CN112686525B - Comprehensive evaluation method of PHM system of electromechanical system - Google Patents
Comprehensive evaluation method of PHM system of electromechanical system Download PDFInfo
- Publication number
- CN112686525B CN112686525B CN202011573187.6A CN202011573187A CN112686525B CN 112686525 B CN112686525 B CN 112686525B CN 202011573187 A CN202011573187 A CN 202011573187A CN 112686525 B CN112686525 B CN 112686525B
- Authority
- CN
- China
- Prior art keywords
- layer
- index
- evaluation
- weight
- priority
- 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
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Complex Calculations (AREA)
Abstract
The invention discloses a comprehensive evaluation method of an electromechanical system PHM system, which comprises the following steps of S1: constructing a performance evaluation index system of the PHM system of the electromechanical system by using an AHP method, and sequencing the priority of each index; s2: cutting the saturation index set by utilizing a rough set theory to obtain a reduced evaluation index set; s3: performing weight calculation on the reduced evaluation index set by using an AHP method, and performing comprehensive weight fusion by combining with a D-S evidence theory, so as to realize comprehensive sequencing of the reduced evaluation index set; s4: based on the weight calculated by the AHP method and the D-S evidence theory, establishing an electromechanical system PHM system performance evaluation model based on fuzzy comprehensive evaluation to obtain a comprehensive conclusion of PHM system evaluation. The uncertainty conclusion of different experts is synthesized by about degeneracy of the index set, the influence degree of each measurement index on the total evaluation target can be reflected more accurately, the comprehensive evaluation conclusion is formed, and important information can be ensured not to be lost in the process of index cutting.
Description
Technical Field
The invention relates to the technical field of electromechanical systems, in particular to a comprehensive evaluation method of a PHM system of an electromechanical system.
Background
The electromechanical system is a strong nonlinear system composed of a plurality of units, and the health state equivalent characterization of the whole health state and reliability of the system by adopting the health state equivalent characterization of each individual part is unreasonable and the evaluation conclusion is unreliable because of the large correlation among the components of the system in the aspects of structure, function, utility and the like. Therefore, the PHM system of the built electromechanical system is also necessarily a complex system with multiple functions, and the performance evaluation of the PHM system is very complex because the coupling effect caused by various factors must be considered. PHM system performance evaluation is a multi-metric comprehensive evaluation problem, and the influence degree of each metric on the overall evaluation target is different. In order to correctly reflect the objective fact, a hierarchical analysis method (Analytic Hierarchy Process, AHP) is often adopted to give weight to measurement, however, when the number of constituent units or sub-systems of the system is large, the requirements on the performance of the system are further refined, so that the index for evaluating the PHM system is huge, and certain redundancy and misjudgment conditions can be generated in the actual use process. Therefore, we need to find a method to cut all index systems describing the evaluation of the PHM system, and in the cutting process, it is required to select those evaluation indexes with the greatest mutual influence to finally form a simplified index system, so that important information is not lost.
Disclosure of Invention
Aiming at the problems, the invention aims to provide a comprehensive evaluation method of an electromechanical system PHM system, which combines an AHP method and a D-S evidence theory, and synthesizes uncertainty conclusions of different experts through about degeneracy of an index set so as to form a comprehensive evaluation conclusion.
In order to achieve the above purpose, the technical scheme adopted by the invention is as follows:
a comprehensive evaluation method of an electromechanical system PHM system is characterized by comprising the following steps,
s1: constructing a performance evaluation index system of the PHM system of the electromechanical system by using an AHP method, and sequencing the priority of each index;
s2: cutting the saturation index set by utilizing a rough set theory to obtain a reduced evaluation index set;
s3: performing weight calculation on the reduced evaluation index set by using an AHP method, and performing comprehensive weight fusion by combining with a D-S evidence theory, so as to realize comprehensive sequencing of the reduced evaluation index set;
s4: based on the weight calculated by the AHP method and the D-S evidence theory, establishing an electromechanical system PHM system performance evaluation model based on fuzzy comprehensive evaluation to obtain a comprehensive conclusion of PHM system evaluation.
Further, the specific operation of step S1 includes the steps of,
s11: dividing the performance evaluation index of the PHM system into 4 layers by using an AHP method, and constructing a multi-layer performance evaluation index system;
s12, constructing a complementary judgment matrix of the index: after a multi-layer performance evaluation index system structure is established by using an AHP method, comparing each index with the priority of the previous layer in pairs, and carrying out weight assignment by adopting a common scale method of 1-9 so as to complete the construction of a pair comparison judgment matrix; by a' = [ a ij ] n×m A pair-wise comparison judgment matrix representing layer B versus layer A, wherein a ij Is the i-th index B in layer B i And the j-th index B j Score comparison value of a) ij The larger B i Ratio B j The priority of the relative layer a is high; a, a ij =1, representing having the same priority; and so on, a comparison judgment matrix of the layer C relative to the layer B is established and expressed as B= [ B ] ij ] n×m ；
S13, calculating the weight and sequencing single-layer priority: the priority of the criterion of layer B with respect to layer a satisfies aw=λ max W, where lambda max As the largest eigenvalue of matrix a', w= (W 1 ，w 2 ，…，w n ) T The vector corresponding to the weight represents the importance degree of classification sequencing in the layer B and is also a single-layer priority sequencing vector of the layer B; and so on, the single-layer priority of the layer C can be obtained;
s14, total priority ranking calculation: the total prioritization is a ranking of weights of all criteria of a single layer relative to the target layer, the criteria layer B contains m criteria b=b 1 ，B 2 ，…，B m The total priority weights of the layers A are respectively as follows
Layer C contains n metrics c=c 1 ，C 2 ，…，C n For criterion B j Is given by a single layer priority weight of
At this time C i The total prioritization weights with respect to layer a are: />
S15, consistency test: definition of the consistency index ci= (λ) max -n)/(n-1), wherein n is the number of measures of layer C, calculating a consistency ratio cr=ci/RI from the average random consistency index RI, and determining that the consistency condition is not satisfied when CR is not less than 0.10; if a certain index of layer C is for criterion B j The uniformity and average uniformity index of (C) are respectively Cl j 、Rl j The total ranking weight of layer B is
The overall rank uniformity ratio of layer C is +.>
Further, the 4 layers in step S11 include a target layer a, a criterion layer B, a metric layer C, and a result layer D;
the target layer A comprises PHM system performance evaluation A;
criterion layer B includes testability B 1 Fault diagnosis B 2 Failure prediction B 3 Occupy resource B 4 Technical maturity B 5 ；
The testability B 1 The corresponding metric layer C comprises a failure detection rate C 1 Fault isolation rate C 2 False alarm rate C 3 Failure prediction rate C 4 ；
The fault diagnosis B 2 The corresponding metric layer C comprises the accuracy C of fault diagnosis 5 Accuracy C of fault diagnosis 6 Stability of fault diagnosis C 7 Sensitivity of fault diagnosisDegree C 8 ；
The failure prediction B 3 The corresponding metric layer C comprises the accuracy C of fault prediction 9 Relative accuracy of failure prediction C 10 Failure prediction interval C 11 Alpha-lambda Performance C 12 Failure prediction coverage C 13 ；
The occupied resource B 4 The corresponding metric layer C includes full period cost C 14 Software code length C 15 Storage capacity C 16 Complexity of system C 17 ；
Degree of maturity of the technology B 5 The corresponding metric layer C includes a testability technology maturity C 18 Technical maturity C of fault diagnosis 19 Technical maturity of failure prediction C 20 ；
The result layer D includes pass, partial pass and no pass.
Further, in step S13, the method for calculating the eigenvector W of the matrix a' is as follows
Further, the specific operation of step S2 includes the steps of,
s21: set of n metrics giving a metrics layer
s22: in the reduction process, the definition of the accuracy and the definition of the accuracy corresponding to the failure prediction in the criterion layer are similar without considering the related indexes of the resource occupation and the technical maturity, and the failure diagnosis and the accuracy in the criterion layer can be combined into a group; thus, 20 metrics may be reduced to 9 by attribute, and each index is divided by "class I-III" to represent "excellent", "good" and "bad", respectively.
Further, the specific operation of step S3 includes the steps of,
s31: based on the judgment matrixes from layer B to layer A and from layer C to layer B given by each expert, calculating corresponding single-layer priority ranking weights according to an AHP method;
s32: fusing the priority ranking weights obtained in the step S31 by adopting a D-S evidence theory to obtain a ranking weight of the layer C on the layer A and a corresponding single-layer measurement priority ranking;
s33: and (3) acquiring judgment matrixes of the layers B to A and the layers C to B according to the result obtained in the step (S32), respectively obtaining a weight vector and a maximum characteristic value, and carrying out consistency test on the expert scoring result to finally obtain a new layer C sequencing weight of the layers A so as to realize comprehensive sequencing of the reduced evaluation index set.
Further, the specific operation steps of step S32 include,
s321: q PHM experts are arranged to participate in the performance evaluation, and based on the judgment matrix from layer B to layer A and from layer C to layer B given by each expert, the corresponding single-layer priority ranking weight W is calculated according to an AHP method k ＝[w 1k ，w 2k ，w 3k ，…，w pk ]K=1, 2,3, …, q; wherein p is the number of measurement indexes contained in the reduced index system;
s322: fusing the single-layer priority ranking weights obtained in the step S321 by adopting a Dempster combination rule in a D-S evidence theory to obtain a ranking weight of the layer C to the layer A and a corresponding single-layer degreeQuantity priority weight W k ＝[w 1 ，w 2 ，w 3 ，…，w p ]；
The sorting weight of the layer C to the layer A is obtained by the following method:
giving trust functions based on condensed metrics
Assuming that BEL1 and BEL2 are two trust functions under the same index system, m 1 、m 2 Respectively corresponding probability assignment, and focal element is A respectively 1 ，A 2 ，…，A k And B 1 ，B 2 ，…，B r Then
Further, the specific operation of step S4 includes the steps of,
s41, selecting a PHM system performance evaluation index fuzzy comment: according to the influence degree of qualitative analysis on the evaluation index, classifying the comment set into 5 grades I= { very serious, general, slight and very slight }, wherein the corresponding weight scores are {9,7,5,3,1};
s42, selecting a fuzzy matrix and a fuzzy membership function: determining a result layer D according to expert experience and PHM system function design requirements i And metric layer C j Is s ij Representation, construction of a fuzzy evaluation matrix s= (S) ij ) Determining a membership function;
s43, fuzzy comprehensive evaluation of performance of the PHM system of the electromechanical system: ranking weight vectors according to overall priority of metric layer relative to target layer
And the membership function matrix S between the evaluation result and each measurement is calculated to obtain the evaluation result of the performance of the PHM system of the electromechanical system as +.>
Further, in step S42, for qualitative and quantitative measurement, the performance measurement of the PHM system is normalized by using a semi-trapezoidal membership function, and then the fuzzy evaluation matrix is converted into a membership matrix s= (S) ij ) Wherein s is ij Is the relative membership of the ith result to the jth metric;
for positive and negative metrics, semi-trapezoidal membership functions are chosen for processing, i.e., x=s ij ，a＝max(e j )，b＝min(e j ) The method comprises the steps of carrying out a first treatment on the surface of the The membership functions of the semi-trapezium, which are larger and smaller, are defined as:
wherein the values in brackets are smaller.
The beneficial effects of the invention are as follows:
1. according to the invention, an AHP method, a rough set theory and a D-S evidence theory are combined, firstly, a performance evaluation index system of an electromechanical system PHM system is constructed by using the AHP method, and the priorities of the indexes are ordered; then, cutting the saturation index set by utilizing a rough set theory to obtain a reduced evaluation index set; secondly, performing weight calculation on the reduced evaluation index set by using an AHP method, and performing comprehensive weight fusion by combining with a D-S evidence theory, so as to realize comprehensive sequencing of the reduced evaluation index set; and finally, based on weights calculated by an AHP method and a D-S evidence theory, establishing an electromechanical system PHM system performance evaluation model based on fuzzy comprehensive evaluation to obtain a comprehensive conclusion of PHM system evaluation. The uncertainty conclusion of different experts is synthesized by about degeneracy of the index set, so that the influence degree of each measurement index on the total evaluation target can be more accurately reflected, and a comprehensive evaluation conclusion is formed; and in the process of index clipping, important information can be ensured not to be lost.
2. According to the electromechanical system PHM system comprehensive evaluation method, the actual fact that the PHM system performance evaluation itself needs multi-metric comprehensive evaluation is fully considered, the rough set is applied to simplify the calculation complexity in the actual engineering technology operation and remove redundant information, the problem of different priority results in the multi-expert decision process is solved by using the AHP and D-S evidence theory, and the problem of different multi-metric index dimensions is solved by using fuzzy comprehensive evaluation. The combination of the technical scheme not only avoids the excessively complex calculation process of the actual engineering application, but also ensures that different indexes of different experts generate the influence of the PHM system performance evaluation final result, thereby ensuring the objectivity and the effectiveness of the result. In addition, the technical scheme also has a certain engineering application value for other similar multi-metric index evaluation methods.
Drawings
FIG. 1 shows an electromechanical system PHM system performance evaluation index system based on an AHP method in the invention.
Detailed Description
In order to enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention is further described below with reference to the accompanying drawings and examples.
A comprehensive evaluation method of an electromechanical system PHM system comprises the following steps,
s1: constructing a performance evaluation index system of the PHM system of the electromechanical system by using an AHP method, and sequencing the priority of each index;
because the electromechanical system is a complex system, indexes influencing the performance of the electromechanical system are various, the coupling degree between various indexes is high, and the electromechanical system cannot be quantitatively expressed by a single function. It is very difficult and impractical to directly analyze the performance of an electromechanical system PHM system without any additional processing using existing evaluation metrics. The AHP method can simplify the problems, and is specifically as follows:
s11: the performance evaluation index of the PHM system is divided into 4 layers by using an AHP method, a multi-layer performance evaluation index system is constructed, various measurement index strips which influence the performance of the PHM system are subjected to physical chemistry and clustering by adopting the multi-layer hierarchical structure, and indexes with similar or compact influence degree on a certain index of an electromechanical system are put on the same layer, so that the multi-layer performance evaluation system is constructed, and the multi-layer performance evaluation system is shown in a figure 1.
Specifically, the 4 layers comprise a target layer A, a criterion layer B, a measurement layer C and a result layer D; the target layer A comprises PHM system performance evaluation A;
criterion layer B includes testability B 1 Fault diagnosis B 2 Failure prediction B 3 Occupy resource B 4 Technical maturity B 5 ；
The testability B 1 The corresponding metric layer C comprises a failure detection rate C 1 Fault isolation rate C 2 False alarm rate C 3 Failure prediction rate C 4 ；
The fault diagnosis B 2 The corresponding metric layer C comprises the accuracy C of fault diagnosis 5 Accuracy C of fault diagnosis 6 Stability of fault diagnosis C 7 Sensitivity C of fault diagnosis 8 ；
The failure prediction B 3 The corresponding metric layer C comprises the accuracy C of fault prediction 9 Relative accuracy of failure prediction C 10 Failure prediction interval C 11 Alpha-lambda Performance C 12 Failure prediction coverage C 13 ；
The occupied resource B 4 The corresponding metric layer C includes full period cost C 14 Software code length C 15 Storage capacity C 16 Complexity of system C 17 ；
Degree of maturity of the technology B 5 The corresponding metric layer C includes a testability technology maturity C 18 Technical maturity C of fault diagnosis 19 Technical maturity of failure prediction C 20 ；
The result layer D includes pass, partial pass and no pass.
S12, constructing a complementary judgment matrix of the index: after a multi-layer performance evaluation index system structure is established by using an AHP method, comparing each index with the priority of the previous layer in pairs, and carrying out weight assignment by adopting a common scale method of 1-9 so as to complete the construction of a pair comparison judgment matrix; by a' = [ a ij ] n×m A pair-wise comparison judgment matrix representing layer B versus layer A, wherein a ij Is the ith in layer BIndex B i And the j-th index B j Score comparison value of a) ij The larger B i Ratio B j The priority of the relative layer a is high; a, a ij =l, representing having the same priority; and so on, a comparison judgment matrix of the layer C relative to the layer B is established and is expressed as B' = [ B ] ij ] n×m ；
S13, calculating the weight and sequencing single-layer priority: the priority weights of the criteria of layer B relative to layer a satisfy a' w=λ max W, where lambda max As the largest eigenvalue of matrix a', w= (W 1 ，w 2 ，…，w n ) T The vector corresponding to the weight represents the importance degree of classification sequencing in the layer B and is also a single-layer priority sequencing vector of the layer B; and so on, the single-layer priority of the layer C can be obtained;
in the AHP method, the eigenvectors are approximated by summing and square root methods, the method is simple, but the calculation accuracy is not high, and in order to improve the accuracy, the eigenvectors W of the matrix A are solved by adopting a first row summing normalization method of the judgment matrix so as to obtain the measurement weights of each layer. The feature vector W of the matrix A' is calculated by the following method
S14, total priority ranking calculation: the total prioritization is a ranking of weights of all criteria of a single layer relative to the target layer, the criteria layer B contains m criteria b=b 1 ，B 2 ，…，B m The total priority weights of the layers A are respectively as follows
Layer C contains n metrics c=c 1 ，C 2 ，…，C n For criterion B j Is given by a single layer priority weight of
At this time C i The total prioritization weights with respect to layer a are: />
S15, consistency test: because of the complexity of the evaluation index itself and the instability of subjective judgment in qualitative measurement, it is difficult to divide the measurement difference of the same criterion, and at this time, a consistency index ci= (λ) is defined max -n)/(n-1), wherein n is the number of layer C metrics, and the corresponding average random uniformity index RI is obtained according to table 1 below, and further, the uniformity ratio cr=ci/RI is calculated, and when CR is equal to or greater than 0.10, it is determined that the uniformity condition is not satisfied.
TABLE 1 average random uniformity index RI
If a certain index of layer C is for criterion B j The uniformity and average uniformity index of (C) are respectively Cl j 、Rl j The total ranking weight of layer B is
The overall rank uniformity ratio of layer C is +.>
Further, S2: cutting the saturation index set by utilizing a rough set theory to obtain a reduced evaluation index set;
specifically, S21: set of n metrics giving a metrics layer
Wherein A is a target layer; b is a criterion layer corresponding to A, B= { B 1 ，B 2 ，…，B m }，B i For the ith criterion, B i E B; c is the metric layer set c= { C 1 ，c 2 ，…，c N }，C i Represents the ith metric, c i ＝(f i ，c，φ)，c i ∈C；/>
For measuring c i Weights for target layer a; phi is a measurement constraint, phi epsilon g x theta, g= (yes, no), yes represents that the measurement can be cut, and no represents that the measurement cannot be cut; θ is a clipping condition set; f (f) i C is i Is a measurement algorithm of (2);
s22: in the reduction process, the definition of the accuracy and the definition of the accuracy corresponding to the failure prediction in the criterion layer are similar without considering the related indexes of the resource occupation and the technical maturity, and the failure diagnosis and the accuracy in the criterion layer can be combined into a group; thus, the 20 metrics may be reduced to 9 by attribute, and each index divided by "class I-III" to represent "excellent", "good" and "bad", respectively, as shown in Table 2 below.
TABLE 2 PHM System Performance metrics 3 class separation Standard Table
Further, S3: performing weight calculation on the reduced evaluation index set by using an AHP method, and performing comprehensive weight fusion by combining with a D-S evidence theory, so as to realize comprehensive sequencing of the reduced evaluation index set;
specific: s31: based on the judgment matrixes from layer B to layer A and from layer C to layer B given by each expert, calculating corresponding single-layer priority ranking weights according to an AHP method;
s32: fusing the priority ranking weights obtained in the step S31 by adopting a D-S evidence theory to obtain a ranking weight of the layer C on the layer A and a corresponding single-layer measurement priority ranking;
specifically, S321: q PHM experts are arranged to participate in the performance evaluation, and based on the judgment matrix from layer B to layer A and from layer C to layer B given by each expert, the corresponding single-layer priority ranking weight W is calculated according to an AHP method k ＝[w 1k ，w 2k ，w 3k ，…，w pk ]，k=1, 2,3, …, q; wherein p is the number of measurement indexes contained in the reduced index system;
s322: fusing the single-layer priority ranking weights obtained in the step S321 by adopting a Dempster combination rule in a D-S evidence theory to obtain a ranking weight of the layer C to the layer A and a corresponding single-layer measurement priority weight W k ＝[w 1 ，w 2 ，w 3 ，…，w p ]；
The sorting weight of the layer C to the layer A is obtained by the following method:
giving trust functions based on condensed metrics
Assuming that BEL1 and BEL2 are two trust functions under the same index system, m 1 、m 2 Respectively corresponding probability assignment, and focal element is A respectively 1 ，A 2 ，…，A k And B 1 ，B 2 ，…，B r Then
S33: and (3) acquiring judgment matrixes of the layers B to A and the layers C to B according to the result obtained in the step (S32), respectively obtaining a weight vector and a maximum characteristic value, and carrying out consistency test on the expert scoring result to finally obtain a new layer C sequencing weight of the layers A so as to realize comprehensive sequencing of the reduced evaluation index set.
Specifically, 2 experts are arranged, and through the evaluation of expert 1, a judgment matrix from layer B to layer A and from layer C to layer B is given
Calculating to obtain weight vector and maximum eigenvalue, and performing consistency test to obtain expert 1 ordering vector as
Similarly, the ranking weights of layer C to layer A are obtained as follows:
W 1 ＝[0.4168，0.1778，0.0387，0.1955，0.0494，0.0156，0.0740，0.0246，0.0076]the method comprises the steps of carrying out a first treatment on the surface of the Its corresponding prioritization is: c 1 ＞c 4 ＞c 2 ＞c 7 ＞c 5 ＞c 3 ＞c 8 ＞c 6 ＞c 9 。
Through the evaluation of expert 2, the judgment matrix from layer B to layer A and from layer C to layer B is given as
Calculating to obtain weight vector and maximum eigenvalue, and performing consistency test to obtain expert 1 ordering vector as
Similarly, the ranking weights of layer C to layer A are obtained as follows:
W 2 ＝[0.4344，0.1909，0.0498，0.1708，0.0728，0.0159，0.0437，0.0175，0.0042]the method comprises the steps of carrying out a first treatment on the surface of the Its corresponding prioritization is: c 1 ＞c 2 ＞c 4 ＞c 5 ＞c 3 ＞c 7 ＞c 8 ＞c 6 ＞c 9 。
In fact, because the expert' S weighting has strong subjectivity, the two experts have different importance evaluations of specific index measurement, namely different weighting coefficients are given, but the evaluation result of each expert is still used as a favorable evidence, and only the D-S theory is needed to be adopted to fuse the evaluation information, so that the ranking weight of the layer C to the layer A is W D-s ＝[0.7022，0.1316，0.0075，0.1295，0.0139，0.0009，0.0125，0.0017，0.0001]The corresponding single layer metric is prioritized as c 1 ＞c 2 ＞c 4 ＞c 5 ＞c 7 ＞c 3 ＞c 8 ＞c 6 ＞c 9 ；
Acquiring judgment matrixes from layer B to layer A and from layer C to layer B according to the obtained result as follows
Respectively obtaining a weight vector and a maximum characteristic value, and carrying out consistency test on expert scoring results to finally obtain new ranking weight of the layer C to the layer A so as to realize comprehensive ranking of the reduced evaluation index set
W new ＝[0.2893，0.2572，0.1286，0.1211，0.1038，0.0346，0.0363，0.0218，0.0073]。
Further, S4: based on the weight calculated by the AHP method and the D-S evidence theory, establishing an electromechanical system PHM system performance evaluation model based on fuzzy comprehensive evaluation to obtain a comprehensive conclusion of PHM system evaluation.
Specifically, S41, selecting fuzzy comments of PHM system performance evaluation indexes: according to the influence degree of qualitative analysis on the evaluation index, classifying the comment set into 5 grades I= { very serious, general, slight and very slight }, wherein the corresponding weight scores are {9,7,5,3,1};
s42, selecting a fuzzy matrix and a fuzzy membership function: determining a result layer D according to expert experience and PHM system function design requirements i And metric layer C j Is s ij Representation, construction of a fuzzy evaluation matrix s= (S) ij ) And determining a membership function;
for qualitative and quantitative measurement, performing normalization treatment on PHM system performance measurement by adopting a semi-trapezoidal membership function, and then converting the fuzzy judgment matrix into a membership matrix S= (S) ij ) Wherein s is ij Is the relative membership of the ith result to the jth metric;
for positive and negative metrics, semi-trapezoidal membership functions are chosen for processing, i.e., x=s ij ，a＝max(e j )，b＝min(e j ) The method comprises the steps of carrying out a first treatment on the surface of the The membership functions of the semi-trapezium, which are larger and smaller, are defined as:
wherein the values in brackets are smaller.
S43, fuzzy comprehensive evaluation of performance of the PHM system of the electromechanical system: ranking weight vectors according to overall priority of metric layer relative to target layer
And the membership function matrix S between the evaluation result and each measurement is calculated to obtain the evaluation result of the performance of the PHM system of the electromechanical system as +.>
The foregoing has shown and described the basic principles, principal features and advantages of the invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, and that the above embodiments and descriptions are merely illustrative of the principles of the present invention, and various changes and modifications may be made without departing from the spirit and scope of the invention, which is defined in the appended claims. The scope of the invention is defined by the appended claims and equivalents thereof.
Claims (2)
1. A comprehensive evaluation method of an electromechanical system PHM system is characterized by comprising the following steps,
s1: constructing a performance evaluation index system of the PHM system of the electromechanical system by using an AHP method, and sequencing the priority of each index;
s2: cutting the saturation index set by utilizing a rough set theory to obtain a reduced evaluation index set;
s3: performing weight calculation on the reduced evaluation index set by using an AHP method, and performing comprehensive weight fusion by combining with a D-S evidence theory, so as to realize comprehensive sequencing of the reduced evaluation index set;
s4: based on the weight calculated by the AHP method and the D-S evidence theory, establishing an electromechanical system PHM system performance evaluation model based on fuzzy comprehensive evaluation to obtain a comprehensive conclusion of PHM system evaluation;
the specific operation of step S1 includes the following steps,
s11: dividing the performance evaluation index of the PHM system into 4 layers by using an AHP method, and constructing a multi-layer performance evaluation index system;
s12, constructing a complementary judgment matrix of the index: after a multi-layer performance evaluation index system structure is established by using an AHP method, comparing each index with the priority of the previous layer in pairs, and carrying out weight assignment by adopting a common scale method of 1-9 so as to complete the construction of a pair comparison judgment matrix; by a' = [ a ij ] n×m A pair-wise comparison judgment matrix representing layer B versus layer A, wherein a ij Is the i-th index B in layer B i And the j-th index B j Score comparison value of a) ij The larger B i Ratio B j The priority of the relative layer a is high; a, a ij =1, representing having the same priority; and so on, a comparison judgment matrix of the layer C relative to the layer B is established and is expressed as B' = [ B ] ij ] n×m ；
S13, calculating the weight and sequencing single-layer priority: the priority weights of the criteria of layer B relative to layer a satisfy a' w=λ max W, where lambda max As the largest eigenvalue of matrix a', w= (W 1 ，w 2 ，…，w n ) T The vector corresponding to the weight represents the importance degree of classification sequencing in the layer B and is also a single-layer priority sequencing vector of the layer B; and so on, the single-layer priority of the layer C can be obtained;
s14, total priority ranking calculation: the total prioritization is a ranking of weights of all criteria of a single layer relative to the target layer, the criteria layer B contains m criteria b=b 1 ，B 2 ，…，B m The total priority weights of the layers A are respectively as follows
Layer C contains n metrics c=c 1 ，C 2 ，…，C n For criterion B j Is given by a single layer priority weight of
At this time C i The total prioritization weights with respect to layer a are: />
S15, consistency test: definition of the consistency index ci= (λ) max -n)/(n-1), wherein n is the number of measures of layer C, calculating a consistency ratio cr=ci/RI from the average random consistency index RI, and determining that the consistency condition is not satisfied when CR is not less than 0.10; if a certain index of layer C is for criterion B j The uniformity and average uniformity index of (C) are respectively Cl j 、Rl j The total ranking weight of layer B is
The overall rank uniformity ratio of layer C is +.>
The 4 layers in the step S11 comprise a target layer A, a criterion layer B, a measurement layer C and a result layer D;
the target layer A comprises PHM system performance evaluation A;
criterion layer B includes testability B 1 Fault diagnosis B 2 Failure prediction B 3 Occupy resource B 4 Technical maturity B 5 ；
The testability B 1 The corresponding metric layer C comprises a failure detection rate C 1 Fault isolation rate C 2 False alarm rate C 3 Failure prediction rate C 4 ；
The fault diagnosis B 2 The corresponding metric layer C comprises the accuracy C of fault diagnosis 5 Accuracy C of fault diagnosis 6 Stability of fault diagnosis C 7 Sensitivity C of fault diagnosis 8 ；
The failure prediction B 3 The corresponding metric layer C includes the accuracy of the fault predictionC 9 Relative accuracy of failure prediction C 10 Failure prediction interval C 11 Alpha-lambda Performance C 12 Failure prediction coverage C 13 ；
The occupied resource B 4 The corresponding metric layer C includes full period cost C 14 Software code length C 15 Storage capacity C 16 Complexity of system C 17 ；
Degree of maturity of the technology B 5 The corresponding metric layer C includes a testability technology maturity C 18 Technical maturity C of fault diagnosis 19 Technical maturity of failure prediction C 20 ；
The result layer D includes pass, partial pass, and no pass;
in step S13, the method for calculating the eigenvector W of the matrix A' is as follows
The specific operation of step S2 includes the following steps,
s21: set of n metrics giving a metrics layer
Wherein A is a target layer; b is a criterion layer corresponding to A, B= { B 1 ，B 2 ，…，B m }，B i For the ith criterion, B i E B; c is the metric layer set c= { C 1 ，c 2 ，…，c N }，C i Represents the ith metric, c i ＝(f i ，c，φ)，c i ∈C；/>
For measuring c i Weights for target layer a; phi is a measurement constraint, phi epsilon g x theta, g= (yes, no), yes represents that the measurement can be cut, and no represents that the measurement cannot be cut; θ is a clipping condition set; f (f) i C is i Is a measurement algorithm of (2);
s22: in the reduction process, the definition of the accuracy and the definition of the accuracy corresponding to the failure prediction in the criterion layer are similar without considering the related indexes of the resource occupation and the technical maturity, and the failure diagnosis and the accuracy in the criterion layer can be combined into a group; thus, 20 metrics may be reduced to 9 by attribute, and each index is divided by "class I-III" to represent "excellent", "good" and "bad", respectively;
the specific operation of step S3 includes the following steps,
s31: based on the judgment matrixes from layer B to layer A and from layer C to layer B given by each expert, calculating corresponding single-layer priority ranking weights according to an AHP method;
s32: fusing the priority ranking weights obtained in the step S31 by adopting a D-S evidence theory to obtain a ranking weight of the layer C on the layer A and a corresponding single-layer measurement priority ranking;
s33: obtaining judgment matrixes of the layers B to A and the layers C to B according to the result obtained in the step S32, respectively obtaining a weight vector and a maximum characteristic value, and carrying out consistency test on the expert scoring result to finally obtain a new layer C sequencing weight of the layer A so as to realize comprehensive sequencing of a reduced evaluation index set;
the specific operation steps of step S32 include,
s321: q PHM experts are arranged to participate in the performance evaluation, and based on the judgment matrix from layer B to layer A and from layer C to layer B given by each expert, the corresponding single-layer priority ranking weight W is calculated according to an AHP method k ＝[w 1k ，w 2k ，w 3k ，…，w pk ]K=1, 2,3, …, q; wherein p is the number of measurement indexes contained in the reduced index system;
s322: fusing the single-layer priority ranking weights obtained in the step S321 by adopting a Dempster combination rule in a D-S evidence theory to obtain a ranking weight of the layer C to the layer A and a corresponding single-layer measurement priority weight W k ＝[w 1 ，w 2 ，w 3 ，…，w p ]；
The sorting weight of the layer C to the layer A is obtained by the following method:
giving trust functions based on condensed metrics
Assuming that BEL1 and BEL2 are two trust functions under the same index system, m 1 、m 2 Respectively corresponding probability assignment, and focal element is A respectively 1 ，A 2 ，…，A k And B 1 ，B 2 ，…，B r Then
The specific operation of step S4 includes the following steps,
s41, selecting a PHM system performance evaluation index fuzzy comment: according to the influence degree of qualitative analysis on the evaluation index, classifying the comment set into 5 grades I= { very serious, general, slight and very slight }, wherein the corresponding weight scores are {9,7,5,3,1};
s42, selecting a fuzzy matrix and a fuzzy membership function: determining a result layer D according to expert experience and PHM system function design requirements i And metric layer C j Is s ij Representation, construction of a fuzzy evaluation matrix s= (S) ij ) And determining a membership function;
s43, fuzzy comprehensive evaluation of performance of the PHM system of the electromechanical system: ranking weight vectors according to overall priority of metric layer relative to target layer
And the membership function matrix S between the evaluation result and each measurement is calculated to obtain the evaluation result of the performance of the PHM system of the electromechanical system as +.>
2. An electromechanical system PHM according to claim 1The comprehensive evaluation method of the system is characterized in that: in step S42, for qualitative and quantitative metrics, the PHM performance metrics are normalized using a semi-trapezoidal membership function, and then the fuzzy evaluation matrix is converted into a membership matrix s= (S) ij ) Wherein s is ij Is the relative membership of the ith result to the jth metric;
for positive and negative metrics, semi-trapezoidal membership functions are chosen for processing, i.e., x=s ij ，a＝max(e j )，b＝min(e j ) The method comprises the steps of carrying out a first treatment on the surface of the The membership functions of the semi-trapezium, which are larger and smaller, are defined as:
wherein the values in brackets are smaller.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011573187.6A CN112686525B (en) | 2020-12-25 | 2020-12-25 | Comprehensive evaluation method of PHM system of electromechanical system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011573187.6A CN112686525B (en) | 2020-12-25 | 2020-12-25 | Comprehensive evaluation method of PHM system of electromechanical system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112686525A CN112686525A (en) | 2021-04-20 |
| CN112686525B true CN112686525B (en) | 2023-06-23 |
Family
ID=75452065
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011573187.6A Active CN112686525B (en) | 2020-12-25 | 2020-12-25 | Comprehensive evaluation method of PHM system of electromechanical system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112686525B (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112116276A (en) * | 2020-09-29 | 2020-12-22 | 广东电网有限责任公司广州供电局 | Transformer substation operation risk assessment method considering time-varying state of electrical main equipment |
-
2020
- 2020-12-25 CN CN202011573187.6A patent/CN112686525B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112116276A (en) * | 2020-09-29 | 2020-12-22 | 广东电网有限责任公司广州供电局 | Transformer substation operation risk assessment method considering time-varying state of electrical main equipment |
Non-Patent Citations (3)
| Title |
|---|
| A Comprehensive Evaluation assessment model to aviation maintenance safety risk Based on grey fuzzy theory;Wang Xiaolin, etc.;《ACM Digital Library》;全文 * |
| 基于PHM的复杂机电***健康管理技术研究;王林秋等;《计算机测量与控制》;第26卷(第1期);第27-30页 * |
| 基于证据理论的Bow-tie航空风险评估方法及应用;徐吉辉等;《火力与指挥控制》;第44卷(第10期);第1-7页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112686525A (en) | 2021-04-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hartson et al. | Criteria for evaluating usability evaluation methods | |
| Cheng et al. | Contractor selection using the analytic network process | |
| CN112633601B (en) | Method, device, equipment and computer medium for predicting disease event occurrence probability | |
| Chang et al. | A theoretical survey on Mahalanobis-Taguchi system | |
| Garg et al. | Granulized Z-VIKOR model for failure mode and effect analysis | |
| CN105512465B (en) | Based on the cloud platform safety quantitative estimation method for improving VIKOR methods | |
| CN110135708A (en) | Ball valve quality multi-step Fuzzy Comprehensive Evaluation based on AHP and comentropy | |
| CN106549826A (en) | Intelligent substation switch applied in network performance test appraisal procedure | |
| CN110543500A (en) | Power transmission and transformation equipment health assessment platform based on big data | |
| CN111882198A (en) | Project performance evaluation method and system | |
| Eddy et al. | Credit scoring models: Techniques and issues | |
| Ettensperger | Comparing supervised learning algorithms and artificial neural networks for conflict prediction: performance and applicability of deep learning in the field | |
| Zaied et al. | An integrated neutrosophic and TOPSIS for evaluating airline service quality | |
| Kumar et al. | Credit score prediction system using deep learning and k-means algorithms | |
| CN103345552A (en) | Method and device for assessing reliability of power ICT communication network | |
| CN112686525B (en) | Comprehensive evaluation method of PHM system of electromechanical system | |
| Dammak et al. | Crisp multi-criteria decision making methods: State of the art | |
| Terceño et al. | Prediction of business failure with fuzzy models | |
| CN115981778A (en) | Method and system for determining early warning model for monitoring container cluster | |
| CN115936841A (en) | Method and device for constructing credit risk assessment model | |
| Achimugu et al. | An adaptive fuzzy decision matrix model for software requirements prioritization | |
| CN113722195A (en) | Local area network operation evaluation system and method based on AHP (advanced health process) hierarchy analysis method | |
| CN114092216A (en) | Enterprise credit rating method, apparatus, computer device and storage medium | |
| CN107230005B (en) | Data processing method and equipment | |
| Azadfallah | Two new Eigenvector-based approaches to assign weights to decision makers in group decision making under multiple criteria |
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 |