CN101329697A - Method for predicting analog circuit state based on immingle algorithm - Google Patents
Method for predicting analog circuit state based on immingle algorithm Download PDFInfo
- Publication number
- CN101329697A CN101329697A CNA2008100446712A CN200810044671A CN101329697A CN 101329697 A CN101329697 A CN 101329697A CN A2008100446712 A CNA2008100446712 A CN A2008100446712A CN 200810044671 A CN200810044671 A CN 200810044671A CN 101329697 A CN101329697 A CN 101329697A
- Authority
- CN
- China
- Prior art keywords
- analog circuit
- model
- circuit state
- predicted value
- predicting
- 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
Abstract
The invention discloses a forecasting method for the status of an analog circuit, which is based on a hybrid algorithm and comprises the following steps: (1) a test signal with multi-frequency is selected for a particular analog circuit, and the spectrum characteristic of the test signal is picked up at intervals, and mapped to be an one-dimensional spectrum characteristic through principal component analysis; (2) wavelet technology is adopted to de-noise; (3) the one-dimensional spectrum characteristic is used as original data, an ARMA-GM(1,1) combined model is adopted to predict the trend term, an LSSVM model is adopted to predict the nonlinear term, and the predictive value of the analog circuit status is obtained by the superposition of the predictive value of the trend term and the predictive value of the nonlinear term. The method is realized by selecting the test signal with multi-frequency to extract the signal representing for the spectrum characteristic of a failure status, carrying out de-dimension and de-noise so as to obtain the original data, and finally obtaining the prediction of the analog circuit status by the hybrid algorithm. Tested by experiments, the method has very high precision in prediction of the analog circuit status and can accurately predict the analog circuit status.
Description
Technical field
The present invention relates to the electric signal process field, specifically, relate to a kind of method for predicting analog circuit state based on immingle algorithm.
Background technology
Mimic channel is the main composition module of sophisticated electronic system, the quality of its state directly influences the normal operation of electronic system or equipment, often cause irremediable loss if break down, carry out effective status monitoring and failure prediction, in time find initial failure, as yet not before the complete failure, in time change fault element in electronic system, fault is prevented trouble before it happens, can bring huge economic benefit.Therefore, mimic channel is carried out effective status monitoring and prediction has important effect.
The reason that the sophisticated electronic system breaks down is diversified, the information that characterizes malfunction has a lot of signals, these signals possess numerous characteristics, as linear or non-linear, stationarity or non-stationary even possess chaotic characteristic, may be one or possess a plurality of characteristics simultaneously to have very strong uncertainty and randomness.In addition, because the structural complexity of electronic equipment and the wearability of element, the appearance that gradual fault occurs is a progressive process, failure symptom is fainter, the influence of component tolerance is especially arranged based on the equipment of mimic channel, and the signal of the sign fault status information that obtains has stronger ground unrest.Therefore, need reasonably select, handle the signal that characterizes fault status information, and on this basis, analog circuit state be predicted.
Summary of the invention
The objective of the invention is to provides a kind of method for predicting analog circuit state based on immingle algorithm in that the signal that characterizes fault status information is reasonably selected, handled on the basis.
For achieving the above object, a kind of method for predicting analog circuit state based on immingle algorithm of the present invention may further comprise the steps:
(1), at the physical simulation circuit, select the test signal input of a plurality of frequencies, and at test point interval certain hour, extract the spectrum signature of test point signal, acquisition comprises the higher-dimension spectrum signature of failure message, take the pivot componential analysis, the higher-dimension spectrum signature is mapped as the one-dimensional spectrum feature;
(2), with wavelet technique the one-dimensional spectrum feature is carried out denoising;
(3), with the one-dimensional spectrum feature as raw data, adopt ARMA-GM (1,1) built-up pattern is predicted the trend term of reflection fault gradual change, adopt the least square method supporting vector machine model that the nonlinear terms of reflection fault gradual change are predicted, trend term predicted value and nonlinear terms predicted value are superposeed, obtain the analog circuit state predicted value.
The present invention is by selecting the test signal of a plurality of frequencies, extract the spectrum signature of the signal that characterizes fault status information in test point, carry out dimension-reduction treatment, and denoising, obtained to characterize the raw data of malfunction, immingle algorithm by the present invention design at last, promptly adopt ARMA-GM (1,1) model is predicted the trend term of reflection fault gradual change, adopt the least square method supporting vector machine model that the nonlinear terms of reflection fault gradual change are predicted, trend term prediction and nonlinear terms prediction are superposeed, obtained the analog circuit state prediction.Checking by experiment, method of the present invention is very high to the analog circuit state accuracy of predicting, can predict analog circuit state exactly.
Description of drawings
Fig. 1 is a kind of embodiment theory diagram of the present invention;
Fig. 2 is physical simulation circuit theory diagrams of checking Forecasting Methodology of the present invention;
Fig. 3 is the variation diagram of the amplitude-frequency characteristic of test frequency correspondence along with parameter;
Fig. 4 is that variation PCA shown in Figure 3 analyzes the one dimension amplitude-frequency characteristic figure that obtains;
Fig. 5 is the amplitude-frequency characteristic figure after the amplitude-frequency characteristic denoising shown in Figure 4;
Fig. 6 is the signatures to predict comparison diagram.
Embodiment
Below in conjunction with accompanying drawing, the preferred embodiments of the present invention is described.In the following description, when perhaps the detailed description of known function that adopts and design can desalinate subject content of the present invention, these were described in here and will be left in the basket.
Fig. 1 is a kind of embodiment theory diagram of the present invention.In the present embodiment, the method for predicting analog circuit state based on immingle algorithm of the present invention is specially:
Step ST101:, select the test signal input of a plurality of frequencies at the physical simulation circuit;
Step ST102: at test point interval certain hour, extract the spectrum signature of test point signal, obtain to comprise the higher-dimension spectrum signature of failure message
Step ST103: take the pivot componential analysis, carry out dimension-reduction treatment, the higher-dimension spectrum signature is mapped as the one-dimensional spectrum feature;
Step ST2: the one-dimensional spectrum feature is carried out denoising with wavelet technique;
Step ST301: the one-dimensional spectrum feature as raw data, is asked for the model prediction exponent number;
Step ST302: adopt ARMA-GM (1,1) built-up pattern that the trend term of reflection fault gradual change is predicted;
Step ST303: adopt the least square method supporting vector machine model, promptly the LSSVM model is predicted the nonlinear terms of reflection fault gradual change,
Step ST304: trend term predicted value and nonlinear terms predicted value are superposeed, obtain the analog circuit state predicted value.
Below immingle algorithm of the present invention is further described:
1, the model prediction exponent number determines
In the present embodiment, step ST301 determines ARMA-GM (1,1) the prediction order p of built-up pattern and LSSVM model, specifically, be to ask autocorrelation matrix with raw data x (n), choose appropriate threshold by svd and check the order of determining this correlation matrix, determine the prediction order p of model by the model testing method.
2, ARMA-GM (1,1) built-up pattern is predicted trend term
2.1, the foundation of arma modeling
The raw data { x (n) } that obtains has contained the information of mimic channel running status, in step ST302, utilizes model prediction exponent number p and raw data { x (n) }, at first sets up arma modeling, and the mimic channel running status meets:
Wherein e (n) is a discrete white noise, a
iAnd b
jBe respectively autoregression (AR) parameter and running mean (MA) parameter.
The AR model can approach arma modeling to a certain extent, and in the present embodiment, the AR model comes match, and above-mentioned formula (1) is reduced to:
In above-mentioned formula (2), as long as determine model prediction exponent number p and auto-regressive parameter a
iJust can obtain the predicted value of AR model.Model prediction exponent number p determines in step ST301, only needs to determine auto-regressive parameter a
iGet final product.In the present embodiment, for making predicated error power minimum, adopt the Burg algorithm to determine auto-regressive parameter a
i
2.2, the foundation of GM (1,1) model
Gray model is a kind of better dynamic system modeling method, and by the principle that raw data adds up, the randomness of reduction data message demonstrates the rule that contains in the data.
In the present embodiment, according to raw data { x (n) }, i.e. x (i), i=1 ..., n, definition:
Single order gray model equation is:
If
Be argument sequence:
Y=[x(2)?x(3)…x(n)]
Obtain estimates of parameters by least square method:
Can get parameter a, b.Substitution GM (1,1) model:
Can get the predicted value of GM (1,1) model.
Arbitrarily gray model As time goes on, principle timeorigin, the information of legacy data will progressively reduce, and must consider in time to insert new data in the model and remove legacy data, the GM of foundation (1,1) model could reflect the feature that information is current at any time.In the present embodiment, the method for building up of GM (1,1) model is improved, make that (a is that online adaptive changes b) to model parameter.Be specially: in raw data { x (n) }, we can set up GM (1,1) model prediction output data, when raw data increases to { x (n+1) }, we select x (2) ..., x (n+1} modeling and forecasting output data, and the like, every like this prediction one secondary data will be set up new model, promptly recomputate model parameter (a, b), make that model parameter is online variable, has certain adaptivity.
2.3, combination ARMA-GM (1,1) forecast model
Any Forecasting Methodology all comprises systematic independent information.Arma modeling can dope the random character of signal preferably, and setting up model often needs more data sample, and is very high to the stationarity requirement of signal; GM (1,1) model has good predictive ability to the development trend that contains in signal as periodicity, overall change direction, needs sample data less, is applicable to the data that variation is bigger, but the randomness predicated error to data is bigger, and need data message be on the occasion of.In order effectively to utilize the advantage of two kinds of Forecasting Methodologies, in the present embodiment, when two kinds of forecast models are combined, be weighted combination, can increase the estimated performance of system like this.Specific as follows:
Wherein,
Be the predicted value of ARMA-GM (1,1) built-up pattern to raw data { x (n) },
Be the predicted value that adopts arma modeling to obtain,
Be the predicted value that adopts GM (1,1) model to obtain, 0<w<1st, the combination weights, its selection directly affects the precision of combined prediction, is decided by the precision of forecast model separately.If arma modeling squared prediction error and more little, the expression precision of prediction is high more, and then the w value is just big more, means that the shared weights of arma modeling predicted value just should be big more in built-up pattern.Around this principle, arma modeling squared prediction error and y
1, GM (1,1) model prediction error sum of squares y
2, combination weights w=y
2/ (y
1+ y
2).In the present embodiment, this ARMA-GM (1,1) built-up pattern combines the advantage of two kinds of models well, and the trend term that reflects the fault gradual change is had prediction comparatively accurately.
3, the LSSVM model is predicted nonlinear terms
In the present embodiment, adopt the population intelligent algorithm that nuclear parameter σ in the LSSVM model and penalty coefficient γ are optimized.
4, trend term predicted value and nonlinear terms predicted value
Trend term predicted value and nonlinear terms predicted value are superposeed, obtain the analog circuit state predicted value, promptly realized prediction analog circuit state.
Case verification
Fig. 2 is physical simulation circuit theory diagrams of checking Forecasting Methodology of the present invention.Among the figure, this mimic channel comprises that three operational amplifiers and resistance R 1~5, resistance R _ f, capacitor C 1~2 form.Test point 1~3 is arranged, test point 3 output test point signals, selecting resistance R 4 according to the characteristic of this circuit is fault element, and it is 0~45% that the slow varying parameter variation range is set, and fault type is that parameter increases gradually.
Select 7 frequencies, the test signal of 14Khz input promptly 8,9,10,11,12,13,, the spectrum signature that the change of parameter obtains is an amplitude-frequency characteristic, this amplitude-frequency characteristic is equivalent to the spectrum signature of interval of the present invention certain hour sampling, just seem more more directly perceived like this, as can be seen fault actual spectrum feature that changes and the relation of predicting spectrum signature.The amplitude-frequency characteristic of each test frequency correspondence of this circuit is along with the variation of parameter is also changing gradually, as shown in Figure 3.
Step 1: by observing the variation of the pairing amplitude-frequency characteristic of each test frequency, original as can be known 7 dimension amplitude-frequency characteristics have higher-dimension and redundancy, adopt the pivot constituent analysis, and promptly PCA analyzes and obtains the one dimension amplitude-frequency characteristic, as shown in Figure 4;
Step 2: because the amplitude-frequency characteristic that the random variation of normal component in allowing range of tolerable variance causes gathering has noise, this also is embodied in the one dimension amplitude-frequency characteristic after the processing simultaneously, adopt wavelet technique that it is carried out denoising, the amplitude-frequency characteristic after the denoising as shown in Figure 5;
Step 3: the amplitude-frequency characteristic after the denoising is predicted with method of the present invention, obtains predicted value and measured value amplitude-frequency characteristic, as shown in Figure 6.
In Fig. 6, we can find clearly that the method for predicting analog circuit state based on immingle algorithm of the present invention obtains predicted value and measured value is basic coincideing.When the analog circuit state predicted value exceeds fence coverage, just can accurately report to the police in advance like this, avoid the generation of losing.
Although above the illustrative embodiment of the present invention is described; but should be understood that; the invention is not restricted to the scope of embodiment; to those skilled in the art; as long as various variations appended claim limit and the spirit and scope of the present invention determined in; these variations are conspicuous, and all utilize innovation and creation that the present invention conceives all at the row of protection.
Claims (5)
1, a kind of method for predicting analog circuit state based on immingle algorithm may further comprise the steps:
(1), at the physical simulation circuit, select the test signal input of a plurality of frequencies, and at test point interval certain hour, extract the spectrum signature of test point signal, acquisition comprises the higher-dimension spectrum signature of failure message, take the pivot componential analysis, the higher-dimension spectrum signature is mapped as the one-dimensional spectrum feature;
(2), with wavelet technique the one-dimensional spectrum feature is carried out denoising;
(3), with the one-dimensional spectrum feature as raw data, adopt ARMA-GM (1,1) built-up pattern is predicted the trend term of reflection fault gradual change, adopt the least square method supporting vector machine model that the nonlinear terms of reflection fault gradual change are predicted, trend term predicted value and nonlinear terms predicted value are superposeed, obtain the analog circuit state predicted value.
2, the method for predicting analog circuit state based on immingle algorithm according to claim 1 is characterized in that, described GM (1,1) foundation of model, in raw data { x (n) }, set up GM (1,1) model prediction output data, when raw data increases to { x (n+1) }, we select x (2) ..., x (n+1} modeling and forecasting output data, and the like, every prediction one secondary data will be set up new model.
3, the method for predicting analog circuit state based on immingle algorithm according to claim 1 is characterized in that, described ARMA-GM (1,1) built-up pattern makes up as follows:
4, the method for predicting analog circuit state based on immingle algorithm according to claim 3 is characterized in that, described combination weights:
w=y
2/(y
1+y
2)
Wherein, y
1Be the arma modeling squared prediction error and, y
2It is GM (1,1) model prediction error sum of squares.
5, the method for predicting analog circuit state based on immingle algorithm according to claim 1, it is characterized in that described least square method supporting vector machine model adopts the population intelligent algorithm that nuclear parameter σ in the LSSVM model and penalty coefficient γ are optimized.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2008100446712A CN101329697B (en) | 2008-06-11 | 2008-06-11 | Method for predicting analog circuit state based on immingle algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2008100446712A CN101329697B (en) | 2008-06-11 | 2008-06-11 | Method for predicting analog circuit state based on immingle algorithm |
Publications (2)
Publication Number | Publication Date |
---|---|
CN101329697A true CN101329697A (en) | 2008-12-24 |
CN101329697B CN101329697B (en) | 2011-03-23 |
Family
ID=40205504
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2008100446712A Expired - Fee Related CN101329697B (en) | 2008-06-11 | 2008-06-11 | Method for predicting analog circuit state based on immingle algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101329697B (en) |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102222151A (en) * | 2011-07-21 | 2011-10-19 | 电子科技大学 | Analog circuit fault prediction method based on ARMA (Autoregressive Moving Average) |
CN102499651A (en) * | 2011-10-24 | 2012-06-20 | 华南理工大学 | Alarm method for monitoring system |
CN102663513A (en) * | 2012-03-13 | 2012-09-12 | 华北电力大学 | Combination forecast modeling method of wind farm power by using gray correlation analysis |
CN102663412A (en) * | 2012-02-27 | 2012-09-12 | 浙江大学 | Power equipment current-carrying fault trend prediction method based on least squares support vector machine |
CN102930168A (en) * | 2012-11-06 | 2013-02-13 | 昆山北极光电子科技有限公司 | Analysis method of signal processing wavelets |
CN103309235A (en) * | 2012-03-15 | 2013-09-18 | 华北计算机系统工程研究所 | Industrial control system active safety method |
CN104198922A (en) * | 2014-08-15 | 2014-12-10 | 电子科技大学 | Frequency selection method in early fault diagnosis of analog circuit |
CN104268658A (en) * | 2014-09-29 | 2015-01-07 | 招商局重庆交通科研设计院有限公司 | Bridge structure safety monitoring data prediction method |
CN104881705A (en) * | 2014-11-18 | 2015-09-02 | 中国石油大学(华东) | Method based on optimal combined model for predicting state trend of primary air fan |
CN105607575A (en) * | 2016-01-27 | 2016-05-25 | 电子科技大学 | FA-LSSVM based numerical control machine tool main shaft heat drift modeling method |
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 |
-
2008
- 2008-06-11 CN CN2008100446712A patent/CN101329697B/en not_active Expired - Fee Related
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102222151B (en) * | 2011-07-21 | 2012-08-22 | 电子科技大学 | Analog circuit fault prediction method based on ARMA (Autoregressive Moving Average) |
CN102222151A (en) * | 2011-07-21 | 2011-10-19 | 电子科技大学 | Analog circuit fault prediction method based on ARMA (Autoregressive Moving Average) |
CN102499651A (en) * | 2011-10-24 | 2012-06-20 | 华南理工大学 | Alarm method for monitoring system |
CN102663412A (en) * | 2012-02-27 | 2012-09-12 | 浙江大学 | Power equipment current-carrying fault trend prediction method based on least squares support vector machine |
CN102663412B (en) * | 2012-02-27 | 2014-11-05 | 浙江大学 | Power equipment current-carrying fault trend prediction method based on least squares support vector machine |
CN102663513B (en) * | 2012-03-13 | 2016-04-20 | 华北电力大学 | Utilize the wind power combined prediction modeling method of grey relational grade analysis |
CN102663513A (en) * | 2012-03-13 | 2012-09-12 | 华北电力大学 | Combination forecast modeling method of wind farm power by using gray correlation analysis |
CN103309235A (en) * | 2012-03-15 | 2013-09-18 | 华北计算机系统工程研究所 | Industrial control system active safety method |
CN102930168A (en) * | 2012-11-06 | 2013-02-13 | 昆山北极光电子科技有限公司 | Analysis method of signal processing wavelets |
CN104198922A (en) * | 2014-08-15 | 2014-12-10 | 电子科技大学 | Frequency selection method in early fault diagnosis of analog circuit |
CN104198922B (en) * | 2014-08-15 | 2017-02-01 | 电子科技大学 | Frequency selection method in early fault diagnosis of analog circuit |
CN104268658A (en) * | 2014-09-29 | 2015-01-07 | 招商局重庆交通科研设计院有限公司 | Bridge structure safety monitoring data prediction method |
CN104268658B (en) * | 2014-09-29 | 2017-10-10 | 招商局重庆交通科研设计院有限公司 | A kind of Forecasting Methodology of bridge structure safe Monitoring Data |
CN104881705A (en) * | 2014-11-18 | 2015-09-02 | 中国石油大学(华东) | Method based on optimal combined model for predicting state trend of primary air fan |
CN105607575A (en) * | 2016-01-27 | 2016-05-25 | 电子科技大学 | FA-LSSVM based numerical control machine tool main shaft heat drift modeling method |
CN105607575B (en) * | 2016-01-27 | 2017-12-26 | 电子科技大学 | Main shaft of numerical control machine tool thermal drift modeling method based on FA LSSVM |
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 |
Also Published As
Publication number | Publication date |
---|---|
CN101329697B (en) | 2011-03-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101329697B (en) | Method for predicting analog circuit state based on immingle algorithm | |
US8606418B1 (en) | Wind prediction for wind farms through the use of weather radar | |
CN108089099A (en) | The diagnostic method of distribution network failure based on depth confidence network | |
US20230023931A1 (en) | Hydraulic turbine cavitation acoustic signal identification method based on big data machine learning | |
WO2018133924A1 (en) | Methods and apparatus for analysing performance of a telecommunications network | |
CN105631578A (en) | Risk assessment-orientated modeling method of power transmission and transformation equipment failure probability model | |
CN105071983A (en) | Abnormal load detection method for cloud calculation on-line business | |
CN102111312A (en) | Multi-scale principle component analysis-based network abnormity detection method | |
Hussain et al. | Vibration analysis and time series prediction for wind turbine gearbox prognostics | |
CN112001644B (en) | Method, device, terminal and storage medium for detecting operation reliability of power distribution network | |
CN102495949A (en) | Fault prediction method based on air data | |
CN102179722A (en) | Estimation method for operational reliability of numerical control machine based on proportional hazards model | |
CN104598736A (en) | Roller bearing service life predicting model of self-adaptive multi-kernel combination relevance vector machine | |
Tagliaferri et al. | Wind modelling with nested Markov chains | |
CN105487009A (en) | Motor fault diagnosis method based on k-means RBF neural network algorithm | |
CN102915515A (en) | Cascade outage recognition and risk evaluation method of power grid | |
CN103197663A (en) | Method and system of failure prediction | |
CN115983374A (en) | Cable partial discharge database sample expansion method based on optimized SA-CACGAN | |
CN109635430B (en) | Power grid transmission line transient signal monitoring method and system | |
Abdel-Karim et al. | Short term wind speed prediction by finite and infinite impulse response filters: A state space model representation using discrete markov process | |
CN104239689A (en) | Optimal relevance vector machine based short-term wind speed prediction method | |
CN109948925A (en) | Consider the power communication system reliability estimation method that weather influences | |
CN103441806B (en) | Pure discontinuous Markov process spectrum sensing method for cognitive radio | |
CN104699994A (en) | FBF (fuzzy basis function) neural network based motor air gap eccentricity fault diagnosis method | |
CN105991362B (en) | The fluctuation threshold range setting method and device of data traffic |
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 | ||
C17 | Cessation of patent right | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20110323 Termination date: 20130611 |