CN113076708A - Analog circuit fault diagnosis method based on optimization matrix random forest algorithm - Google Patents

Analog circuit fault diagnosis method based on optimization matrix random forest algorithm Download PDF

Info

Publication number
CN113076708A
CN113076708A CN202110349866.3A CN202110349866A CN113076708A CN 113076708 A CN113076708 A CN 113076708A CN 202110349866 A CN202110349866 A CN 202110349866A CN 113076708 A CN113076708 A CN 113076708A
Authority
CN
China
Prior art keywords
matrix
random forest
fault diagnosis
data
feature
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.)
Pending
Application number
CN202110349866.3A
Other languages
Chinese (zh)
Inventor
谈恩民
阮济民
黄顺梅
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.)
Guilin University of Electronic Technology
Original Assignee
Guilin University of Electronic Technology
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 Guilin University of Electronic Technology filed Critical Guilin University of Electronic Technology
Priority to CN202110349866.3A priority Critical patent/CN113076708A/en
Publication of CN113076708A publication Critical patent/CN113076708A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/30—Circuit design
    • G06F30/32—Circuit design at the digital level
    • G06F30/33—Design verification, e.g. functional simulation or model checking
    • G06F30/3308—Design verification, e.g. functional simulation or model checking using simulation
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/24323—Tree-organised classifiers
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/259—Fusion by voting

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Computer Hardware Design (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Geometry (AREA)
  • Tests Of Electronic Circuits (AREA)

Abstract

The invention discloses an analog circuit fault diagnosis method based on an optimized matrix random forest algorithm, which comprises the following steps of 1) completing a simulation experiment through a circuit schematic diagram, and measuring original data of components with different set values; 2) performing local mean decomposition on the measured original data to obtain an optimized matrix; 3) calculating the optimized matrix to obtain a reduced-dimension output voltage matrix; 4) equally dividing the output voltage matrix obtained in the step 3), and using one part of the output voltage matrix as a training set to obtain an optimal solution of the parameters of the decision tree; the other part is used as a test set; 5) inputting the test set data in the step 4) into the decision tree which is trained in the step 4) and used for finding the optimal solution, and obtaining the fault diagnosis rate through random forest calculation of the obtained optimal solution parameters. The method uses an algorithm to complete feature extraction and feature classification, saves a large amount of time and saves the test cost.

Description

Analog circuit fault diagnosis method based on optimization matrix random forest algorithm
Technical Field
The invention relates to the field of analog circuit fault diagnosis, in particular to analog circuit fault feature extraction and feature classification, and particularly relates to an analog circuit fault diagnosis method based on an optimization matrix random forest algorithm.
Background
As technology and times continue to advance, analog circuits are used in more and more fields, but as the application of analog circuits increases, many problems such as faults occur in analog circuits, so research on fault diagnosis of analog circuits is necessary at present. The most important of the analog circuit fault diagnosis is two parts, namely feature extraction and feature classification, and the feature extraction and the feature classification have a plurality of methods, wherein a good feature extraction can save a large amount of time for diagnosis and can save test cost, the feature extraction methods have a plurality of methods, such as wavelet transformation, particle swarm, Random Forest (RF) algorithm and other methods, and none of the methods can be applied to an actual circuit for analog circuit fault diagnosis in a mature manner, and because the analog circuit original has a tolerance problem of 5%, the feature difference extracted among various faults is reduced, so that the existing analog circuit fault diagnosis method is jointly completed by combining the two algorithms of feature extraction and feature classification, the diagnosis time is greatly increased, and the efficiency is lower.
Disclosure of Invention
The invention aims to provide a fault diagnosis method of an analog circuit based on an optimized matrix random forest algorithm aiming at the defects of the prior art, which solves the problem of tolerance of circuit elements, and the extracted features have large difference among faults and low similarity among the same fault, thereby improving the fault diagnosis rate.
The technical scheme for realizing the purpose of the invention is as follows:
the analog circuit fault diagnosis method based on the optimization matrix random forest algorithm comprises the following steps:
1) completing a simulation experiment through a circuit schematic diagram, and measuring original data of components with different set values;
2) performing local mean decomposition on the measured original data to obtain an optimized matrix;
3) calculating the optimized matrix in MATLAB by using an RF algorithm to obtain a reduced-dimension output voltage matrix;
4) equally dividing the output voltage matrix obtained in the step 3), averagely dividing the output voltage matrix into two parts, wherein one part is used as a training set and is used for obtaining the optimal solution of the parameters of the decision tree; the other part is used as a test set;
5) inputting the test set data in the step 4) into the decision tree which is trained in the step 4) and used for finding the optimal solution, and obtaining the fault diagnosis rate through a random forest algorithm which has obtained the optimal solution parameters.
Further, step 1), carrying out transient analysis and Monte Carlo analysis on the Sallen _ Key circuit and the logarithmic amplification circuit by using Cadence software, and taking half of obtained data as a training set and half as a test set according to a set value of a fault; the training set has more training data, so that the diagnosis rate can be improved.
Further, step 2), the local mean decomposition provides a more optimized feature matrix for the output matrix, and the local mean decomposition is as follows: setting the output voltage values as a function v (t) along with the time, arranging the measured voltage values, and obtaining m based on the formulas (1), (2) and (3) through the principle of local mean decompositioni，ni，HiBased on the formula (4), demodulating H (t), finally obtaining f (t), wherein epsilon in denominator of f (t) is not 0, rearranging values of obtained f (t) functions to form a new matrix, obtaining an optimized matrix, and sending the finally obtained matrix into a random forest algorithm for feature extraction;
Figure BDA0003002074450000021
Figure BDA0003002074450000022
Hi＝vi-mi (3)
Figure BDA0003002074450000023
i represents time: can be 1, 2, 3 … … (ms), miN (t) denotes niH (t) represents HiTime function of (m)iDenotes viAnd vi+1Average value of viDenotes the voltage, v, at the i-th secondi+1Denotes the voltage of i +1 second, niDenotes viAnd vi+1F (t) represents the final demodulation function, HiRepresenting a function of the voltage difference.
Further, step 4), generating a decision tree by using a CART algorithm, wherein the regression tree is divided by the principle of the minimum mean square error Z of the formula (5), wherein A represents any feature, a represents any division point, and D1And D2A representative data set;
Figure BDA0003002074450000024
in the formula
Figure BDA0003002074450000025
Denotes the minimum value of the arbitrary dividing point a in the arbitrarily divided data set A, c1Represents D1Output average value of (1), c2Represents D2The average value of the outputs in (1),
Figure BDA0003002074450000026
represents that c is calculated so that the sum of squared errors is minimized1The value of the one or more of,
Figure BDA0003002074450000027
represents that c is calculated so that the sum of squared errors is minimized2Value, yiRepresenting a continuous variable, xiRepresenting data in a dataset;
the RF pair feature goodness measure is done based on equation (6):
Figure BDA0003002074450000031
import represents the importance degree, and N represents the number of decision trees;
the calculation of the error of the out-of-bag data is performed by selecting the corresponding out of bag data (OOB) from each decision tree, and is recorded as errOOB1By adding a part of interference to the data outside the bag, the errOOB is obtained by calculating the error2Obtaining the most important features in the random forest by the formula (2), and calculating the importance of each feature, wherein the feature selection process comprises the following steps:
4-1) calculating the importance of each feature, and sorting the features according to descending order;
4-2) selecting and removing some data with unobvious characteristics, and removing corresponding data according to the importance of the characteristics to obtain a new characteristic set;
4-3) repeating the above process with the new set of features until 13 features remain;
4-4) selecting the feature set with the lowest out-of-bag error rate according to the feature sets obtained in the process;
the random forest feature classification is mainly completed by a decision tree, and the specific construction process is as follows:
4-a) sampling from the sample set with Bootstrap to select 50% of the samples;
4-b) randomly selecting K attributes from all the attributes, and selecting the best segmentation attribute from the K attributes as a node to create a decision tree;
4-c) repeating the process for m times to establish m decision trees, wherein the decision trees can be established in parallel, so that a large amount of time can be saved;
4-d), the final result is finished by voting m decision trees of the random forest, and the decision with the largest occurrence number determines the result of the predicted data.
Further, step 5), simulation experiments are carried out on the Sallen _ Key circuit and the logarithmic amplifier circuit of the test circuit by using Cadence software, half of training samples in the finally obtained data set are used for training the random forest algorithm, the rest half of samples are used for fault diagnosis, and finally the fault diagnosis rate is obtained.
The analog circuit fault diagnosis method for optimizing the matrix stochastic son algorithm solves the problem of tolerance of circuit elements, the difference between the extracted characteristics of the analog circuit fault diagnosis method is large, the similarity between the same type of faults is low, and the fault diagnosis rate is improved. The RF algorithm provided by the invention has the performance of optimizing parameters, is an algorithm established on the basis of a statistical theory, has a more advanced combined classifier, is extracted by resampling bootstraps to classify the features, is classified by a decision tree algorithm in the RF, is a relatively new machine learning, is rapidly developed in recent years, and has a better effect in the fields of biology, medicine, economy and the like.
The method provided by the invention enables the output power supply to have more accurate characteristic voltage value by using a multi-input voltage source; the voltage value of the output matrix is accurately measured each time in the simulation, and the fault diagnosis is carried out on the decomposed matrix by a random forest algorithm through the principle of local decomposition on the matrix formed by the measured output voltage values, so that the random forest can more effectively carry out feature extraction, the single fault diagnosis rate reaches 100%, the multiple fault diagnosis rate reaches 99.5%, and the method is higher than that of the prior art, and the fault diagnosis method provided by the invention has certain advantages. The method uses an algorithm to complete feature extraction and feature classification, saves a large amount of time and saves the test cost.
Drawings
FIG. 1 is a diagram of an embodiment of a decision tree structure;
FIG. 2 is a flow chart of analog circuit fault diagnosis in an embodiment;
FIG. 3 is a schematic diagram of an embodiment of an RF algorithm;
FIG. 4 is a schematic diagram of a Sallen _ Key circuit in an embodiment;
FIG. 5 is a diagram showing a normal state response of the Sallen _ key circuit in the embodiment;
FIG. 6 is a diagram illustrating selection of a decision tree in an embodiment;
FIG. 7 shows the Sallen _ Key circuit fault diagnosis result in the embodiment;
FIG. 8 is a schematic diagram of a logarithmic amplifier circuit in an embodiment;
FIG. 9 is a normal state response diagram of the logarithmic amplifier circuit in the embodiment;
fig. 10 shows the multi-fault diagnosis result of the logarithmic amplifier circuit in the embodiment.
Detailed Description
The technical solutions in the embodiments of the present invention will be fully described below with reference to the accompanying drawings and tables in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example (b):
an analog circuit fault diagnosis method based on an optimization matrix random forest algorithm comprises the following steps:
1) completing a simulation experiment through a circuit schematic diagram, and measuring original data of components with different set values;
2) performing local mean decomposition on the measured original data to obtain an optimized matrix;
3) calculating the optimized matrix in MATLAB by using an RF algorithm to obtain a reduced-dimension output voltage matrix;
4) equally dividing the output voltage matrix obtained in the step 3), averagely dividing the output voltage matrix into two parts, wherein one part is used as a training set and is used for obtaining the optimal solution of the parameters of the decision tree; the other part is used as a test set;
5) inputting the test set data in the step 4) into the decision tree which is trained in the step 4) and used for finding the optimal solution, and obtaining the fault diagnosis rate through the optimized random forest algorithm, as shown in fig. 2.
Further, step 1), performing transient analysis and monte carlo analysis on the salen _ Key circuit as shown in fig. 4 and the logarithmic amplification circuit as shown in fig. 8 by using Cadence software, wherein time is set to be 1ms, performing 100 monte carlo analyses, and obtaining 1100 data by using a set value of a fault, wherein half of the data is used as a training set and half of the data is used as a test set; the training set has more data, so that the diagnosis rate can be improved, and random forest feature extraction is performed on all fault samples;
selecting a resistor R1, a resistor R2, a resistor R3, a capacitor C1 and a capacitor C2 as fault elements through sensitivity analysis, wherein the tolerance of the resistor in the circuit is 5 percent, the tolerance of the capacitor is 10 percent, and when the corresponding resistance or capacitance of the circuit element exceeds the range, the circuit element is in fault; NF indicates a normal state, "-" indicates that data is not present, "↓" indicates exceeding the standard value in the table, and "↓" indicates falling below the standard value in the table in table 1, and the final diagnosis result is shown in fig. 7.
TABLE 1 Sallen _ Key Circuit element Fault parameter values
Figure BDA0003002074450000051
Sensitivity analysis is carried out on the logarithmic amplifier circuit, the resistor R4, the resistor R6 and the capacitor C1 are selected to be fault elements, 3 fault elements can be divided into 10 faults and 1 normal state, the total number of the circuit states is 11, parameter values of specific fault elements are shown in table 2, Monte Carlo analysis and transient analysis are carried out on the logarithmic amplifier circuit under normal conditions through Cadence, and the final diagnosis result is shown in fig. 10.
TABLE 2 logarithmic amplifier circuit element multiple fault parameter values
Figure BDA0003002074450000061
Further, step 2), the local mean decomposition provides a more optimized characteristic matrix for the output matrix, which well reflects the essence of the fault information, so that the method is very suitable for the characteristic extraction of the analog circuitThe local mean decomposition is as follows: setting the output voltage values as a function v (t) along with the time, arranging the measured voltage values, and obtaining m based on the formulas (1), (2) and (3) through the principle of local mean decompositioni，ni，HiBased on the formula (4), demodulating H (t), finally obtaining f (t), wherein epsilon in denominator of f (t) is not 0, rearranging values of obtained f (t) functions to form a new matrix, obtaining an optimized matrix, and sending the finally obtained matrix into a random forest algorithm for feature extraction;
Figure BDA0003002074450000062
Figure BDA0003002074450000063
Hi＝vi-mi (3)
Figure BDA0003002074450000064
i represents time: can be 1, 2, 3 … … (ms), miN (t) denotes niH (t) represents HiTime function of (m)iDenotes viAnd vi+1Average value of viDenotes the voltage, v, at the i-th secondi+1Denotes the voltage of i +1 second, niDenotes viAnd vi+1F (t) represents the final demodulation function, HiRepresenting a function of the voltage difference.
Further, step 3), during feature extraction, it should be noted that a certain tolerance exists in circuit components, generally about 5%, which affects the final components to have a certain uncertainty, which will increase the difficulty of feature extraction in the fault diagnosis of the analog circuit, and when feature extraction is performed, a feature overlapping phenomenon occurs, which results in a reduction of feature discrimination and ultimately affects the fault diagnosis rate, the method uses a multi-input excitation source, which will increase effective features, can better complete feature extraction through local mean decomposition, can better perform feature extraction on the output voltage value of the tested circuit for a random forest algorithm, and the local mean decomposition provides a more optimized feature matrix for the output matrix, which well reflects the essence of fault information, so the method is very suitable for feature extraction of the analog circuit, and sending the finally obtained matrix into a random forest algorithm for feature extraction.
Further, step 4), a good feature classification can improve the diagnosis rate and improve the classification precision, and at present, the method mainly comprises the following methods for performing feature classification, such as a least square Support Vector Machine (SVM), an extreme learning Machine, an artificial neural network algorithm and the like, wherein the good or bad effect of the classification algorithm depends on the parameters of the algorithm, and the good or bad effect of the parameters depends on the training of the classifier parameters, so that the problem of the classifier parameters is solved and becomes the important factor of feature classification in the analog circuit fault diagnosis, some optimization algorithms can optimize the parameters, a particle swarm algorithm can optimize the parameters of the artificial neural network, a wolfsbane swarm algorithm can optimize the parameters of the extreme learning Machine, a firefly algorithm can optimize the chaotic parameters, and the optimized parameters have higher fault diagnosis rate. The random forest is a machine learning algorithm with supervised learning based on a guided aggregation algorithm (Bagging), wherein Bagging is a randomly extracted training set sample put back from an original set, the Bootstrap algorithm is converted from an integrated learning idea and trained to obtain a single weak learner, the generated weak learner becomes a decision tree in the random forest algorithm, the process is repeated to generate a plurality of decision trees, the regression trees form the random forest together as shown in figure 1, the predicted average value of the decision tree is very important, the average value can influence or change the final value, the decision tree algorithm generally comprises three types of ID3, CART and C4.5, the three algorithms have tree structures, and each tree structure has a tree structureThe greedy algorithm from top to bottom is constructed, splitting is performed at each node, and an optimal feature is selected to be passed to the next tree structure until the required conditions are finally met, as shown in fig. 3. The method uses CART algorithm to generate decision tree, regression tree is divided by the principle of formula (5) minimum mean square error Z, wherein A represents any characteristic, a represents any division point, D represents any division point, and A represents any division point1And D2Representing a data set, a schematic diagram is shown in fig. 1.
Figure BDA0003002074450000081
In the formula (I), the compound is shown in the specification,
Figure BDA0003002074450000082
denotes the minimum value of the arbitrary dividing point a in the arbitrarily divided data set A, c1Represents D1Output average value of (1), c2Represents D2The average value of the outputs in (1),
Figure BDA0003002074450000083
represents that c is calculated so that the sum of squared errors is minimized1The value of the one or more of,
Figure BDA0003002074450000084
represents that c is calculated so that the sum of squared errors is minimized2Value, yiRepresenting a continuous variable, xiRepresenting data in the data set.
The random forest algorithm is mainly used for feature extraction according to the OOB (out of bag) principle, a certain feature is assumed to be important, certain noise is introduced into the feature data in a distributed mode, only the data after the feature is changed are subjected to RF training, the performance of the matrix model is greatly changed, a certain feature is assumed to be unimportant, the changed data are also subjected to RF training, the performance of the matrix model is not greatly changed, and the RF pair feature quality metric is completed based on the formula (6).
Figure BDA0003002074450000085
In the formula, import represents the importance degree, and N represents the number of the decision tree.
The calculation of the error of the out-of-bag data is performed by selecting the corresponding out of bag data (OOB) from each decision tree, and is recorded as errOOB1By adding a part of interference to the data outside the bag, the errOOB is obtained by calculating the error2Obtaining the most important features in the random forest by the formula (2), and calculating the importance of each feature, wherein the feature selection process comprises the following steps:
4-1) calculating the importance of each feature, and sorting the features according to descending order;
4-2) selecting and removing some data with unobvious characteristics, and removing corresponding data according to the importance of the characteristics to obtain a new characteristic set;
4-3) repeating the above process with the new set of features until 13 features remain;
4-4) selecting the feature set with the lowest out-of-bag error rate according to the feature sets obtained in the process described above.
The random forest feature classification is mainly completed by a decision tree, the decision tree is a small classifier in nature, how many trees will generate how many classification results, the random forest determines the final result in a classification voting mode, the random forest with the largest voting times will become the final output result, and the random forest is trained and constructed into a data set in a sampling mode with a put-back mode, so that each subset has the same selection weight, the decision tree is constructed jointly into a random forest through the subset data, an output result is generated by putting prediction data into each subset data set, the mode is considered as a voting mode, and the forest with the largest voting number will become the final output result of the random forest, and the specific construction process comprises the following steps:
4-a) sampling from the sample set with Bootstrap to select 50% of the samples;
4-b) randomly selecting K attributes from all the attributes, and selecting the best segmentation attribute from the K attributes as a node to create a decision tree;
4-c) repeating the above process m times to build m decision trees, which can adopt parallel decision tree building, thus saving a lot of time, as shown in FIG. 6;
4-d), the final result is finished by voting m decision trees of the random forest, and the decision with the largest occurrence number determines the result of the predicted data.
Further, step 5), simulation experiments are carried out on the Sallen _ Key circuit and the logarithmic amplifier circuit of the test circuit by using Cadence software, transient analysis is carried out for 1ms in running time, Monte Carlo analysis is carried out for 100 times in running time, 1100 data sets are finally obtained, 550 training samples in the transient analysis are used for training a random forest algorithm for fault diagnosis of the logarithmic amplifier circuit of the test circuit, and the remaining 550 samples are used for fault diagnosis to finally obtain a fault diagnosis rate.
The preferred embodiments of the present invention have been disclosed for illustrative purposes only and are not intended to limit the invention to the specific embodiments described. Obviously, many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the invention and the practical application, to thereby enable others skilled in the art to best utilize the invention.

Claims (7)

1. The analog circuit fault diagnosis method based on the optimization matrix random forest algorithm is characterized by comprising the following steps:
1) completing a simulation experiment through a circuit schematic diagram, and measuring original data of components with different set values;
2) performing local mean decomposition on the measured original data to obtain an optimized matrix;
3) calculating the optimized matrix in MATLAB by using an RF algorithm to obtain a reduced-dimension output voltage matrix;
4) equally dividing the output voltage matrix obtained in the step 3), averagely dividing the output voltage matrix into two parts, wherein one part is used as a training set and is used for obtaining the optimal solution of the parameters of the decision tree; the other part is used as a test set;
5) inputting the test set data in the step 4) into the decision tree which is trained in the step 4) and used for finding the optimal solution, and obtaining the fault diagnosis rate through random forest calculation of the obtained optimal solution parameters.
2. The analog circuit fault diagnosis method based on the optimized matrix random forest algorithm according to claim 1, wherein in step 1), Cadence software is used for carrying out transient analysis and Monte Carlo analysis on a Sallen _ Key circuit and a logarithmic amplification circuit, and half of obtained data is used as a training set and half is used as a testing set according to a set value of a fault.
3. The analog circuit fault diagnosis method based on the optimization matrix random forest algorithm as claimed in claim 1, wherein in step 2), the local mean decomposition provides a more optimized matrix of features for the output matrix, and the local mean decomposition is as follows: setting the output voltage values as a function v (t) along with the time, arranging the measured voltage values, and obtaining m based on the formulas (1), (2) and (3) through the principle of local mean decompositioni，ni，HiBased on the formula (4), demodulating H (t), finally obtaining f (t), wherein epsilon in denominator of f (t) is not 0, rearranging values of obtained f (t) functions to form a new matrix, obtaining an optimized matrix, and sending the finally obtained matrix into a random forest algorithm for feature extraction;
Figure FDA0003002074440000011
Figure FDA0003002074440000012
Hi＝vi-mi (3)
Figure FDA0003002074440000013
i represents time: can be 1, 2, 3 … … (ms), miN (t) denotes niH (t) represents HiTime function of (m)iDenotes viAnd vi+1Average value of viDenotes the voltage, v, at the i-th secondi+1Denotes the voltage of i +1 second, niDenotes viAnd vi+1F (t) represents the final demodulation function, HiRepresenting a function of the voltage difference.
4. The analog circuit fault diagnosis method based on the optimization matrix random forest algorithm as claimed in claim 1, wherein in step 4), a CART algorithm is used to generate a decision tree, and a regression tree is divided by the principle of the minimum mean square error Z of formula (5), wherein A represents an arbitrary feature, a represents an arbitrary division point, and D represents an arbitrary division point1And D2A representative data set;
Figure FDA0003002074440000021
in the formula
Figure FDA0003002074440000022
Denotes the minimum value of the arbitrary dividing point a in the arbitrarily divided data set A, c1Represents D1Output average value of (1), c2Represents D2The average value of the outputs in (1),
Figure FDA0003002074440000023
represents that c is calculated so that the sum of squared errors is minimized1The value of the one or more of,
Figure FDA0003002074440000024
represents that c is calculated so that the sum of squared errors is minimized2Value, yiRepresenting a continuous variable, xiRepresenting data in a dataset;
the RF pair feature goodness measure is done based on equation (6):
Figure FDA0003002074440000025
import represents the importance degree, and N represents the number of decision trees;
the calculation of the error of the out-of-bag data is performed by selecting the corresponding out of bag data (OOB) from each decision tree, and is recorded as errOOB1By adding a part of interference to the data outside the bag, the errOOB is obtained by calculating the error2。
5. The analog circuit fault diagnosis method based on the optimization matrix random forest algorithm as claimed in claim 4, wherein the most important features in the random forest are obtained through equation (2), the importance of each feature is calculated, and the feature selection process is as follows:
4-1) calculating the importance of each feature, and sorting the features according to descending order;
4-2) selecting and removing some data with unobvious characteristics, and removing corresponding data according to the importance of the characteristics to obtain a new characteristic set;
4-3) repeating the above process with the new set of features until 13 features remain;
4-4) selecting the feature set with the lowest out-of-bag error rate according to the feature sets obtained in the process described above.
6. The analog circuit fault diagnosis method based on the optimization matrix random forest algorithm as claimed in claim 4, wherein the random forest feature classification is mainly completed by a decision tree, and the specific construction process is as follows:
4-a) sampling from the sample set with Bootstrap to select 50% of the samples;
4-b) randomly selecting K attributes from all the attributes, and selecting the best segmentation attribute from the K attributes as a node to create a decision tree;
4-c) repeating the process for m times, establishing m decision trees, and adopting parallel establishment of the decision trees;
4-d), the final result is finished by voting m decision trees of the random forest, and the decision with the largest occurrence number determines the result of the predicted data.
7. The analog circuit fault diagnosis method based on the optimized matrix random forest algorithm as claimed in claim 1, wherein in step 5), Cadence software is used for carrying out simulation experiments on a Sallen _ Key circuit and a logarithmic amplifier circuit of a test circuit, half of training samples in a finally obtained data set are used for training the random forest algorithm, the rest half of the training samples are used for fault diagnosis, and the fault diagnosis rate is finally obtained.
CN202110349866.3A 2021-03-31 2021-03-31 Analog circuit fault diagnosis method based on optimization matrix random forest algorithm Pending CN113076708A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110349866.3A CN113076708A (en) 2021-03-31 2021-03-31 Analog circuit fault diagnosis method based on optimization matrix random forest algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110349866.3A CN113076708A (en) 2021-03-31 2021-03-31 Analog circuit fault diagnosis method based on optimization matrix random forest algorithm

Publications (1)

Publication Number Publication Date
CN113076708A true CN113076708A (en) 2021-07-06

Family

ID=76614728

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110349866.3A Pending CN113076708A (en) 2021-03-31 2021-03-31 Analog circuit fault diagnosis method based on optimization matrix random forest algorithm

Country Status (1)

Country Link
CN (1) CN113076708A (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102809934A (en) * 2012-08-20 2012-12-05 桂林电子科技大学 Boundary scan test controller for mixed signal circuit
CN108303632A (en) * 2017-12-14 2018-07-20 佛山科学技术学院 Circuit failure diagnosis method based on random forests algorithm
CN110287552A (en) * 2019-06-10 2019-09-27 合肥工业大学 Based on the motor bearings fault diagnosis method and system for improving random forests algorithm
CN110879351A (en) * 2019-11-28 2020-03-13 山东科技大学 Fault diagnosis method for non-linear analog circuit based on RCCA-SVM
CN111507518A (en) * 2020-04-14 2020-08-07 中铁隧道局集团有限公司 Wavelet neural network concrete impermeability prediction method based on random forest

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102809934A (en) * 2012-08-20 2012-12-05 桂林电子科技大学 Boundary scan test controller for mixed signal circuit
CN108303632A (en) * 2017-12-14 2018-07-20 佛山科学技术学院 Circuit failure diagnosis method based on random forests algorithm
CN110287552A (en) * 2019-06-10 2019-09-27 合肥工业大学 Based on the motor bearings fault diagnosis method and system for improving random forests algorithm
CN110879351A (en) * 2019-11-28 2020-03-13 山东科技大学 Fault diagnosis method for non-linear analog circuit based on RCCA-SVM
CN111507518A (en) * 2020-04-14 2020-08-07 中铁隧道局集团有限公司 Wavelet neural network concrete impermeability prediction method based on random forest

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
秦喜文: "基于LMD与随机森林的滚动轴承故障诊断", 《长春工业大学学报》 *
陈峰: "基于CART算法的空气质量指数回归预测模型的学习", 《上饶师范学院学报》 *

Similar Documents

Publication Publication Date Title
CN108228716B (en) SMOTE _ Bagging integrated sewage treatment fault diagnosis method based on weighted extreme learning machine
CN110175386B (en) Method for predicting temperature of electrical equipment of transformer substation
CN108170765B (en) Poverty-stricken and living fund assisting recommendation method based on multidimensional analysis of on-school behavior data
CN106485262A (en) A kind of bus load Forecasting Methodology
CN115099519B (en) Oil well yield prediction method based on multi-machine learning model fusion
CN112149873B (en) Low-voltage station line loss reasonable interval prediction method based on deep learning
CN112613536B (en) Near infrared spectrum diesel fuel brand recognition method based on SMOTE and deep learning
CN105572572A (en) WKNN-LSSVM-based analog circuit fault diagnosis method
CN112800115B (en) Data processing method and data processing device
CN111200141B (en) Proton exchange membrane fuel cell performance prediction and optimization method based on deep belief network
CN105608329A (en) Organizational behavior anomaly detection method based on community evolution
CN112016253A (en) High-fidelity chaotic polynomial correction method suitable for CFD uncertainty quantification
CN103440275A (en) Prim-based K-means clustering method
CN103678681B (en) The Multiple Kernel Learning sorting technique of the auto-adaptive parameter based on large-scale data
CN110569883A (en) Air quality index prediction method based on Kohonen network clustering and Relieff feature selection
CN114330486A (en) Power system bad data identification method based on improved Wasserstein GAN
CN116245019A (en) Load prediction method, system, device and storage medium based on Bagging sampling and improved random forest algorithm
CN109164794A (en) Multivariable industrial process Fault Classification based on inclined F value SELM
CN113537556A (en) Household short-term load prediction method based on state frequency memory network
CN104182854A (en) Mixed energy consumption measuring method for enterprise energy management system
CN112418504A (en) Wind speed prediction method based on mixed variable selection optimization deep belief network
CN113076708A (en) Analog circuit fault diagnosis method based on optimization matrix random forest algorithm
CN113537366B (en) Transient stability evaluation method for power system
Ishibashi et al. Knowledge extraction using a genetic fuzzy rule-based system with increased interpretability
CN115293230A (en) Regional ecological harmony LSTM algorithm prediction method

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
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20210706

WD01 Invention patent application deemed withdrawn after publication