CN110109005A - A kind of analog circuit fault test method based on sequential test - Google Patents

A kind of analog circuit fault test method based on sequential test Download PDF

Info

Publication number
CN110109005A
CN110109005A CN201910439946.0A CN201910439946A CN110109005A CN 110109005 A CN110109005 A CN 110109005A CN 201910439946 A CN201910439946 A CN 201910439946A CN 110109005 A CN110109005 A CN 110109005A
Authority
CN
China
Prior art keywords
fault
sample
vector
opt
bpa
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
Application number
CN201910439946.0A
Other languages
Chinese (zh)
Other versions
CN110109005B (en
Inventor
刘震
梅文娟
杜立
杨成林
周秀云
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Electronic Science and Technology of China
Original Assignee
University of Electronic Science and Technology of China
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by University of Electronic Science and Technology of China filed Critical University of Electronic Science and Technology of China
Priority to CN201910439946.0A priority Critical patent/CN110109005B/en
Publication of CN110109005A publication Critical patent/CN110109005A/en
Application granted granted Critical
Publication of CN110109005B publication Critical patent/CN110109005B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/28—Testing of electronic circuits, e.g. by signal tracer
    • G01R31/316—Testing of analog circuits
    • G01R31/3163—Functional testing

Landscapes

  • Engineering & Computer Science (AREA)
  • General Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Test And Diagnosis Of Digital Computers (AREA)

Abstract

The invention discloses a kind of analog circuit fault test methods based on sequential test, the validity of measuring point is assessed by conditional entropy, the characteristic value for the measuring point for selecting fault identification ability strong generates the evidence vector of expression malfunction recognition result as extreme learning machine input layer information；Belief function is generated based on D-S theory again, malfunction set is separated into several failure subsets according to similar spread through sex intercourse, and further generate the diagnostic model of failure subset, until malfunction is totally separated or is separated to optimum state, have the characteristics that fault identification precision is high, fault identification is high-efficient.

Description

A kind of analog circuit fault test method based on sequential test
Technical field
The invention belongs to circuit fault diagnosis and machine learning techniques fields, more specifically, are related to a kind of based on sequence Pass through the analog circuit fault test method of test.
Background technique
As analog circuit integrated level is promoted increasingly, electronic system research field is become to the status monitoring of complicated circuit Major issue.How efficiently to the malfunction of analog circuitry system carry out effectively monitoring become the field research hotspot it One.Existing analog circuit fault state identification method is broadly divided into the two class failures based on Heuristics and based on data-driven Diagnostic method.
Method for diagnosing faults based on Heuristics is the mechanism model relationship by being obtained by circuit, abstract test point with Dependence between each fault mode, and then efficient Test Strategy is generated, it is used for positioning failure, mainly includes that expert is System, graph search algorithm, fault tree method of formation and fault dictionary.However, the input that such method uses is binary information or use In the probabilistic information for indicating uncertain dependence, need to be constructed by the dependence on circuit structure, due to for The building process of the dependence model of complication system is complex, therefore such method relatively simple, mechanism that is confined to structure mostly The fault diagnosis of relatively conventional electronic system.Simultaneously as it is special to have ignored numerical value existing for analog circuit test signal itself It levies (amplitude of such as signal, frequency, variance etc.), such method can not establish quantitative fault identification model to analog circuit.
In recent years, with the rapid development of artificial intelligence technology, the method for diagnosing faults based on data-driven is produced.It should Class method is without establishing failure dependence model, the malfunction for the information identification circuit system that can be directly reflected by measuring point, Therefore become a kind of method of great vitality.Existing intelligent diagnosing method can be divided into: the event based on probabilistic neural network Hinder diagnostic method, SVM method for diagnosing faults, diagnosing information fusion fault method etc..
Intelligent diagnosing method is to complicated circuit system practical significance with higher.Since this method considers the intelligence of people Factor more meets the natural inference to real system, is a kind of up-and-coming method for diagnosing faults.However, existing be based on The method for diagnosing faults of machine learning is all directly to carry out Holistic modeling by collected whole measuring point information, considers mould emphatically The accuracy of identification of type entirety, and the otherness of each failure effective information in identification process is had ignored, therefore affect failure The efficiency of identification process, modeling process lack flexibility.
Summary of the invention
It is an object of the invention to overcome the deficiencies of the prior art and provide a kind of analog circuit faults based on sequential test Test method, the fault diagnosis tree-model based on extreme learning machine have fault identification come the malfunction of test simulation circuit The features such as precision is high, fault identification is high-efficient.
For achieving the above object, a kind of analog circuit fault test method based on sequential test of the present invention, it is special Sign is, comprising the following steps:
(1), the voltage value of different corner frequencies is acquired
Acquisition analog circuit is nonserviceabled voltage characteristic of the N group sample in different corner frequencies under lower and health status Value, is denoted as X={ x1,x2,...,xN, wherein voltage characteristic value xi={ xi,1,xi,2,...,xi,k,...,xi,T, xi,kTo adopt Voltage characteristic value of i-th of the sample of collection under k-th of corner frequency, T are corner frequency number；
Construct the desired output vector Y={ y of each sample1,y2,...,yi,...,yN, wherein yi={ yi,1,yi,2, yi,3,...yi,j,...,yi,M, M is malfunction sum, yi,jThe input Imitating circuit constituted for i-th group of voltage characteristic value The mark value of jth kind failure, if yi,j=1, then i-th group of voltage characteristic value can generate jth kind failure under constituting, otherwise yi,j=0；
(2), initialization failure tree-model
Construct the fault set S={ s of fault tree root nodes stand1,s2,...sN, and be arranged in fault identification model and selected Measuring point information aggregate T' is empty set, measuring point information aggregate T "={ t to be selected1,t2,...,tT, the evidence arrow of the measuring point to be selected is set Duration set m' is empty set；
(3), extreme learning machine fault identification model to be selected is generated
(3.1), measuring point information t to be selected is utilizedkConstruct corresponding extreme learning machine model E LMtk={ wtk,btk,βtk, In, wtkFor hidden layer weighted vector, btkFor the biasing of hidden layer, βtkFor output layer weight；
(3.2), the weighted vector w of hidden layer is generated at randomtk={ wtk,1,wtk,2,...wtk,p,...,wtk,H, wherein H For the total number of hidden layer neuron, wtk,pFor the weight of p-th of neuron；The random offset vector b for generating hidden layertk= {btk,1,btk,2,btk,3,...,btk,p,...,btk,H}；
(3.3), the output valve h of hidden layer neuron is calculatedtk；
Wherein, X " ' is T " and xtkThe input vector collectively formed；
(3.4), the output layer weight β of extreme learning machine is generatedtk；
(3.5), the evidence vector generated by extreme learning machine is calculated；
mtk=βtkhtk
(4), the conditional entropy numerical value f of the extreme learning machine evidence vector generated is calculatedtk；
(5), the corresponding measuring point to be selected of conditional entropy minimum value, which is chosen, as measuring point to be selected increased under the node inputs topt；
topt=argmin { ftk,tk∈T”}
Utilize toptCorresponding hidden layer weight and offset vector and output layer weighted vector construct the failure under the node Diagnostic model ELMS={ wtopt,btopt,βtopt, by mtoptIt is added in m', by toptIt is added in T'；
(6), the probability assignment function BPA of fault set is calculateds={ BPAS,1,BPAS,2,BPAS,3,....BPAS,i, ...BPAS,N, wherein BPAs,iFor the probability assignment function of i-th of sample, calculate as follows:
(7), S is divided by similitude propagation algorithm by multiple sub- fault sets according to probability assignment function
(7.1), the similarity between two sample probability assignment functions is calculated:
(7.2), the specimen amount of each sample is generated by Similarity Algorithm；
(7.2.1), initialization Attraction Degree parameter r={ ri,k|si,sk∈ S } it is null matrix, wherein ri,kIt is suitble to for sample k The accumulative Attraction Degree of the cluster centre of sample i；Initialize degree of membership function a={ ai,k|si,sk∈ S } it is null matrix, wherein ai,k It is suitble to the accumulative degree of membership of the cluster centre of sample i for sample k；
(7.2.2), update Attraction Degree function and degree of membership function are as follows:
ri,k=simii,k-max(ai,k'+simii,k')
(7.2.3), the specimens point e={ e for generating each sample1,e2,...,ei,...,eN, wherein eiFor i-th of sample Specimens point；
(7.2.4), judge whether the value of the specimens point e of each sample no longer changes, if do not changed, enters step (7.3), otherwise (7.2.2) is returned to；
(7.3), sub- fault set S'={ S is partitioned into according to the specimens point e of each sample11,S12,...,S1k,...,S1K, Wherein, S1kFor k-th of failure subset, meet:
S1k={ si|ei∈S1k}
(8), the fault diagnosis model of sub- fault set is generated
Judge sub- fault set S1,jWhether the minimum euclidean distance between the probability assignment function numerical value of middle corresponding sample, which is greater than, sets Definite value ε, if more than then by sub- fault set S1,jAs a child node for fault tree, and generate the fault diagnosis mould of the node Otherwise type generates new failure tree node S1,j, return step (3), until all sub- fault set processing is completed；
(9), fault diagnosis is carried out to analog circuit according to the fault tree models of generation
(9.1), acquisition analog circuit is denoted as x'={ x' in the voltage characteristic value of different corner frequencies1,x'2,...,x 'T, initialization evidence matrix m' is empty set, and input vector X is arrangedoptFor empty set；
(9.2), by the input vector t of the selection in fault diagnosis modeloptXopt is added, and calculates hidden layer output hopt；
It calculates the output evidence vector m " of extreme learning machine and is added in m'；
M "=βopthopt
(9.3), probability assignment function BPA is calculated:
(9.4), according to Similarity Algorithm find that the system nonserviceabled from fault set S', and using S' as new Root node, if S' is leaf node, using the corresponding malfunction of BPA maximum value as diagnostic result, otherwise return step (9.2)。
Goal of the invention of the invention is achieved in that
A kind of analog circuit fault test method based on sequential test of the present invention assesses the effective of measuring point by conditional entropy Property, the characteristic value for the measuring point for selecting fault identification ability strong generates expression malfunction as extreme learning machine input layer information The evidence vector of recognition result；Belief function is generated based on D-S theory again, is separated malfunction set according to similar spread through sex intercourse At several failure subsets, and the diagnostic model of failure subset is further generated, until malfunction is totally separated or is separated to Optimum state has the characteristics that fault identification precision is high, fault identification is high-efficient.
Detailed description of the invention
Fig. 1 is a kind of analog circuit fault test method flow chart based on sequential test of the present invention；
Fig. 2 is the structure chart of fault tree diagnostic model；
Fig. 3 is simulation circuit structure figure to be measured in embodiment；
Fig. 4 is the fault tree diagnostic model structure chart of analog circuit to be measured in embodiment.
Specific embodiment
A specific embodiment of the invention is described with reference to the accompanying drawing, preferably so as to those skilled in the art Understand the present invention.Requiring particular attention is that in the following description, when known function and the detailed description of design perhaps When can desalinate main contents of the invention, these descriptions will be ignored herein.
Embodiment
Fig. 1 is a kind of analog circuit fault test method flow chart based on sequential test of the present invention.
In the present embodiment, as shown in Figure 1, a kind of analog circuit fault test method based on sequential test of the present invention, The following steps are included:
S1: the voltage value of different corner frequencies is acquired
Acquisition analog circuit is nonserviceabled voltage characteristic of the N group sample in different corner frequencies under lower and health status Value, is denoted as X={ x1,x2,...,xN, wherein voltage characteristic value xi={ xi,1,xi,2,...,xi,k,...,xi,T, xi,kTo adopt Voltage characteristic value of i-th of the sample of collection under k-th of corner frequency, T are corner frequency number；
Construct the desired output vector Y={ y of each sample1,y2,...,yi,...,yN, wherein yi={ yi,1,yi,2, yi,3,...yi,j,...,yi,M, M is malfunction sum, yi,jThe input Imitating circuit constituted for i-th group of voltage characteristic value The mark value of jth kind failure, if yi,j=1, then i-th group of voltage characteristic value can generate jth kind failure under constituting, otherwise yi,j=0；
S2: initialization failure tree-model
Construct the fault set S={ s of fault tree root nodes stand1,s2,...sN, and be arranged in fault identification model and selected Measuring point information aggregate T' is empty set, measuring point information aggregate T "={ t to be selected1,t2,...,tT, the evidence arrow of the measuring point to be selected is set Duration set m' is empty set；
S3: extreme learning machine fault identification model to be selected is generated
According to the information of measuring point each in T ", the extreme learning machine model about each measuring point information of T " is constructed.About measuring point Information tk, construct corresponding extreme learning machine model E LMtk={ wtk,btk,βtk, wherein wtkFor hidden layer weighted vector, btk For the biasing of hidden layer, βtkFor output layer weight.ELMtkBuilding process it is as follows:
S3.1, the random weighted vector w for generating hidden layertk={ wtk,1,wtk,2,...wtk,p,...,wtk,H, wherein H is The total number of hidden layer neuron, wtk,pFor the weight of p-th of neuron.The random offset vector b for generating hidden layertk= {btk,1,btk,2,btk,3,...,btk,p,...,btk,H}。
S3.2, the output valve h for calculating hidden layer neurontk:
Wherein, X " ' is T " and xtkThe input vector collectively formed:
X " '={ xi|ti∈Torti=tk}
S3.3, the output layer weight β for generating extreme learning machinetk:
S3.4 calculates the evidence vector generated by extreme learning machine:
mtk=βtkhtk
S4, the conditional entropy numerical value f for calculating the extreme learning machine evidence vector generatedtk:
S5, the corresponding measuring point of conditional entropy minimum value is chosen as measuring point increased under node input topt:
topt=argmin { ftk,tk∈T”}
And use toptCorresponding hidden layer weight and offset vector and output layer weighted vector construct the failure under the node Diagnostic model ELMS={ wtopt,btopt,βtopt, by mtoptIt is added in m', by toptIt is added in T'.
S6, the probability assignment function BPA for calculating fault sets={ BPAS,1,BPAS,2,BPAS,3,....BPAS,i, ...BPAS,N, wherein BPAs,iFor the probability assignment function of i-th of sample, calculate as follows:
The probability assignment function that S7, basis are calculated, is divided into multiple sub- failures for S by similitude propagation algorithm Collection, process are as follows:
Similarity between S7.1, two sample probability assignment functions of calculating:
S7.2, the specimen amount that each sample is generated by Similarity Algorithm, the specific steps are as follows:
S7.2.1, initialization Attraction Degree parameter r={ ri,k|si,sk∈ S } it is null matrix, wherein ri,kIt is suitble to sample for sample k The accumulative Attraction Degree of the cluster centre of this i.Initialize degree of membership function a={ ai,k|si,sk∈ S } it is null matrix, wherein ai,kFor Sample k is suitble to the accumulative degree of membership of the cluster centre of sample i.
S7.2.2, update Attraction Degree function and degree of membership function are as follows:
ri,k=simii,k-max(ai,k'+simii,k')
S7.2.3, the specimens point e={ e for generating each sample1,e2,...,ei,...,eN, wherein eiFor i-th sample Specimens point calculates as follows:
ei=arg max { ai,k+ri,k}
If the value of e no longer changes, S7.3 is entered step, S7.2.2 is otherwise returned.
S7.3, according to specimens point information, be partitioned into sub- fault set S'={ S11,S12,...,S1k,...,S1K, S1kFor kth A failure subset:
S1k={ si|ei∈S1k}
S8, the fault diagnosis model for generating sub- fault set
For sub- fault set S1,j, the minimum for judging to correspond in the fault set between the probability assignment function numerical value of sample is European Distance is greater than the set value ε, if so, illustrating that the sample in the set has separated, by the sub- fault set S1,jAs fault tree One child node, and the fault diagnosis model of the node is generated, otherwise generate new failure tree node S1,j, and return step S3, Until the processing of sub- fault set is completed, to obtain fault tree diagnostic model shown in Fig. 2.
S9, fault diagnosis is carried out to analog circuit according to the fault tree models of generation, its step are as follows:
S9.1, acquisition analog circuit are denoted as x'={ x' in the voltage characteristic value of different corner frequencies1,x'2,...,x'T}。 Initialization evidence matrix m' is empty set, and input vector X is arrangedoptFor empty set
S9.2, the input vector t by the selection in fault diagnosis modeloptXopt is added, and calculates hidden layer output hopt；
It calculates the output evidence vector m " of extreme learning machine and is added in m'；
M "=βopthopt
S9.3, probability assignment function BPA is calculated:
S9.4, find that the system nonserviceabled according to Similarity Algorithm from fault set S', and using S' as new Root node, if S' is leaf node, using the corresponding malfunction of BPA maximum value as diagnostic result, otherwise return step S9.2。
Example
Technical effect to illustrate the invention carries out implementation verifying to the present invention using for certain analog circuit.Such as Fig. 3 institute Show, circuit is made of 4 second order filters and an adder, using Pspice software to its modeling and simulation.R1, R2, R3, The tolerance of R4, R5, R6, R7 and R8 are that the tolerance of ± 10%, C1, C2, C3, C4, C5, C6, C7 and C8 are ± 5%, amplifier The tolerance of gain A v1, Av2, Av3 and Av4 are that the tolerance of ± 1%, R9, R10 and R11 are ± 1%.Analysis learns that the circuit has 4 corner frequencies: 10Hz, 100Hz, 10kHz and 100kHz.Analog circuit single fault probability of occurrence accounts for about 80%, therefore only considers The status monitoring of single fault and health control.
Av is set1,Av2,Av3And Av4In (1.1~1.5%) Xn, (1.6~2%) Xn, (2.1~2.5%) Xn, (2.6 ~3.0%) Xn, (3.1~3.5%) Xn, (3.6~4.0%) Xn, (4.1~4.5%) Xn, (4.6~5.0%) Xn, (5.1~ 5.5%) Xn, (5.6~6.0%) Xn variation and repeatedly Monte-Carlo emulation in totally 10 parameter sections, only consider it is single therefore Hinder element: the parameter value of only one element of any time even variation, remaining element in its different parameters section are holding Even variation in poor range；All elements even variation all in range of tolerable variance when normal state.It is obtained under normal state and malfunction Take voltage characteristic vector under 7 corner frequencies.
To measure and comparing test performance, the diagnosis of typical BP network and Hidden Markov Model HMM and this method is utilized Discrimination is compared, and the fault recognition rate of each method is as shown in table 1, in general, the present invention in fault recognition rate compared with Typical method is obviously improved.
State class Normal state Av1 increases Av1 reduces Av2 increases Av2 reduces
Label S1 S2 S3 S4 S5
BP 46.59 37.11 39.70 49.04 55.41
HMM 100.0 80.0 92.22 75.56 74.44
The present invention 97.0 100.0 98.1 100.0 100.0
State class Av3 increases Av3 reduces Av4 increases Av4 reduces It is average
Label S6 S7 S8 S9 --
BP 67.93 78.44 67.85 57.93 55.56
HMM 100.0 100.0 100.0 100.0 91.36
The present invention 99.0 98.6 100.0 100.0 99.28
Table 1
Fig. 4 is the fault tree models for producing sequential Diagnostic Strategy generated by the invention, it can be seen that by this hair The fault diagnosis model of bright generation reduces the use of voltage characteristic value, improves the service efficiency of mode input characteristic value, increases The strong flexibility of diagnostic model.
Although the illustrative specific embodiment of the present invention is described above, in order to the technology of the art Personnel understand the present invention, it should be apparent that the present invention is not limited to the range of specific embodiment, to the common skill of the art For art personnel, if various change the attached claims limit and determine the spirit and scope of the present invention in, these Variation is it will be apparent that all utilize the innovation and creation of present inventive concept in the column of protection.

Claims (3)

1. a kind of analog circuit fault test method based on sequential test, which comprises the following steps:
(1), the voltage value of different corner frequencies is acquired
N group sample is remembered in the voltage characteristic value of different corner frequencies under acquisition analog circuit is nonserviceabled and under health status For X={ x1,x2,...,xN, wherein voltage characteristic value xi={ xi,1,xi,2,...,xi,k,...,xi,T, xi,kIt is the of acquisition Voltage characteristic value of the i sample under k-th of corner frequency, T are corner frequency number；
Construct the desired output vector Y={ y of each sample1,y2,...,yi,...,yN, wherein yi={ yi,1,yi,2,yi,3, ...yi,j,...,yi,M, M is malfunction sum, yi,jThe input Imitating circuit jth kind constituted for i-th group of voltage characteristic value The mark value of failure, if yi,j=1, then i-th group of voltage characteristic value can generate jth kind failure under constituting, otherwise yi,j=0；
(2), initialization failure tree-model
Construct the fault set S={ s of fault tree root nodes stand1,s2,...sN, and be arranged in fault identification model and selected measuring point Information aggregate T' is empty set, measuring point information aggregate T "={ t to be selected1,t2,...,tT, the evidence vector set of the measuring point to be selected is set Conjunction m' is empty set；
(3), extreme learning machine fault identification model to be selected is generated
(3.1), measuring point information t to be selected is utilizedkConstruct corresponding limit study habit machine model E LMtk={ wtk,btk,βtk, wherein wtkFor hidden layer weighted vector, btkFor the biasing of hidden layer, βtkFor output layer weight；
(3.2), the weighted vector w of hidden layer is generated at randomtk={ wtk,1,wtk,2,...wtk,p,...,wtk,H, wherein H is hidden Hide the total number of layer neuron, wtk,pFor the weight of p-th of neuron；The random offset vector b for generating hidden layertk={ btk,1, btk,2,btk,3,...,btk,p,...,btk,H}；
(3.3), the output valve h of hidden layer neuron is calculatedtk；
Wherein, X " ' is T " and xtkThe input vector collectively formed；
(3.4), the output layer weight β of extreme learning machine is generatedtk；
(3.5), the evidence vector generated by extreme learning machine is calculated；
mtk=βtkhtk
(4), the conditional entropy numerical value f of the extreme learning machine evidence vector generated is calculatedtk；
(5), the corresponding measuring point to be selected of conditional entropy minimum value, which is chosen, as measuring point to be selected increased under the node inputs topt；
topt=argmin { ftk,tk∈T”}
Utilize toptCorresponding hidden layer weight and offset vector and output layer weighted vector construct the fault diagnosis under the node Model E LMS={ wtopt,btopt,βtopt, by mtoptIt is added in m', by toptIt is added in T'；
(6), the probability assignment function BPA of fault set is calculateds={ BPAS,1,BPAS,2,BPAS,3,....BPAS,i,...BPAS,N, Wherein, BPAs,iFor the probability assignment function of i-th of sample, calculate as follows:
(7), S is divided by similitude propagation algorithm by multiple sub- fault sets according to probability assignment function
(7.1), the similarity between two sample probability assignment functions is calculated:
(7.2), the specimen amount of each sample is generated by Similarity Algorithm；
(7.2.1), initialization Attraction Degree parameter r={ ri,k|si,sk∈ S } it is null matrix, wherein ri,kIt is suitble to sample i for sample k Cluster centre accumulative Attraction Degree；Initialize degree of membership function a={ ai,k|si,sk∈ S } it is null matrix, wherein ai,kFor sample This k is suitble to the accumulative degree of membership of the cluster centre of sample i；
(7.2.2), update Attraction Degree function and degree of membership function are as follows:
ri,k=simii,k-max(ai,k'+simii,k')
(7.2.3), the specimens point e={ e for generating each sample1,e2,...,ei,...,eN, wherein eiFor the mark of i-th of sample This point；
(7.2.4), judge whether the value of the specimens point e of each sample no longer changes, if do not changed, enter step (7.3), Otherwise (7.2.2) is returned；
(7.3), sub- fault set S'={ S is partitioned into according to the specimens point e of each sample11,S12,...,S1k,...,S1K, wherein S1kFor k-th of failure subset, meet:
S1k={ si|ei∈S1k}
(8), the fault diagnosis model of sub- fault set is generated
Judge sub- fault set S1,jWhether the minimum euclidean distance between the probability assignment function numerical value of middle corresponding sample is greater than the set value ε, if more than then by sub- fault set S1,jAs a child node for fault tree, and the fault diagnosis model of the node is generated, it is no Then generate new failure tree node S1,j, return step (3), until all sub- fault set processing is completed；
(9), fault diagnosis is carried out to analog circuit according to the fault tree models of generation
(9.1), acquisition analog circuit is denoted as x'={ x' in the voltage characteristic value of different corner frequencies1,x'2,...,x'T, just Beginningization evidence matrix m' is empty set, and input vector X is arrangedoptFor empty set；
(9.2), by the input vector t of the selection in fault diagnosis modeloptXopt is added, and calculates hidden layer output hopt；
It calculates the output evidence vector m " of extreme learning machine and is added in m'；
M "=βopthopt
(9.3), probability assignment function BPA is calculated:
(9.4), according to Similarity Algorithm find that the system nonserviceabled from fault set S', and using S' as new root section Point, if S' is leaf node, using the corresponding malfunction of BPA maximum value as diagnostic result, otherwise return step (9.2).
2. the analog circuit fault test method according to claim 1 based on sequential test, which is characterized in that described defeated Enter vector X " ' satisfaction:
X " '={ xi|ti∈Torti=tk}。
3. the analog circuit fault test method according to claim 1 based on sequential test, which is characterized in that described The specimens point e of i sampleiCalculation are as follows:
ei=arg max { ai,k+ri,k}
Wherein, ri,kIt is suitble to the accumulative Attraction Degree of the cluster centre of sample i, a for sample ki,kIt is suitble to the cluster of sample i for sample k The accumulative degree of membership at center.
CN201910439946.0A 2019-05-24 2019-05-24 Analog circuit fault testing method based on sequential testing Active CN110109005B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910439946.0A CN110109005B (en) 2019-05-24 2019-05-24 Analog circuit fault testing method based on sequential testing

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910439946.0A CN110109005B (en) 2019-05-24 2019-05-24 Analog circuit fault testing method based on sequential testing

Publications (2)

Publication Number Publication Date
CN110109005A true CN110109005A (en) 2019-08-09
CN110109005B CN110109005B (en) 2021-03-30

Family

ID=67492154

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910439946.0A Active CN110109005B (en) 2019-05-24 2019-05-24 Analog circuit fault testing method based on sequential testing

Country Status (1)

Country Link
CN (1) CN110109005B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111368885A (en) * 2020-02-24 2020-07-03 大连理工大学 Aero-engine gas circuit fault diagnosis method based on deep learning and information fusion
CN113484738A (en) * 2021-05-25 2021-10-08 北京航空航天大学 Circuit fault diagnosis method based on multi-feature information fusion

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH09135168A (en) * 1995-10-13 1997-05-20 Lg Semicon Co Ltd Converting characteristic test circuit of a/d converter and its method
JP2008172634A (en) * 2007-01-12 2008-07-24 Denso Corp Failure detector of ad converter
CN101231672A (en) * 2008-02-02 2008-07-30 湖南大学 Method for diagnosing soft failure of analog circuit base on modified type BP neural network
US20120218674A1 (en) * 2011-02-28 2012-08-30 Schultz Neal H Overcurrent Protection System
CN104090227A (en) * 2014-07-16 2014-10-08 电子科技大学 Measuring point selection method in analog circuit fault diagnosis
CN105629156A (en) * 2016-03-10 2016-06-01 电子科技大学 Genetic programming-based analog circuit fault test optimal sequential search method
CN106483449A (en) * 2016-09-09 2017-03-08 电子科技大学 Based on deep learning and the analog-circuit fault diagnosis method of Complex eigenvalues
US20170193538A1 (en) * 2016-01-06 2017-07-06 Oracle International Corporation System and method for determining the priority of mixed-type attributes for customer segmentation
WO2018096492A1 (en) * 2016-11-24 2018-05-31 Tvs Motor Company Limited Fault diagnostic system for a two-wheeled vehicle and a control method thereof
CN109581203A (en) * 2018-11-07 2019-04-05 电子科技大学 Survey post-simulation method for diagnosing faults based on genetic algorithm

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH09135168A (en) * 1995-10-13 1997-05-20 Lg Semicon Co Ltd Converting characteristic test circuit of a/d converter and its method
JP2008172634A (en) * 2007-01-12 2008-07-24 Denso Corp Failure detector of ad converter
CN101231672A (en) * 2008-02-02 2008-07-30 湖南大学 Method for diagnosing soft failure of analog circuit base on modified type BP neural network
US20120218674A1 (en) * 2011-02-28 2012-08-30 Schultz Neal H Overcurrent Protection System
CN104090227A (en) * 2014-07-16 2014-10-08 电子科技大学 Measuring point selection method in analog circuit fault diagnosis
US20170193538A1 (en) * 2016-01-06 2017-07-06 Oracle International Corporation System and method for determining the priority of mixed-type attributes for customer segmentation
CN105629156A (en) * 2016-03-10 2016-06-01 电子科技大学 Genetic programming-based analog circuit fault test optimal sequential search method
CN106483449A (en) * 2016-09-09 2017-03-08 电子科技大学 Based on deep learning and the analog-circuit fault diagnosis method of Complex eigenvalues
WO2018096492A1 (en) * 2016-11-24 2018-05-31 Tvs Motor Company Limited Fault diagnostic system for a two-wheeled vehicle and a control method thereof
CN109581203A (en) * 2018-11-07 2019-04-05 电子科技大学 Survey post-simulation method for diagnosing faults based on genetic algorithm

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
GUOMING SONG等: "Multi-classifier fusion approach based on data clustering for analog circuits fault diagnosis", 《2009 IEEE 8TH INTERNATIONAL CONFERENCE ON ASIC》 *
何星等: "基于限定样本序贯极端学习机的模拟电路在线故障诊断", 《控制与决策》 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111368885A (en) * 2020-02-24 2020-07-03 大连理工大学 Aero-engine gas circuit fault diagnosis method based on deep learning and information fusion
CN111368885B (en) * 2020-02-24 2021-12-03 大连理工大学 Gas circuit fault diagnosis method for aircraft engine
CN113484738A (en) * 2021-05-25 2021-10-08 北京航空航天大学 Circuit fault diagnosis method based on multi-feature information fusion

Also Published As

Publication number Publication date
CN110109005B (en) 2021-03-30

Similar Documents

Publication Publication Date Title
Wu et al. Evolving RBF neural networks for rainfall prediction using hybrid particle swarm optimization and genetic algorithm
CN109767312B (en) Credit evaluation model training and evaluation method and device
CN105487526B (en) A kind of Fast RVM sewage treatment method for diagnosing faults
CN109669087A (en) A kind of method for diagnosing fault of power transformer based on Multi-source Information Fusion
CN110007652A (en) A kind of Hydropower Unit degradation trend interval prediction method and system
CN106295186A (en) A kind of method and system of aided disease diagnosis based on intelligent inference
CN109145516A (en) A kind of analog circuit fault recognition methods based on modified extreme learning machine
CN103810266B (en) Semantic network target recognition sentences card method
CN110458373A (en) A kind of method of crime prediction and system of the fusion of knowledge based map
Widiputra et al. Multiple time-series prediction through multiple time-series relationships profiling and clustered recurring trends
CN110109005A (en) A kind of analog circuit fault test method based on sequential test
CN110164129A (en) Single Intersection multi-lane traffic flow amount prediction technique based on GERNN
CN110533072A (en) Based on the SOAP service similarity calculation and clustering method of Bigraph structure under Web environment
da Silva et al. Validity index-based vigilance test in adaptive resonance theory neural networks
Galea et al. Simultaneous ant colony optimization algorithms for learning linguistic fuzzy rules
Hong et al. A cooperative method for supervised learning in spiking neural networks
CN109408896A (en) A kind of anerobic sowage processing gas production multi-element intelligent method for real-time monitoring
CN115293249A (en) Power system typical scene probability prediction method based on dynamic time sequence prediction
Li et al. A short-term wind power forecasting method based on NWP wind speed fluctuation division and clustering
CN103077404A (en) Local incremental visual clustering method based on Markov random field
CN106202986A (en) A kind of tonsillitis Forecasting Methodology based on increment type neural network model and prognoses system
Stammer et al. Learning by Self-Explaining
Liu et al. Fault prediction for satellite communication equipment based on deep neural network
Lasota et al. Enhancing intelligent property valuation models by merging similar cadastral regions of a municipality
Gök Semantic and goal-oriented signal processing: semantic extraction

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