x from underwater robot monitoring signal1,x2,…,xNThe fault signature constitutes a fault sample, X ═ X1 x2 … xN]Obtaining fault samples of M fault levels, wherein the fault levels are I, II, III, … and M respectively, and the corresponding fault degrees are lambda respectivelyI、λII、λIII、…、λM(ii) a The fault samples are divided into training samples and test samples.
(2) Calculating first characteristics x of fault samples with different fault grades1Maximum, average and minimum values of;
(3) taking training samples corresponding to the fault level I as target samples, taking training samples corresponding to the rest fault levels as non-target samples, and establishing a fuzzy membership function;
(4) calculating the fuzzy membership coefficient s of all training samples according to the fuzzy membership function S (X)IiAccording to the target sample and its fuzzy membership coefficient sI1Training the hypersphere model I to obtain the center CIAnd a radius RICalculating the non-target sample to the center CIGeneralized distance D ofIiAccording to the generalized distance D of the non-target sampleIiCalculating the monitoring coefficient, and calculating the average value of the training sample monitoring coefficients corresponding to the fault levels I-MI1A~IMAEstablishing a mapping function of the monitoring coefficient and the fault degree to obtain a single-core identification model I established by taking a training sample corresponding to the fault grade I as a target sample; wherein, the generalized distance DIiThe calculation formula is as follows:
wherein D isIiFor the ith training sample XiGeneralized distance to the ith hypersphere, sIiFor the ith training sample XiFuzzy membership coefficient in the I-th fuzzy membership function, K () is kernel function, XIj、XIkTraining samples corresponding to the failure class I, sIjAnd sIkIs XIj、XIkFuzzy membership coefficient, alpha, in the I-th fuzzy membership functionIjAnd alphaIkIs the global optimal solution of the I-th hyper-sphere, N6The number of training samples.
(5) Sequentially taking training samples corresponding to the fault classes II, III, IV, … and M as target samples, establishing a fuzzy membership function, and obtaining single-core identification models II, III, IV, … and M which are respectively established by taking the training samples corresponding to the fault classes II, III, IV, … and M as the target samples according to the step (4), wherein the multiple single-core identification models I-M jointly form a multi-core identification model;
(6) establishing a test sample XU=[xU1 xU2 … xUN]The fault degree of the test sample is lambdaUFuzzy membership coefficient s of the test sampleiUAnd fuzzy membership coefficient s in hypersphere models I-MijAll are set to be zero, and the generalized distance D from the test sample to the hypersphere models I-M is calculatediUAccording to the obtained generalized distance DiUAnd determining the fault degree of the test sample as lambda by the ratio of the sum to the radius of the corresponding hypersphere modelU=λi;
(7) According to the first characteristic x of the test sampleU1Determining a fault degree interval;
(8) within the fault degree interval, testing a first characteristic x of a sampleU1Substituting corresponding fuzzy membership function to calculate fuzzy membership coefficient s of test sampleiUCalculating the monitoring coefficient from the test sample to the hypersphere model I-MIU,IIU,…,MUThe monitoring coefficient is substituted into corresponding single-core identification models I-M to obtain a plurality of fault degree identification results lambdaIU,λIIU,…,λMUTo find lambdaIU~λMUThe average value of the values is obtained to obtain the final identification result.
Wherein, the first characteristic x in the step (1)1The characteristic of the fault is most sensitive to the fault degree, and the fault level I corresponds to the normal work of the propeller.
The fault grade I in the step (2) and the first characteristic x of the fault sampleI1Maximum value of (2) is xI1HMinimum value of xI1LAverage value of xI1A(ii) a Fault class II, fault sampleFirst characteristic xII1Maximum value of (2) is xII1HMinimum value of xII1LAverage value of xII1A(ii) a … …, respectively; fault class M, fault sample first feature xM1Maximum value of (2) is xM1HMinimum value of xM1LAverage value of xM1A。
Establishing a fuzzy membership function in the step (3) as follows:
wherein, L ═ xI1H,H=xM1L。
The step (4) is specifically as follows:
based on target sample and its fuzzy membership coefficient sX1Training the hypersphere model I to obtain the center CIAnd a radius RICalculating the non-target sample to the center CIGeneralized distance D ofIiCalculating the monitoring coefficientIi=(DIi-RI)/RICalculating the average value of the training sample monitoring coefficients corresponding to the fault level II1ARespectively calculating the average value of the training sample monitoring coefficients corresponding to the fault levels II-MI2A~IMAThereby obtaining (I1A,λI),(I2A,λII),…,(IMA,λM);
Will (a) toI1A,λI),(I2A,λII),…,(IMA,λM) Determining a straight line according to the two points, establishing a mapping function from the monitoring coefficient to the fault degree between the two fault grades, and obtaining a mononuclear identification model I established by taking a training sample corresponding to the fault grade I as a target sample, wherein the mapping function is as follows:
when lambda isI≤λ<λIIWhen k is equal to λI1×+bI1;
When lambda isII≤λ<λIIIWhen k is equal to λI2×+bI2;
……;
When lambda isM-1≤λ<λMWhen k is equal to λI(M-1)×+bI(M-1);
Wherein k isI1=(λII-λI)/(I2A-I1A),bI1=λI-I1A(λII-λI)/(I2A-I1A);
kI2=(λIII-λII)/(I3A-I2A),bI2=λII-I2A(λIII-λII)/(I3A-I2A);
……;
kI(M-1)=(λM-λM-1)/(IMA-I(M-1)A),
bI(M-1)=λM-1-I(M-1)A(λM-λM-1)/(IMA-I(M-1)A)。
The step (5) is specifically as follows:
training samples corresponding to the fault level II are used as target samples, training samples corresponding to the rest fault levels are used as non-target samples, a fuzzy membership function is established as follows,
at λI≤λ<λIIIn the interval, L ═ xI1H,H=xII1L,
At λII≤λ<λMIn the interval, L ═ xII1H,H=xM1L,
Obtaining a single-core identification model II established by taking the training sample corresponding to the fault level II as a target sample according to the step (4), and further obtaining single-core identification models III, IV, … and M-1 established by taking the training samples corresponding to the fault levels III, IV, … and M-1 as target samples;
taking training samples corresponding to the fault level M as target samples, taking training samples corresponding to the rest fault levels as non-target samples, and establishing a fuzzy membership function as follows:
wherein, L ═ xI1H,H=xM1LAnd (5) obtaining a single-core identification model M established by taking the training sample corresponding to the fault level M as a target sample according to the step (4).
The generalized distance D obtained in the step (6) is used as the basisiUAnd determining the fault degree of the test sample as lambda by the ratio of the sum to the radius of the corresponding hypersphere modelUThe formula of (1) is as follows:
if D isIU/RILambda is less than or equal to 1U=λI;
If D isIIU/RIILambda is less than or equal to 1U=λII;
……,
If D isMU/RMLambda is less than or equal to 1U=λM;
And (7) if the test sample does not meet the condition, executing the step.
The step (7) is carried out according to the first characteristic x of the test sampleU1Determining a fault degree interval specifically as follows:
if xU1<xI1AThen λU=λI;
If xI1A≤xU1<xII1AThen λI≤λU<λII;
If xII1A≤xU1<xIII1AThen λII≤λU<λIII;
……;
If x(M-1)1A≤xU1<xM1AThen λM-1≤λU<λM;
If xM1A≤xU1Then λU=λM。
Has the advantages that: on the basis of a fuzzy support vector field, training samples corresponding to different fault grades are respectively used as target samples to respectively train a plurality of hypersphere models, then training samples corresponding to other fault grades except the target grade are used as non-target samples, monitoring coefficients are calculated, a plurality of single-core identification models are established, and finally the plurality of single-core identification models are integrated into a multi-core identification model.
Drawings
FIG. 1 is a flow chart of the off-line modeling phase of the present invention;
FIG. 2 is a flow chart of the online fault identification phase of the present invention;
FIG. 3 is a flow chart of an offline modeling phase of a conventional fuzzy support vector domain description method;
FIG. 4 is a flow chart of an online fault identification phase of a conventional fuzzy support vector domain description method;
FIG. 5 is a schematic diagram of a training sample distribution;
FIG. 6 is a schematic view of a test sample distribution;
FIG. 7 is a schematic diagram of each of the single-core identification models;
FIG. 8 is a diagram illustrating a fault identification result of a conventional SVM domain description method;
FIG. 9 is a diagram illustrating a fault identification result of the multi-core model method according to the present invention.
Detailed Description
The invention is further described below with reference to the accompanying drawings.
The invention discloses a method for monitoring the state of a propeller of an underwater robot based on a multi-core model, which comprises the following steps:
(1) offline establishment of multi-core identification model
(a) From underwater robot supervisionExtracting propeller fault feature x from measured signal1,x2,…,xNWherein the first characteristic x1Is the fault signature most sensitive to the degree of the fault, the fault signature constitutes a fault sample, X ═ X1 x2 …xN]The fault samples are divided into training samples and testing samples, a group of example results are shown in fig. 5 and 6, and it can be known from the graphs that the speed signal characteristics are sensitive to the fault degree, so that the speed signal fault characteristics form first fault characteristics, through experiments, training samples of 5 fault levels are obtained, the fault levels are I, II, III, iv and V, and the corresponding fault degrees are 0%, 10%, 20%, 30% and 40% respectively;
(b) calculating first characteristics x of training samples with different fault grades1Maximum, average, minimum:
failure class I, training sample first feature xI1Has a maximum value of 3.37, a minimum value of 2.85, and an average value of 3.14;
failure class II, training sample first feature xII1A maximum value of 9.34, a minimum value of 8.53, and an average value of 8.86;
failure class III, training sample first feature xIII1Has a maximum value of 11.52, a minimum value of 10.11 and an average value of 11.05;
failure class IV, training sample first feature xⅣ1Has a maximum value of 13.80, a minimum value of 13.59, and an average value of 13.70;
failure class V, training sample first feature xV1Has a maximum value of 15.12, a minimum value of 14.98 and an average value of 15.03;
(c) training samples corresponding to the fault level I are used as target samples, training samples corresponding to the rest fault levels are used as non-target samples, and a fuzzy membership function is established as follows, wherein L is xI1H=3.37,H=xV1L=14.98。
(d) Based onCalculating fuzzy membership coefficient s of all training samples by using fuzzy membership functionIiBased on the target sample and its fuzzy membership coefficient sI1Training the hypersphere model I to obtain the center CIAnd a radius RI. Calculating non-target sample to center of sphere CIGeneralized distance D ofIiCalculating the monitoring coefficientIi=(DIi-RI)/RI. Calculating the average value of the training sample monitoring coefficients corresponding to the fault grade II1ASimilarly, the average value of the training sample monitoring coefficients corresponding to the fault levels II-V is calculated respectivelyI2A~I5A,
I1A=-0.24,I2A=10.00,I3A=12.16,I4A=13.46,I5A=13.61,
Thereby obtaining (I1A,λI),(I2A,λII),…,(I5A,λV);
(e) Will (a) toI1A,λI),(I2A,λII),…,(I5A,λV) Regarding the points in the plane, determining a straight line according to the two points, and establishing a mapping function from the monitoring coefficient to the fault degree between the two fault levels, which is specifically as follows:
when lambda isI≤λ<λIIWhen k is equal to λI1×+bI1;
When lambda isII≤λ<λIIIWhen k is equal to λI2×+bI2;
When lambda isIII≤λ<λIVWhen k is equal to λI3×+bI3;
When lambda isIV≤λ<λVWhen k is equal to λI4×+bI4;
In the above formula, kI1,kI2,…,kI4,bI1,bI2,…,bI4Respectively by reactingI1A,λI),(I2A,λII),…,(I5A,λV) The above equation was substituted.
kI1=(λII-λI)/(I2A-I1A)=0.010,bI1=λI-I1A(λII-λI)/(I2A-I1A)=0.002;
kI2=(λIII-λII)/(I3A-I2A)=0.046,bI2=λII-I2A(λIII-λII)/(I3A-I2A)=-0.360;
kI3=(λIV-λIII)/(I4A-I3A)=0.077,bI3=λIII-I3A(λIV-λIII)/(I4A-I3A)=-0.736;
kI4=(λV-λIV)/(I5A-I4A)=0.667,bI4=λIV-I4A(λV-λIV)/(I5A-I4A)=-8.678;
(f) Through the steps 1(c) to (e), a single-core identification model I established by taking the training sample corresponding to the fault level I as a target sample can be obtained, wherein the model I is shown in fig. 7 (a);
(g) training samples corresponding to the fault level II are used as target samples, training samples corresponding to the rest fault levels are used as non-target samples, a fuzzy membership function is established as follows,
at λI≤λ<λIIIn this interval, L ═ xI1H=3.37,H=xII1L=8.53。
At λII≤λ<λVIn this interval, L ═ xII1H=9.34,H=xV1L=14.98。
(h) Referring to steps 1(d) - (f), obtaining a single-core identification model II established by taking a training sample corresponding to the fault level II as a target sample, wherein the model II is shown in FIG. 7 (b);
(i) referring to steps 1(g) - (h), single-core identification models III, IV, V established by using training samples corresponding to fault levels III, IV, V as target samples can be obtained, where model III is shown in fig. 7(c), model IV is shown in fig. 7(d), and model V is shown in fig. 7(e), respectively;
(j) a plurality of single-core identification models I-V jointly form a multi-core identification model;
(2) on-line fault degree identification
(a) According to the fuzzy support vector field description theory, when the fuzzy membership coefficient is set to be zero, the fuzzy support vector field description is degenerated to a support vector field description algorithm. Therefore, this patent first obtains a test sample X according to step 1(a)U=[xU1 xU2 … xUN]Degree of failure λ of test sampleUUnknown, the fuzzy membership coefficient s of the test sampleiUAnd fuzzy membership coefficient s in hypersphere models I-VijAll are set to be zero, and the generalized distance D from the test sample to the hypersphere models I-V is calculatediU,
If D isIU/RILambda is less than or equal to 1U=λI=0%;
If D isIIU/RIILambda is less than or equal to 1U=λII=10%;
……,
If D isVU/RVLambda is less than or equal to 1U=λV=40%;
If the test sample does not meet the condition, executing the following steps;
(b) and determining a fault degree interval. According to the first characteristic x of the test sampleU1Determining a fault degree interval specifically as follows:
xI1A=3.14,xII1A=8.86,xIII1A=11.05,xIV1A=13.70,xV1A=15.03。
if xU1<xI1AThen λU=0%;
If xI1A≤xU1<xII1AWhen the ratio is more than or equal to 0% < lambda >U<10%;
If xII1A≤xU1<xIII1AWhen the ratio is more than or equal to 10% < lambda >U<20%;
If xIII1A≤xU1<xIV1AWhen the ratio is more than or equal to 20% < lambda >U<30%;
If xIV1A≤xU1<xV1AWhen the ratio is more than or equal to 30 percent, lambda is more than or equal toU<40%;
If xV1A≤xU1Then λU=40%;
(c) Within the fault degree interval, testing a first characteristic x of a sampleU1Substituting corresponding fuzzy membership function to calculate fuzzy membership coefficient s of test sampleiU. Calculating the monitoring coefficient from the test sample to the hypersphere model I-VIU,IIU,…,VUThe monitoring coefficient is substituted into corresponding single-core identification models I-V to obtain a plurality of fault degree identification results lambdaIU,λIIU,…,λVUTo find lambdaIU~λVUThe average value of the data is used as a final identification result;
by comparing fig. 1 and fig. 3, it can be found that, in the existing fuzzy support vector field description algorithm, the training samples when the propeller normally works are only used as target samples for training the hypersphere model, and the training samples corresponding to other fault classes are only used as non-target samples for calculating the monitoring coefficient, so that the fault sample information is not fully utilized. On the basis of the existing fuzzy support vector field, training samples corresponding to different fault levels are respectively used as target samples to respectively train a plurality of hypersphere models, then training samples corresponding to other fault levels except the target levels (including training samples when a propeller normally works) are used as non-target samples, monitoring coefficients are calculated, a plurality of single-core identification models are established, and finally, the plurality of single-core identification models are integrated into a multi-core identification model.
By comparing fig. 2 and fig. 4, it can be found that the existing fuzzy support vector field description algorithm only calculates the monitoring coefficient of the corresponding single-core hypersphere model from the test sample to the propeller when the propeller normally works, and then brings the monitoring coefficient into the single-core identification model to obtain the fault degree.
As shown in fig. 8 and fig. 9, when the test sample is at some fault levels, such as fault levels 10%, 20%, and 30%, the fault identification result of the conventional fuzzy support vector field description method fluctuates around the true fault level, and the identification result of the present invention coincides with the true fault level; when the test sample is between two fault levels, such as 11%, 23%, 26% and 37%, compared with the fault identification result of the existing fuzzy support vector field description method, the patent identification result of the invention is closer to the real fault degree;
for describing the fault identification performance quantitatively, the identification results shown in fig. 8-9 are counted in table 1, and the data format in table 1 is, the average value ± standard deviation, which is specifically as follows:
TABLE 1 identification results of failure degree of underwater propeller by different methods
As can be seen from table 1, at the fault levels, for example, 10%, 20%, 30%, and 40%, compared with the fault identification method of the conventional fuzzy support vector field description method, the average identification accuracy of the conventional fuzzy support vector field description method is 96.39%, the average identification accuracy of the present method is 100%, and the identification accuracy is improved by 3.61% on average, and when the test sample is located between two fault level levels, for example, 11%, 23%, 26%, and 37%, the average identification accuracy of the conventional fuzzy support vector field description method is 86.71%, the average identification accuracy of the present invention is 95.53%, and the identification accuracy is improved by 8.82% on average. Overall, compared with the fault identification method of the existing fuzzy support vector field description method, the identification precision of the method is averagely improved by 6.19%, and the standard deviation of the identification precision is reduced by 6.25%. The experimental result proves the effectiveness of the method in the aspects of improving the identification precision and reducing the standard deviation of the identification precision.
Claims (8)
1. A method for monitoring the state of a propeller of an underwater robot based on a multi-core model is characterized by comprising the following steps:
(1) method for extracting propeller fault feature x from underwater robot monitoring signal1,x2,…,xNThe fault signature constitutes a fault sample, X ═ X1 x2 … xN]Obtaining fault samples of M fault levels, wherein the fault levels are I, II, III, … and M respectively, and the corresponding fault degrees are lambda respectivelyI、λII、λIII、…、λMDividing the test sample into a training sample and a test sample;
(2) calculating first characteristics x of fault samples with different fault grades1Maximum, average and minimum values of;
(3) taking training samples corresponding to the fault level I as target samples, taking training samples corresponding to the rest fault levels as non-target samples, and establishing a fuzzy membership function;
(4) calculating the fuzzy membership coefficient s of all training samples according to the fuzzy membership function S (X)IiAccording to the target sample and its fuzzy membership coefficient sI1Training the hypersphere model I to obtain the center CIAnd a radius RICalculating the non-target sample to the center CIGeneralized distance D ofIiAccording to the generalized distance D of the non-target sampleIiCalculating the monitoring coefficient, and calculating the average value of the training sample monitoring coefficients corresponding to the fault levels I-MI1A~IMAEstablishing a mapping function of the monitoring coefficient and the fault degree to obtain a single-core identification model I established by taking a training sample corresponding to the fault grade I as a target sample;
(5) sequentially taking training samples corresponding to the fault classes II, III, IV, … and M as target samples, establishing a fuzzy membership function, and obtaining single-core identification models II, III, IV, … and M which are respectively established by taking the training samples corresponding to the fault classes II, III, IV, … and M as the target samples according to the step (4), wherein the multiple single-core identification models I-M jointly form a multi-core identification model;
(6) establishing a test sample XU=[xU1 xU2 … xUN]The fault degree of the test sample is lambdaUFuzzy membership coefficient s of the test sampleiUAnd fuzzy membership coefficient s in hypersphere models I-MijAll are set to be zero, and the generalized distance D from the test sample to the hypersphere models I-M is calculatediUAccording to the obtained generalized distance DiUAnd determining the fault degree of the test sample as lambda by the ratio of the sum to the radius of the corresponding hypersphere modelU=λi;
(7) According to the first characteristic x of the test sampleU1Determining a fault degree interval;
(8) within the fault degree interval, testing a first characteristic x of a sampleU1Substituting corresponding fuzzy membership function S (X) to calculate fuzzy membership coefficient s of test sampleiUCalculating the monitoring coefficient from the test sample to the hypersphere model I-MIU,IIU,…,MThe monitoring coefficient is substituted into corresponding single-core identification models I-M to obtain a plurality of fault degree identification results lambdaIU,λIIU,…,λMUTo find lambdaIU~λMUThe average value of the values is obtained to obtain the final identification result.
2. The multi-core model-based underwater robot propeller state monitoring method according to claim 1, wherein the first feature x in the step (1)1The characteristic of the fault is most sensitive to the fault degree, and the fault level I corresponds to the normal work of the propeller.
3. The method for monitoring the state of a multi-core model-based underwater robot propeller of claim 1, wherein the fault in the step (2)Class I, failure sample first feature xI1Maximum value of (2) is xI1HMinimum value of xI1LAverage value of xI1A(ii) a Fault class II, fault sample first feature xII1Maximum value of (2) is xII1HMinimum value of xII1LAverage value of xII1A(ii) a … …, respectively; fault class M, fault sample first feature xM1Maximum value of (2) is xM1HMinimum value of xM1LAverage value of xM1A。
5. The method for monitoring the state of the underwater robot propeller based on the multi-core model according to claim 1, wherein the step (4) is specifically as follows:
based on target sample and its fuzzy membership coefficient sI1Training the hypersphere model I to obtain the center CIAnd a radius RICalculating the non-target sample to the center CIGeneralized distance D ofIiCalculating the monitoring coefficientIi=(DIi-RI)/RICalculating the average value of the training sample monitoring coefficients corresponding to the fault level II1ARespectively calculating the average value of the training sample monitoring coefficients corresponding to the fault levels II-MI2A~IMAThereby obtaining (I1A,λI),(I2A,λII),…,(IMA,λM);
Will (a) toI1A,λI),(I2A,λII),…,(IMA,λM) Regarding as points in a plane, a straight line is determined from the two pointsEstablishing a mapping function from the monitoring coefficient to the fault degree between the two fault grades to obtain a single-core identification model I established by taking a training sample corresponding to the fault grade I as a target sample, wherein the mapping function is as follows:
when lambda isI≤λ<λIIWhen k is equal to λI1×+bI1;
When lambda isII≤λ<λIIIWhen k is equal to λI2×+bI2;
……;
When lambda isM-1≤λ<λMWhen k is equal to λI(M-1)×+bI(M-1);
Wherein k isI1=(λII-λI)/(I2A-I1A),bI1=λI-I1A(λII-λI)/(I2A-I1A),
kI2=(λIII-λII)/(I3A-I2A),bI2=λII-I2A(λIII-λII)/(I3A-I2A),
……,
kI(M-1)=(λM-λM-1)/(IMA-I(M-1)A),
bI(M-1)=λM-1-I(M-1)A(λM-λM-1)/(IMA-I(M-1)A)。
6. The method for monitoring the state of the underwater robot propeller based on the multi-core model according to claim 1, wherein the step (5) is specifically as follows:
training samples corresponding to the fault level II are used as target samples, training samples corresponding to the rest fault levels are used as non-target samples, a fuzzy membership function is established as follows,
at λI≤λ<λIIIn the interval, L ═ xI1H,H=xII1L,
At λII≤λ<λMIn the interval, L ═ xII1H,H=xM1L,
Obtaining a single-core identification model II established by taking the training sample corresponding to the fault level II as a target sample according to the step (4), and further obtaining single-core identification models III, IV, … and M-1 established by taking the training samples corresponding to the fault levels III, IV, … and M-1 as target samples;
taking training samples corresponding to the fault level M as target samples, taking training samples corresponding to the rest fault levels as non-target samples, and establishing a fuzzy membership function as follows:
wherein, L ═ xI1H,H=xM1LAnd (5) obtaining a single-core identification model M established by taking the training sample corresponding to the fault level M as a target sample according to the step (4).
7. The method for monitoring the state of a multi-core model-based underwater robot propeller of claim 1, wherein the step (6) is performed according to the obtained generalized distance DiUAnd determining the fault degree of the test sample as lambda by the ratio of the sum to the radius of the corresponding hypersphere modelUThe formula of (1) is as follows:
if D isIU/RILambda is less than or equal to 1U=λI;
If D isIIU/RIILambda is less than or equal to 1U=λII;
……,
If D isMU/RMLambda is less than or equal to 1U=λM;
And (7) if the test sample does not meet the condition, executing the step.
8. The method for monitoring the state of a multinuclear model-based underwater robot propeller of claim 1, wherein in the step (7), the first characteristic x is determined according to a test sampleU1Determining a fault degree interval specifically as follows:
if xU1<xI1AThen λU=λI;
If xI1A≤xU1<xII1AThen λI≤λU<λII;
If xII1A≤xU1<xIII1AThen λII≤λU<λIII;
……;
If x(M-1)1A≤xU1<xM1AThen λM-1≤λU<λM;
If xM1A≤xU1Then λU=λM。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010951862.8A CN112051836B (en) | 2020-09-11 | 2020-09-11 | Underwater robot propeller state monitoring method based on multi-core model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010951862.8A CN112051836B (en) | 2020-09-11 | 2020-09-11 | Underwater robot propeller state monitoring method based on multi-core model |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112051836A true CN112051836A (en) | 2020-12-08 |
CN112051836B CN112051836B (en) | 2021-09-24 |
Family
ID=73611061
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010951862.8A Active CN112051836B (en) | 2020-09-11 | 2020-09-11 | Underwater robot propeller state monitoring method based on multi-core model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112051836B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117291921A (en) * | 2023-11-27 | 2023-12-26 | 哪吒港航智慧科技(上海)有限公司 | Container sporadic damage sample mining and learning method, device, equipment and medium |
Citations (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5566092A (en) * | 1993-12-30 | 1996-10-15 | Caterpillar Inc. | Machine fault diagnostics system and method |
US20100085669A1 (en) * | 2008-10-03 | 2010-04-08 | General Electric Company | Arc fault detection using fuzzy logic |
CN101907681A (en) * | 2010-07-15 | 2010-12-08 | 南京航空航天大学 | Analog circuit dynamic online failure diagnosing method based on GSD-SVDD |
CN102735447A (en) * | 2012-06-29 | 2012-10-17 | 西安交通大学 | Method for quantitatively identifying performance degradation degree of rolling bearing |
US20140324747A1 (en) * | 2013-04-30 | 2014-10-30 | Raytheon Company | Artificial continuously recombinant neural fiber network |
CN104252575A (en) * | 2014-08-06 | 2014-12-31 | 哈尔滨工程大学 | Behavior based UUV (unmanned underwater vehicle) propulsion operating system exception identification method |
CN105759201A (en) * | 2016-03-11 | 2016-07-13 | 江苏镇安电力设备有限公司 | High voltage circuit breaker self-diagnosis method based on abnormal sample identification |
CN106203500A (en) * | 2016-06-30 | 2016-12-07 | 浙江群力电气有限公司 | A kind of Fault Classification based on the fuzzy support vector machine improved |
CN107132760A (en) * | 2017-03-27 | 2017-09-05 | 哈尔滨工程大学 | The underwater robot state monitoring method described based on fuzzy support vector domain |
CN107450572A (en) * | 2017-07-26 | 2017-12-08 | 江苏科技大学 | Underwater robot attitude regulation control system and processing method based on sliding formwork control |
CN108830218A (en) * | 2018-06-15 | 2018-11-16 | 哈尔滨工程大学 | A kind of underwater robot propeller method for diagnosing faults based on improvement Isomap algorithm ISOMAP |
CN109033965A (en) * | 2018-06-22 | 2018-12-18 | 江苏科技大学 | A kind of underwater robot propeller failure time-frequency characteristics Enhancement Method |
CN109633270A (en) * | 2019-01-02 | 2019-04-16 | 江苏科技大学 | The identification of fault energy zone boundary and feature extracting method based on instantaneous spectrum entropy and noise energy difference |
CN109683591A (en) * | 2018-12-27 | 2019-04-26 | 江苏科技大学 | Underwater propeller fault degree discrimination method based on fusion signal time domain energy and time-frequency entropy |
CN109696906A (en) * | 2018-12-27 | 2019-04-30 | 江苏科技大学 | Underwater robot propeller method for diagnosing faults based on small echo amendment Bayes's convolution energy |
KR102055915B1 (en) * | 2018-11-14 | 2019-12-13 | 주식회사 케이티 | System and method for fault prediction in core network based on autoencoder |
CN111542794A (en) * | 2018-01-03 | 2020-08-14 | 高通股份有限公司 | Adjustable object avoidance proximity threshold for robotic vehicles based on detected presence of payload |
CN111626374A (en) * | 2020-06-02 | 2020-09-04 | 上海电力大学 | Switch cabinet fault classification method based on semi-supervised learning |
-
2020
- 2020-09-11 CN CN202010951862.8A patent/CN112051836B/en active Active
Patent Citations (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5566092A (en) * | 1993-12-30 | 1996-10-15 | Caterpillar Inc. | Machine fault diagnostics system and method |
US20100085669A1 (en) * | 2008-10-03 | 2010-04-08 | General Electric Company | Arc fault detection using fuzzy logic |
CN101907681A (en) * | 2010-07-15 | 2010-12-08 | 南京航空航天大学 | Analog circuit dynamic online failure diagnosing method based on GSD-SVDD |
CN102735447A (en) * | 2012-06-29 | 2012-10-17 | 西安交通大学 | Method for quantitatively identifying performance degradation degree of rolling bearing |
US20140324747A1 (en) * | 2013-04-30 | 2014-10-30 | Raytheon Company | Artificial continuously recombinant neural fiber network |
CN104252575A (en) * | 2014-08-06 | 2014-12-31 | 哈尔滨工程大学 | Behavior based UUV (unmanned underwater vehicle) propulsion operating system exception identification method |
CN105759201A (en) * | 2016-03-11 | 2016-07-13 | 江苏镇安电力设备有限公司 | High voltage circuit breaker self-diagnosis method based on abnormal sample identification |
CN106203500A (en) * | 2016-06-30 | 2016-12-07 | 浙江群力电气有限公司 | A kind of Fault Classification based on the fuzzy support vector machine improved |
CN107132760A (en) * | 2017-03-27 | 2017-09-05 | 哈尔滨工程大学 | The underwater robot state monitoring method described based on fuzzy support vector domain |
CN107450572A (en) * | 2017-07-26 | 2017-12-08 | 江苏科技大学 | Underwater robot attitude regulation control system and processing method based on sliding formwork control |
CN111542794A (en) * | 2018-01-03 | 2020-08-14 | 高通股份有限公司 | Adjustable object avoidance proximity threshold for robotic vehicles based on detected presence of payload |
CN108830218A (en) * | 2018-06-15 | 2018-11-16 | 哈尔滨工程大学 | A kind of underwater robot propeller method for diagnosing faults based on improvement Isomap algorithm ISOMAP |
CN109033965A (en) * | 2018-06-22 | 2018-12-18 | 江苏科技大学 | A kind of underwater robot propeller failure time-frequency characteristics Enhancement Method |
KR102055915B1 (en) * | 2018-11-14 | 2019-12-13 | 주식회사 케이티 | System and method for fault prediction in core network based on autoencoder |
CN109683591A (en) * | 2018-12-27 | 2019-04-26 | 江苏科技大学 | Underwater propeller fault degree discrimination method based on fusion signal time domain energy and time-frequency entropy |
CN109696906A (en) * | 2018-12-27 | 2019-04-30 | 江苏科技大学 | Underwater robot propeller method for diagnosing faults based on small echo amendment Bayes's convolution energy |
CN109633270A (en) * | 2019-01-02 | 2019-04-16 | 江苏科技大学 | The identification of fault energy zone boundary and feature extracting method based on instantaneous spectrum entropy and noise energy difference |
CN111626374A (en) * | 2020-06-02 | 2020-09-04 | 上海电力大学 | Switch cabinet fault classification method based on semi-supervised learning |
Non-Patent Citations (1)
Title |
---|
殷宝吉,等: "船舶螺旋桨水下清洗机器人推进器驱动及关键部件状态监测研究", 《江苏科技大学学报( 自然科学版)》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117291921A (en) * | 2023-11-27 | 2023-12-26 | 哪吒港航智慧科技(上海)有限公司 | Container sporadic damage sample mining and learning method, device, equipment and medium |
CN117291921B (en) * | 2023-11-27 | 2024-03-12 | 哪吒港航智慧科技(上海)有限公司 | Container sporadic damage sample mining and learning method, device, equipment and medium |
Also Published As
Publication number | Publication date |
---|---|
CN112051836B (en) | 2021-09-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2021135630A1 (en) | Rolling bearing fault diagnosis method based on grcmse and manifold learning | |
CN105930976B (en) | Node voltage sag severity comprehensive evaluation method based on weighted ideal point method | |
CN109765490B (en) | Power battery fault detection method and system based on high-dimensional data diagnosis | |
WO2022021726A1 (en) | Pmu-based power system state estimation performance evaluation method | |
CN108508865B (en) | A kind of fault detection method based on distributing OSC-PLS regression model | |
CN111680875B (en) | Unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on probability baseline model | |
Zhang et al. | A novel data-driven method based on sample reliability assessment and improved CNN for machinery fault diagnosis with non-ideal data | |
CN115187832A (en) | Energy system fault diagnosis method based on deep learning and gram angular field image | |
CN108830006B (en) | Linear-nonlinear industrial process fault detection method based on linear evaluation factor | |
WO2021114320A1 (en) | Wastewater treatment process fault monitoring method using oica-rnn fusion model | |
CN112051836B (en) | Underwater robot propeller state monitoring method based on multi-core model | |
CN114266289A (en) | Complex equipment health state assessment method | |
CN109886314B (en) | Kitchen waste oil detection method and device based on PNN neural network | |
CN109902731B (en) | Performance fault detection method and device based on support vector machine | |
CN111612149A (en) | Main network line state detection method, system and medium based on decision tree | |
CN111400966A (en) | Static voltage stability evaluation method of power system based on improved AdaBoost | |
CN113359665B (en) | Industrial process fault detection method and system based on weighted key pivot | |
CN114330486A (en) | Power system bad data identification method based on improved Wasserstein GAN | |
CN105759217B (en) | Online fault diagnosis method for lead-acid storage battery pack based on measurable data | |
El‐Shaarawi et al. | Parametric and nonparametric tests for dependent data 1 | |
CN111914471A (en) | Rectification tower fault detection method based on rapid nuclear independent component analysis | |
CN105824785A (en) | Rapid abnormal point detection method based on penalized regression | |
CN113033683B (en) | Industrial system working condition monitoring method and system based on static and dynamic joint analysis | |
Yu et al. | Design for an SPRT control scheme based on linguistic data | |
CN111814394A (en) | Power system safety assessment method based on correlation and redundancy detection |
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 | ||
TR01 | Transfer of patent right |
Effective date of registration: 20220804 Address after: 212021 room 1-413, high tech entrepreneurship service center, Zhenjiang high tech Zone, Jiangsu Province Patentee after: Zhenjiang Haiyun Marine Technology Co.,Ltd. Address before: 212003, No. 2, Mengxi Road, Zhenjiang, Jiangsu Patentee before: JIANGSU University OF SCIENCE AND TECHNOLOGY |
|
TR01 | Transfer of patent right |