CN103824135A - Analogue circuit failure prediction method - Google Patents
Analogue circuit failure prediction method Download PDFInfo
- Publication number
- CN103824135A CN103824135A CN201410088347.6A CN201410088347A CN103824135A CN 103824135 A CN103824135 A CN 103824135A CN 201410088347 A CN201410088347 A CN 201410088347A CN 103824135 A CN103824135 A CN 103824135A
- Authority
- CN
- China
- Prior art keywords
- fault
- failure
- analog circuit
- vector
- health degree
- 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
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses an analogue circuit failure prediction method which comprises the following steps: performing Monte Carlo analysis on various elements of an analogue circuit in a failure-free section and extracting various frequency band signal energy, normalizing the extracted frequency band signal energy to obtain a feature vector; training a BP neural network; judging failure modes with occurrence trends, extracting a failure prediction feature vector when the element is at the initial value, extracting the failure prediction feature vector when a detected circuit is in work, computing the cosine angle distance to represent the health degree of the element, computing the health degree threshold value when the element is in failure, and optimally selecting a kernel function width factor of a relevance vector machine algorithm, and performing the failure prediction on the analogue circuit. The method can be used for a real-time system, and can be further used for a non-real-time system, a failure of the linear analogue circuit can be predicted, and a failure of the non-linear analogue circuit can be predicted, and failures of main elements such as resistor, inductor and the capacitor in the analogue circuit can be predicted.
Description
Technical field
The present invention relates to a kind of analog circuit fault Forecasting Methodology, specifically, relate to a kind of method that failure prediction model is predicted analog circuit fault of setting up.
Background technology
Mimic channel is widely used in the equipment such as Jia Yong electricity Qi ﹑ commercial production Xian ﹑ automobile and Aero-Space, and the fault of mimic channel will cause performance Xia Jiang ﹑ function Shi Ling ﹑ delay of response and other electronic failures of equipment.Therefore the state of mimic channel being assessed, is very necessary.
The state estimation of mimic channel generally comprises fault diagnosis and failure prediction.Wherein fault diagnosis development is very fast, and in a large amount of research work, the accuracy rate of fault diagnosis all can reach 99% left and right.The achievement in research of current analog circuit fault prediction is generally the particular element for mimic channel, rather than for circuit entirety.A difficulty cannot predicting circuit entirety is seldom to have method can accurately describe the hydraulic performance decline of each element of mimic channel, i.e. health degree decline.Simultaneously current rarely have method to predict the fault of non-linear simulation circuit.
Method Using Relevance Vector Machine is a kind of regression forecasting algorithm based on Bayesian frame, and fast operation, is applicable to online detection, and the existing precision of prediction that studies have shown that Method Using Relevance Vector Machine is higher than the algorithms most in use such as support vector machine and neural network.In Method Using Relevance Vector Machine algorithm, the width factor of its kernel function has great impact to precision of prediction, and the empirical methods that adopt obtained more in the past.
Summary of the invention
The technical problem to be solved in the present invention is, overcomes the above-mentioned defect that prior art exists, and the analog circuit fault Forecasting Methodology that a kind of failure prediction precision is high is provided.First the method has the element of the trend of breaking down and occurring mode thereof (forward bias from nominal value or oppositely depart from nominal value) by BP neural network recognization circuit, obtain this element health degree by the cosine similarity of calculating different time points and put in time the data of variation, then put and whether break down sometime future predicting this element by the Method Using Relevance Vector Machine algorithm based on particle cluster algorithm optimization, or directly predict the time of failure point of this element.The method has popular style, and to resistance in mimic channel, the main elements such as electric capacity and inductance are all effective.
The technical solution adopted for the present invention to solve the technical problems is:
A kind of analog circuit fault Forecasting Methodology, comprises the following steps:
(1) the each element of mimic channel is carried out to Monte Carlo Analysis in non-fault interval, extract test node signal, the signal extracting is carried out to wavelet package transforms denoising Processing, extract each band signal energy, wherein test node signal is generally branch voltage;
(2) the band signal energy of extraction is normalized, obtains fault diagnosis proper vector;
(3) using fault diagnosis proper vector as training data, training BP neural network;
(4) while extracting work, circuit-under-test node signal, carries out wavelet package transforms and normalization, generates corresponding fault diagnosis proper vector, utilizes the judgement of BP neural network to have the failure mode of occurrence tendency;
(5) circuit-under-test test node frequency sweep response signal when extraction element is positioned at initial value, the failure prediction proper vector that forms initial value with this, described response signal is generally branch voltage;
(6) tested analog circuit test node frequency sweep response signal while pressing Fixed Time Interval extraction work, with the failure prediction proper vector of a time point sequence of this composed component;
(7) calculate the cosine angle distance between element fault predicted characteristics vector and the failure prediction proper vector of element initial value of pressing Fixed Time Interval extraction, in order to characterize the health degree of element in different time points, generate corresponding Healthy sequence;
(8) health degree threshold value when computing element breaks down;
(9), based on element Healthy sequence and time point sequence data, application particle cluster algorithm is in optimized selection the kernel function width factor of Method Using Relevance Vector Machine algorithm;
(10) the Method Using Relevance Vector Machine algorithm of application after particle cluster algorithm optimization carries out failure prediction to mimic channel.
In above-mentioned steps (1), non-fault interval is [the normal tolerance lower limit of 50%, 1+] of analog circuit element nominal value and [the normal tolerance upper limit of 1+, 150%] of nominal value.The each element of mimic channel is carried out Monte Carlo Analysis and can be adopted in non-fault interval the Monte Carlo function of Pspice software.
In above-mentioned steps (4), have the failure mode of occurrence tendency to refer to the element breaking down, and element depart from the direction of initial value in the time breaking down.
In above-mentioned steps (5), initial value is defined as: in the time that element forward bias breaks down from nominal value, initial value equals nominal value × (the normal tolerance upper limit of 1+), and in the time that element oppositely departs from nominal value and breaks down, initial value equals nominal value × (the normal tolerance lower limit of 1+).The frequency sweep response signal of extracting initial value can adopt the ac sweep function of Pspice software.
In above-mentioned steps (6), the fixed interval that extracts tested analog circuit test node frequency sweep response signal is followed successively by
t 1 , T 2 ..., T n , this time point sequence be [
t 1 , T 2 ..., T n ],
it is the sum of time point.
In above-mentioned step (7), the method for cosine angle distance is:
Wherein,
what characterize is the failure prediction proper vector of element while being positioned at initial value,
represent vector
in
jindividual feature,
,
mthe quantity of the feature in the stack features vector extracting;
what characterize is the
ithe element fault predicted characteristics vector that individual time point extracts,
represent vector
in
jindividual feature,
,
being the sum of the characteristic quantity of extraction, is also the sum of time point.
Health degree
, the Healthy sequence that the element fault predicted characteristics vector extracting by Fixed Time Interval generates be [
h 1 , H 2 ..., H n ], corresponding time point sequence be [
t 1 , T 2 ..., T n ].
In above-mentioned steps (8), when element breaks down, the computing method of health degree threshold value are: calculate the cosine angle distance of final deviation value in the non-fault interval of this element place offset direction, be the health degree threshold value that break down of element at this offset direction.When element breaks down, health degree threshold value also can be referred to as health degree fault threshold.
In above-mentioned steps (9), the step that application particle cluster algorithm is in optimized selection the kernel function width factor of Method Using Relevance Vector Machine algorithm is:
(1) initialization particle cluster algorithm parameter, comprises position, speed, Search Range and iterations, wherein width factor is mapped as to the position of particle;
(2) calculate the fitness of each particle, draw the personal best particle of each particle and the global optimum position of population according to fitness;
(3) each particle is carried out to the renewal of speed and position;
(4) repeat (2) and (3) until iteration finishes, Output rusults.
In above-mentioned steps (10), mimic channel is carried out to failure prediction and comprises two kinds of modes:
(1) the time of failure point of prediction element;
(2) whether prediction element will be put and will be broken down sometime future.
In the present invention, adopt the health degree of element in the simple and effective sign mimic channel of cosine angle distance.In the time that in circuit, main element progressively departs from its initial value arbitrarily, the cosine angle calculating is apart from declining, and the health degree that characterizes element declines, and the health degree that is equal to circuit declines.Adopt particle cluster algorithm to be in optimized selection the kernel function width factor of Method Using Relevance Vector Machine algorithm, can further improve the precision of failure prediction.This invention is simply effective, both can be for real-time system, and also can be for non real-time system; Both can predict the fault of In Linear Analog Circuits, also can predict the fault of non-linear simulation circuit; Can carry out failure prediction to main elements such as resistance, inductance and electric capacity in mimic channel.
The present invention compares background technology tool and has the following advantages:
(1) failure prediction proper vector is extracted simply, is applicable to real time environment;
(2) the failure prediction proper vector of application fetches can directly be calculated cosine angle distance, without pre-service, calculates simply, is equally applicable to real time environment;
(3) owing to being that extraction branch voltage is failure prediction proper vector, therefore the method can be applied to the failure prediction of non-linear simulation circuit equally;
(4) proposed to calculate cosine angle distance in order to characterize the method for the each main element health degree of mimic channel;
(5) Method Using Relevance Vector Machine algorithm is optimized after kernel function width factor at particle cluster algorithm, and its estimated performance will be significantly improved.
Accompanying drawing explanation
Fig. 1 is the FB(flow block) of analog circuit fault Forecasting Methodology of the present invention;
Fig. 2 is the FB(flow block) that particle cluster algorithm is optimized Method Using Relevance Vector Machine algorithm;
Fig. 3 is Sallen – Key band pass filter circuit schematic diagram;
Fig. 4 is the variation diagram that the health degree of R2 is put in time;
Fig. 5 is the prediction effect figure of Method Using Relevance Vector Machine algorithm optimizing through particle cluster algorithm that the present invention proposes.
Embodiment
Below in conjunction with drawings and Examples, the invention will be further described.
With reference to Fig. 1, overall flow figure of the present invention is made up of 8 steps:
Step 1: the each element of mimic channel is carried out to Monte Carlo Analysis in non-fault interval, extract test node signal, the signal extracting is carried out to wavelet package transforms denoising Processing, extract each band signal energy, wherein test node signal is generally branch voltage; The band signal energy of extraction is normalized, obtains fault diagnosis proper vector;
Step 2: using fault diagnosis proper vector as training data, training BP neural network;
Step 3: circuit-under-test node signal while extracting work, carry out wavelet package transforms and normalization, generate corresponding fault diagnosis proper vector, utilize the judgement of BP neural network to have the failure mode of occurrence tendency;
Step 4: the failure mode definite according to step 3, circuit-under-test test node frequency sweep response signal when extraction fault element is positioned at initial value, form the failure prediction proper vector of initial value with this, tested analog circuit test node frequency sweep response signal while pressing Fixed Time Interval extraction work, with the failure prediction proper vector of a time point sequence of this composed component;
Step 5: computing element is pressed the cosine angle distance between element fault predicted characteristics vector and the failure prediction proper vector of element initial value that Fixed Time Interval extracts, in order to characterize the health degree of element in different time points, and element Healthy sequence is corresponding with time point sequence;
Step 6: health degree threshold value when computing element breaks down;
Step 7: the Healthy sequence based on element and time point sequence data, application particle cluster algorithm is in optimized selection the kernel function width factor of Method Using Relevance Vector Machine algorithm;
Step 8: the Method Using Relevance Vector Machine algorithm after optimizing application carries out failure prediction to mimic channel.
These 8 steps are divided into two parts of serial, are respectively fault diagnosis of partial and failure prediction part, and fault diagnosis part comprises step 1, step 2 and step 3, and failure prediction part comprises step 4, step 5, step 6, step 7 and step 8.The object of fault diagnosis part is the failure mode that identification has occurrence tendency, and failure prediction part is according to the failure mode that has occurrence tendency of fault diagnosis part identification, set up targetedly failure prediction model, select corresponding health degree fault threshold, mimic channel is carried out to failure prediction.
In step 1, the signal extracting is carried out to wavelet package transforms denoising Processing, the method for extracting each band signal energy is:
(1) signal extracting is carried out to N layer Orthogonal Wavelet Packet and decompose, obtain high frequency wavelet bag coefficient of dissociation sequence and low frequency wavelet bag coefficient of dissociation sequence on each scaling function space;
(2) high frequency coefficient is carried out to denoising Processing;
(3) calculate the energy of each layer of WAVELET PACKET DECOMPOSITION coefficient sequence.
The each element of mimic channel is carried out Monte Carlo Analysis and can be adopted in non-fault interval the Monte Carlo function of Pspice software.
Non-fault interval is [50% of element nominal value, the normal tolerance lower limit of 1+] and [the normal tolerance upper limit of 1+ of nominal value, 150%], when device parameter values departs from nominal value more than 50%, this element can be judged to be fault element, and the normal tolerance of resistance is ± 5%, and the normal tolerance of electric capacity is ± 10%, non-fault is interval characterizes the duty that an analog circuit element has departed from normal value, but does not also reach the degree breaking down.The normal value of element is nominal value [the normal tolerance lower limit of 1+, the normal tolerance upper limit of 1+].
The object being normalized is to make to eliminate each input data dimension difference to calculating the impact causing.
In step 2, when training BP neural network, the input number of nodes of training BP neural network equals M(N+1), M is the nodes extracting, the number of plies that N is WAVELET PACKET DECOMPOSITION; Output node number equals typical number of faults.The activation function of hidden neuron adopts Sigmoid function.
In step 3, the output sequence of application BP neural network is
, wherein in sequence, q number is 1, shows that q class fault will occur circuit, in sequence, other numbers are 0, show that other class fault does not occur.Have the failure mode of occurrence tendency to refer to the element breaking down, and element departs from the direction of initial value in the time breaking down, forward bias is from initial value or oppositely depart from initial value.
In step 4, initial value is defined as: in the time that element forward bias breaks down from nominal value, initial value equals nominal value × (the normal tolerance upper limit of 1+), and in the time that element oppositely departs from nominal value and breaks down, initial value equals nominal value × (the normal tolerance lower limit of 1+).The frequency sweep response signal of extracting initial value can adopt the ac sweep function of Pspice software.
The fixed interval point that extracts tested analog circuit test node frequency sweep response signal is followed successively by
t 1 , T 2 ..., T n , this time point sequence be [
t 1 , T 2 ..., T n ]
it is the sum of time point.
In step 5, the method for calculating cosine angle distance is:
Wherein,
what characterize is the failure prediction proper vector of element while being positioned at initial value,
represent vector
in
jindividual feature,
,
mthe quantity of the feature in the stack features vector extracting;
what characterize is the
ithe element fault predicted characteristics vector that individual time point extracts,
represent vector
in
jindividual feature,
,
being the sum of the characteristic quantity of extraction, is also the sum of time point.
Application cosine angle is apart from characterizing element health degree, therefore health degree
, the Healthy sequence that the element fault predicted characteristics vector extracting by Fixed Time Interval generates be [
h 1 , H 2 ..., H n ], corresponding time point sequence be [
t 1 , T 2 ..., T n ].
In the time that element is positioned at initial value, the cosine angle distance that it calculates is 1, and the health degree of its sign is also 1, and in the time that the value of element progressively departs from initial value, its corresponding cosine angle distance progressively declines, and health degree declines.
In step 6, the cosine angle distance of the interval final deviation value of the non-fault of computing element place offset direction, is the health degree threshold value that break down of this element at this offset direction.When element breaks down, health degree threshold value also can be referred to as health degree fault threshold.The interval final deviation value of non-fault can be expressed as nominal value × 0.5 and nominal value × 1.5 on two offset directions.
In step 7, application particle cluster algorithm is that Method Using Relevance Vector Machine algorithm optimization is selected applicable kernel function width factor, and the step of particle cluster algorithm optimization Method Using Relevance Vector Machine algorithm can be with reference to Fig. 2, is divided into following 5 steps:
(1) initialization particle cluster algorithm parameter, comprises position, speed, Search Range and iterations etc., wherein width factor is mapped as to the position of particle;
(2) calculate the fitness of each particle according to fitness function, draw the personal best particle of each particle and the global optimum position of population according to fitness;
(3) each particle is carried out to the renewal of speed and position;
(4) repeat (2) and (3) until iteration cycle finishes, reach the maximum times of iteration;
(5) optimum results of particle cluster algorithm is output as to the kernel function width factor of Method Using Relevance Vector Machine algorithm.
The computing formula of particle cluster algorithm is:
Wherein,
it is the number of times of iteration;
,
the quantity of particle in population;
it is particle
the position in optimizing;
it is particle
speed in optimizing;
with
it is speedup factor;
with
it is the random number between 0-1;
it is inertia weight.
particle personal best particle in searching process,
it is the global optimum position of population.
The fitness function of selecting is:
Wherein
square error, for minimizing objective function, the tightness degree of its sign predicted data and True Data;
,
it is the data bulk for training;
with
respectively predicted data and True Data.
In step 8, the step of Method Using Relevance Vector Machine algorithm predicts is as follows:
The prediction of Method Using Relevance Vector Machine algorithm is output as:
Wherein
it is training sample;
be that average is 0, variance is
noise;
be output function, its expression formula is:
Wherein
for the weights of forecast model;
it is deviation;
be u training sample;
for kernel function, gaussian radial basis function kernel function one is to there being outstanding Nonlinear Processing ability, still adopt this function:
Wherein
for width factor, it has vital effect to the estimated performance of Method Using Relevance Vector Machine algorithm, is selected by particle cluster algorithm optimization.
Mimic channel is carried out to failure prediction and comprises two kinds of modes:
(1) the time of failure point of prediction element;
(2) whether prediction element will be put and will be broken down sometime future.
The Forecasting Methodology of mode (1) is:
Utilize the Method Using Relevance Vector Machine algorithm pair after optimizing
t n the element health degree of each time point is predicted continuously afterwards, finds satisfied in time point sequence
h n+g >=
h threshold and
h n+g-1
<H
thresholdcondition
t n+g time point, mimic channel will be in future
t n+g time point breaks down.Wherein
h threshold it is the health degree threshold value of mimic channel corresponding failure classification while breaking down.
The Forecasting Methodology of mode (2) is:
Suppose that prediction element exists
t n after time point the
kindividual time point
t n+k whether break down.Utilize the Method Using Relevance Vector Machine algorithm predicts after optimizing
t n+k time point health degree, is assumed to be
h n+k .
If 1.
h n+k >H threshold , mimic channel is following the
kindividual time point does not break down;
If 2.
h n+k =
h threshold , mimic channel is following the
kindividual time point breaks down;
If 3.
h n+k <
h threshold , mimic channel will be in future
kto break down in individual time point, judge the Forecasting Methodology that the concrete time point that breaks down can reference pattern (1).
For showing the estimated performance of Method Using Relevance Vector Machine algorithm after particle cluster algorithm is optimized, at this with an example explanation.Fig. 3 is Sallen – Key band pass filter circuit schematic diagram, wherein V
outfor test node, each element nominal value is identified in figure, the time of failure point of prediction R2 as an example of R2 example.If R2 passing resistance value in time in ageing process progressively rises, each regular time the changing value of interval resistance identical.R2 at the initial value in its non-fault interval is: 3 × (1+0.05)=3.15k Ω, resistance value when R2 breaks down is: 3 × 1.5=4.5 k Ω, 5 Ω if each time point rises, after 270 time points, the parameter value of R2 reaches fault value 4.5 k Ω by initial value 3.15k Ω.Check for convenience prediction effect, the present invention arranges R2, and to rise to resistance value from initial value 3.15k Ω be 4.75 k Ω always, totally 320 time points.
Failure prediction proper vector when extraction R2 is positioned at initial value, and the failure prediction proper vector of 320 time points, calculate the health degree of these 320 time points, it is showed in to Fig. 4 by increasing progressively of time point, wherein the 270th time point resistance value will reach 4.5k Ω, R2 will break down, and its health degree fault threshold is 0.999765181360359.The data that particle cluster algorithm is applied front 100 time points are training data, optimize the kernel functional parameter factor of selecting Method Using Relevance Vector Machine algorithm, the value obtaining after optimization is 0.443725069454054, Method Using Relevance Vector Machine algorithm after optimizing with this is predicted the fault of mimic channel, the Contrast on effect of predicted data and raw data is shown in Fig. 5, prediction down time is the 268th time point, error is-2 time points, and the Mse between predicted data and the raw data of each time point is 5.3340e-11.The kernel functional parameter factor obtaining due to optimization approaches 0.45, therefore at this, respectively with 0.1,0.2,0.3,0.4,0.45,0.5,0.6,0.7,0.8,0.9 and 1 is the kernel functional parameter factor, time of failure is predicted, prediction the results are shown in Table 1.Wherein >320 shows to have exceeded the maximum magnitude of data, and >50 representative has exceeded the maximum magnitude of error.
Table 1 parameter factors is 0.1,0.2,0.3,0.4,0.45,0.5,0.6,0.7,0.8,0.9 and 1 o'clock prediction result
| Parameter factors | Prediction down time point | Error | Mse |
| 0.1 | 143 | -127 | 1.1876e-08 |
| 0.2 | 186 | -84 | 4.8176e-09 |
| 0.3 | 207 | -63 | 2.6312e-09 |
| 0.4 | 245 | -25 | 3.9886e-10 |
| 0.45 | 213 | -57 | 2.4730e-09 |
| 0.5 | 228 | -42 | 1.4181e-09 |
| 0.6 | 258 | -12 | 3.4975e-10 |
| 0.7 | 286 | 16 | 1.7554e-10 |
| 0.8 | 307 | 37 | 3.9992e-10 |
| 0.9 | >320 | >50 | 7.5851 |
| 1 | >320 | >50 | 1.1708e-09 |
Can learn from every result of table 1, optimize the Method Using Relevance Vector Machine algorithm of parameter factors through particle cluster algorithm, its prediction down time point is the most accurate, and predicated error minimum, only has-2 time points, and Mse is also minimum.And Method Using Relevance Vector Machine algorithm is 0.1,0.2,0.3 in the parameters factor, 0.4,0.45,0.5,0.6,0.7,0.8,0.9 and 1 o'clock, what the error of prediction and Mse were all greater than that the present invention proposes has optimized the Method Using Relevance Vector Machine algorithm of parameter factors through particle cluster algorithm.This explanation Method Using Relevance Vector Machine algorithm is optimized after kernel function width factor at particle cluster algorithm, and its estimated performance will be significantly improved.
Because the method Characteristic Extraction is simple, calculated amount is little, thus not only can be for real-time system, also can be for non real-time system.
Owing to also there is equally parameter value variation in electric capacity and inductance in ageing process, therefore the method also can be applied to electric capacity and inductance.
Because non-linear simulation circuit exists equally the feature of parameter value variation in component ageing process, non-linear simulation circuit also can extract the frequency sweep response signal of test node simultaneously, therefore the method both can be predicted the fault of In Linear Analog Circuits, also can predict the fault of non-linear simulation circuit.
Claims (9)
1. an analog circuit fault Forecasting Methodology, is characterized in that, comprises the following steps:
(1) the each element of mimic channel is carried out to Monte Carlo Analysis in non-fault interval, extract test node signal, the signal extracting is carried out to wavelet package transforms denoising Processing, extract each band signal energy, wherein test node signal is generally branch voltage;
(2) the band signal energy of extraction is normalized, obtains fault diagnosis proper vector;
(3) using fault diagnosis proper vector as training data, training BP neural network;
(4) while extracting work, circuit-under-test node signal, carries out wavelet package transforms and normalization, generates corresponding fault diagnosis proper vector, utilizes the judgement of BP neural network to have the failure mode of occurrence tendency;
(5) circuit-under-test test node frequency sweep response signal when extraction element is positioned at initial value, the failure prediction proper vector that forms initial value with this, described response signal is generally branch voltage;
(6) tested analog circuit test node frequency sweep response signal while pressing Fixed Time Interval extraction work, with the failure prediction proper vector of a time point sequence of this composed component;
(7) calculate the cosine angle distance between element fault predicted characteristics vector and the failure prediction proper vector of element initial value of pressing Fixed Time Interval extraction, in order to characterize the health degree of element in different time points, generate corresponding Healthy sequence;
(8) health degree threshold value when computing element breaks down;
(9), based on element Healthy sequence and time point sequence data, application particle cluster algorithm is in optimized selection the kernel function width factor of Method Using Relevance Vector Machine algorithm;
(10) the Method Using Relevance Vector Machine algorithm of application after particle cluster algorithm optimization carries out failure prediction to mimic channel.
2. analog circuit fault Forecasting Methodology according to claim 1, is characterized in that, in described step (1), non-fault interval is [the normal tolerance lower limit of 50%, 1+] of analog circuit element nominal value and [the normal tolerance upper limit of 1+, 150%] of nominal value.
3. analog circuit fault Forecasting Methodology according to claim 1, is characterized in that, in described step (4), have the failure mode of occurrence tendency to refer to the element breaking down, and element departs from the direction of initial value in the time breaking down.
4. analog circuit fault Forecasting Methodology according to claim 1, it is characterized in that, in described step (5), initial value is defined as: in the time that element forward bias breaks down from nominal value, initial value equals nominal value × (the normal tolerance upper limit of 1+), in the time that element oppositely departs from nominal value and breaks down, initial value equals nominal value × (the normal tolerance lower limit of 1+).
5. analog circuit fault Forecasting Methodology according to claim 1, is characterized in that, in described step (6), the fixed interval point that extracts tested analog circuit test node frequency sweep response signal is followed successively by
t 1 , T 2 ..., T n , thus form time point sequence be [
t 1 , T 2 ..., T n ],
it is the sum of time point.
6. analog circuit fault Forecasting Methodology according to claim 1, is characterized in that, in described step (7), the method for calculating cosine angle distance is:
Wherein,
what characterize is the failure prediction proper vector of element while being positioned at initial value,
represent vector
in
jindividual feature,
,
mthe quantity of the feature in the stack features vector extracting;
what characterize is the
ithe element fault predicted characteristics vector that individual time point extracts,
represent vector
in
jindividual feature,
,
being the sum of the characteristic quantity of extraction, is also the sum of time point;
Health degree
, the Healthy sequence that the element fault predicted characteristics vector extracting by Fixed Time Interval generates be [
h 1 , H 2 ..., H n ], corresponding time point sequence be [
t 1 , T 2 ..., T n ].
7. analog circuit fault Forecasting Methodology according to claim 1, it is characterized in that, in described step (8), when element breaks down, the computing method of health degree threshold value are: calculate the cosine angle distance of final deviation value in the non-fault interval of this element place offset direction, be the health degree threshold value that break down of element at this offset direction; When element breaks down, health degree threshold value also can be referred to as health degree fault threshold.
8. analog circuit fault Forecasting Methodology according to claim 1, is characterized in that, in described step (9), the step that application particle cluster algorithm is in optimized selection the kernel function width factor of Method Using Relevance Vector Machine algorithm is:
(1) initialization particle cluster algorithm parameter, comprises position, speed, Search Range and iterations, wherein width factor is mapped as to the position of particle;
(2) calculate the fitness of each particle, draw the personal best particle of each particle and the global optimum position of population according to fitness;
(3) each particle is carried out to the renewal of speed and position;
(4) repeat (2) and (3) until iteration finishes, Output rusults.
9. analog circuit fault Forecasting Methodology according to claim 1, is characterized in that, in described step (10),
Mimic channel is carried out to failure prediction and comprises two kinds of modes:
(1) the time of failure point of prediction element;
(2) whether prediction element will be put and will be broken down sometime future.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410088347.6A CN103824135B (en) | 2014-03-11 | A kind of analog circuit fault Forecasting Methodology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410088347.6A CN103824135B (en) | 2014-03-11 | A kind of analog circuit fault Forecasting Methodology |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103824135A true CN103824135A (en) | 2014-05-28 |
| CN103824135B CN103824135B (en) | 2016-11-30 |
Family
ID=
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104198924A (en) * | 2014-09-11 | 2014-12-10 | 合肥工业大学 | Novel analog circuit early fault diagnosis method |
| CN105044590A (en) * | 2015-08-07 | 2015-11-11 | 合肥工业大学 | Single measurement node simulation circuit fault diagnosis method |
| CN105467305A (en) * | 2016-01-13 | 2016-04-06 | 安徽理工大学 | Buck-Boost circuit fault feature extraction method under variable conditions |
| CN105467304A (en) * | 2015-12-26 | 2016-04-06 | 哈尔滨理工大学 | Fault diagnosis instrument of nonlinear analog circuit based on nuclear characteristics and fault diagnosis method |
| CN105699883A (en) * | 2016-01-22 | 2016-06-22 | 合肥工业大学 | Analog circuit health prediction method |
| CN106347578A (en) * | 2016-10-27 | 2017-01-25 | 江苏科技大学 | Method for monitoring marine propulsion shafting bearing vibration failure |
| CN104238546B (en) * | 2014-08-28 | 2017-09-01 | 中国南方电网有限责任公司超高压输电公司检修试验中心 | Based on the direct current master-failure diagnostic method for mixing neutral net |
| CN107894564A (en) * | 2017-11-09 | 2018-04-10 | 合肥工业大学 | A kind of analog-circuit fault diagnosis method based on intersection wavelet character |
| CN108414923A (en) * | 2018-02-05 | 2018-08-17 | 武汉大学 | A kind of analog-circuit fault diagnosis method based on the extraction of depth confidence network characterization |
| CN108490344A (en) * | 2018-03-28 | 2018-09-04 | 电子科技大学 | Analog-circuit fault diagnosis method based on training sample optimization |
| TWI639907B (en) * | 2017-06-06 | 2018-11-01 | 國立彰化師範大學 | Tool machine residual service life prediction system and method thereof |
| CN108898249A (en) * | 2018-06-28 | 2018-11-27 | 鹿寨知航科技信息服务有限公司 | A kind of electric network fault prediction technique |
| CN109447152A (en) * | 2018-10-29 | 2019-03-08 | 中国石油大学(华东) | A kind of failure prediction method based on Monte Carlo tree search and neural network |
| CN109828860A (en) * | 2019-02-01 | 2019-05-31 | 重庆科技学院 | A kind of method for diagnosing faults based on GSA optimization SVM |
| CN110058133A (en) * | 2019-04-15 | 2019-07-26 | 杭州拓深科技有限公司 | A kind of electrical circuit fault electric arc wrong report optimization method based on feedback mechanism |
| CN111324083A (en) * | 2020-03-02 | 2020-06-23 | 西南交通大学 | Real-time monitoring and early warning method and system for key components of mechanical equipment |
| CN111337819A (en) * | 2020-03-02 | 2020-06-26 | 武汉大学 | Analog circuit fault feature extraction method and system based on optimized wavelet basis function |
| CN112485650A (en) * | 2020-11-30 | 2021-03-12 | 电子科技大学 | Analog circuit fault parameter range identification method based on PBI |
| CN113341305A (en) * | 2021-05-12 | 2021-09-03 | 西安建筑科技大学 | Analog circuit fault prediction method based on fusion modeling |
| CN113985261A (en) * | 2021-07-07 | 2022-01-28 | 四川大学锦城学院 | Analog circuit fault diagnosis method based on neural network |
Cited By (28)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104238546B (en) * | 2014-08-28 | 2017-09-01 | 中国南方电网有限责任公司超高压输电公司检修试验中心 | Based on the direct current master-failure diagnostic method for mixing neutral net |
| CN104198924A (en) * | 2014-09-11 | 2014-12-10 | 合肥工业大学 | Novel analog circuit early fault diagnosis method |
| CN105044590B (en) * | 2015-08-07 | 2017-08-25 | 合肥工业大学 | A kind of single measuring node analog-circuit fault diagnosis method |
| CN105044590A (en) * | 2015-08-07 | 2015-11-11 | 合肥工业大学 | Single measurement node simulation circuit fault diagnosis method |
| CN105467304A (en) * | 2015-12-26 | 2016-04-06 | 哈尔滨理工大学 | Fault diagnosis instrument of nonlinear analog circuit based on nuclear characteristics and fault diagnosis method |
| CN105467305B (en) * | 2016-01-13 | 2018-06-26 | 安徽理工大学 | Buck-Boost faults feature extracting method under a kind of variable working condition |
| CN105467305A (en) * | 2016-01-13 | 2016-04-06 | 安徽理工大学 | Buck-Boost circuit fault feature extraction method under variable conditions |
| CN105699883A (en) * | 2016-01-22 | 2016-06-22 | 合肥工业大学 | Analog circuit health prediction method |
| CN105699883B (en) * | 2016-01-22 | 2018-08-14 | 合肥工业大学 | A kind of analog circuit health forecast method |
| CN106347578A (en) * | 2016-10-27 | 2017-01-25 | 江苏科技大学 | Method for monitoring marine propulsion shafting bearing vibration failure |
| TWI639907B (en) * | 2017-06-06 | 2018-11-01 | 國立彰化師範大學 | Tool machine residual service life prediction system and method thereof |
| CN107894564A (en) * | 2017-11-09 | 2018-04-10 | 合肥工业大学 | A kind of analog-circuit fault diagnosis method based on intersection wavelet character |
| CN108414923A (en) * | 2018-02-05 | 2018-08-17 | 武汉大学 | A kind of analog-circuit fault diagnosis method based on the extraction of depth confidence network characterization |
| CN108490344B (en) * | 2018-03-28 | 2020-05-22 | 电子科技大学 | Analog circuit fault diagnosis method based on training sample optimization |
| CN108490344A (en) * | 2018-03-28 | 2018-09-04 | 电子科技大学 | Analog-circuit fault diagnosis method based on training sample optimization |
| CN108898249A (en) * | 2018-06-28 | 2018-11-27 | 鹿寨知航科技信息服务有限公司 | A kind of electric network fault prediction technique |
| CN109447152A (en) * | 2018-10-29 | 2019-03-08 | 中国石油大学(华东) | A kind of failure prediction method based on Monte Carlo tree search and neural network |
| CN109447152B (en) * | 2018-10-29 | 2022-02-15 | 中国石油大学(华东) | Fault prediction method based on Monte Carlo tree search and neural network |
| CN109828860A (en) * | 2019-02-01 | 2019-05-31 | 重庆科技学院 | A kind of method for diagnosing faults based on GSA optimization SVM |
| CN110058133A (en) * | 2019-04-15 | 2019-07-26 | 杭州拓深科技有限公司 | A kind of electrical circuit fault electric arc wrong report optimization method based on feedback mechanism |
| CN110058133B (en) * | 2019-04-15 | 2021-03-02 | 杭州拓深科技有限公司 | Feedback mechanism-based electric circuit fault arc false alarm optimization method |
| CN111324083A (en) * | 2020-03-02 | 2020-06-23 | 西南交通大学 | Real-time monitoring and early warning method and system for key components of mechanical equipment |
| CN111337819A (en) * | 2020-03-02 | 2020-06-26 | 武汉大学 | Analog circuit fault feature extraction method and system based on optimized wavelet basis function |
| CN111337819B (en) * | 2020-03-02 | 2021-02-19 | 武汉大学 | Analog circuit fault feature extraction method and system based on optimized wavelet basis function |
| CN112485650A (en) * | 2020-11-30 | 2021-03-12 | 电子科技大学 | Analog circuit fault parameter range identification method based on PBI |
| CN112485650B (en) * | 2020-11-30 | 2021-09-14 | 电子科技大学 | Analog circuit fault parameter range identification method based on PBI |
| CN113341305A (en) * | 2021-05-12 | 2021-09-03 | 西安建筑科技大学 | Analog circuit fault prediction method based on fusion modeling |
| CN113985261A (en) * | 2021-07-07 | 2022-01-28 | 四川大学锦城学院 | Analog circuit fault diagnosis method based on neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10408882B2 (en) | Method for predicting remaining useful life of lithium battery based on wavelet denoising and relevance vector machine | |
| CN105279365B (en) | For the method for the sample for learning abnormality detection | |
| Li et al. | Efficient missing data imputing for traffic flow by considering temporal and spatial dependence | |
| CN102721941B (en) | Method for fusing and diagnosing fault information of circuit of electric meter on basis of SOM (self-organized mapping) and D-S (Dempster-Shafer) theories | |
| CN103995237A (en) | Satellite power supply system online fault diagnosis method | |
| Arif et al. | Cascade quality prediction method using multiple PCA+ ID3 for multi-stage manufacturing system | |
| CN103064009B (en) | Artificial circuit fault diagnosis method based on wavelet analysis and limited gauss mixed model expectation maximization (EM) method | |
| CN105005197B (en) | Time-varying neurodynamics system identifying method based on Chebyshev polynomials expansion | |
| Wang et al. | Hankel-structured tensor robust PCA for multivariate traffic time series anomaly detection | |
| CN115081673A (en) | Oil and gas pipeline abnormity prediction method and device, electronic equipment and medium | |
| CN113554128A (en) | Unconventional anomaly detection method and system and storage medium | |
| Bodda et al. | Comparative Analysis of Deep Learning and Machine Learning Techniques for Power System Fault type Classification and Location Prediction | |
| CN103824135A (en) | Analogue circuit failure prediction method | |
| CN104091047A (en) | Traffic flow missing data estimation system and method based on traffic time-spatial information | |
| CN111262248A (en) | Random power flow analysis and calculation method and system | |
| Bechtold et al. | Nonlinear model order reduction in nanoelectronics: combination of POD and TPWL | |
| CN103824135B (en) | A kind of analog circuit fault Forecasting Methodology | |
| CN114239948A (en) | Deep traffic flow prediction method, medium and equipment based on time sequence decomposition unit | |
| CN113240904A (en) | Traffic flow prediction method based on feature fusion | |
| Heydarzadeh et al. | Model-free testing of analog circuits | |
| Inazumi et al. | Use of prior knowledge in a non-Gaussian method for learning linear structural equation models | |
| EP3850495A1 (en) | Computer-implemented method and electronic device for detecting influential components in a netlist representing an electrical circuit | |
| Li et al. | Data-driven localization of forced oscillations in power systems | |
| Gamage et al. | Detecting and adapting to concept drift in continually evolving stochastic processes | |
| Zhao et al. | Statistical rare event analysis using smart sampling and parameter guidance |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |