CN115186564A - Analog circuit fault diagnosis method based on feature fusion and improved particle swarm algorithm - Google Patents
Analog circuit fault diagnosis method based on feature fusion and improved particle swarm algorithm Download PDFInfo
- Publication number
- CN115186564A CN115186564A CN202210515698.5A CN202210515698A CN115186564A CN 115186564 A CN115186564 A CN 115186564A CN 202210515698 A CN202210515698 A CN 202210515698A CN 115186564 A CN115186564 A CN 115186564A
- Authority
- CN
- China
- Prior art keywords
- particle
- particles
- fault
- particle swarm
- analog circuit
- 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
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/25—Design optimisation, verification or simulation using particle-based methods
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/28—Testing of electronic circuits, e.g. by signal tracer
- G01R31/316—Testing of analog circuits
- G01R31/3163—Functional testing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides an analog circuit fault diagnosis method based on feature fusion and improved particle swarm optimization, which comprises the steps of obtaining fault one-dimensional time domain data of an analog circuit, and converting the one-dimensional time domain data into two-dimensional time frequency image data; obtaining a first characteristic according to the one-dimensional time domain data and the one-dimensional convolutional neural network; obtaining a second characteristic according to the two-dimensional time-frequency image data and the depth residual error network; splicing the first characteristic and the second characteristic to obtain a final fault characteristic; obtaining a fault type according to the fault characteristics and a pre-trained support vector machine; the method comprises the steps of optimizing punishment parameters and kernel parameters of a support vector machine by adopting an improved particle swarm algorithm, calculating the comprehensive fraction of each particle according to a balance factor, the distance between the particle and the optimal population fitness particle and a fitness value in the improved particle swarm algorithm, and selecting the particle with the highest comprehensive fraction as the comprehensive optimal particle according to the order of the fractions of the particles from high to low; the invention greatly improves the fault diagnosis precision of the analog circuit.
Description
Technical Field
The invention relates to the technical field of analog circuit fault diagnosis, in particular to an analog circuit fault diagnosis method based on feature fusion and improved particle swarm optimization.
Background
The statements in this section merely provide background information related to the present disclosure and may not constitute prior art.
In electronic products, circuits are the core components. The circuit can be divided into a digital circuit and an analog circuit. Relevant research shows that in modern electronic products, the integrated digital circuit accounts for up to 80%, and the analog circuit accounts for only 20%. But 80% of the faults in the circuit are analog circuit faults, and the faults occurring in the digital circuit are only 20%. Faults in digital circuits are relatively easy to identify and locate, since the analysis of digital signals is relatively simple and easy to observe. And because the components such as the capacitor and the resistor of the analog circuit have tolerance, faults are difficult to identify and the probability of the occurrence of the faults is higher. Therefore, the performance of the analog circuit has a decisive influence on the stability and reliability of the entire circuit system.
The difficulty of analog circuit fault diagnosis research is mainly as follows:
(1) The fault bound is fuzzy. Unlike digital circuits, the signals in analog circuits are continuous and cannot be simply quantified; meanwhile, the boundary between the normal working state and the fault state of the circuit is fuzzy, so that the fault is difficult to diagnose and position.
(2) The degree of fault coupling is high. The coupling between each component in the analog circuit is strong, and if one component fails, other components are likely to fail in a short time; single faults are mixed with faults caused by two or even a plurality of components, and the difficulty of diagnosis is increased.
(3) Is susceptible to the external environment. Information in the analog circuit is transmitted in the form of analog signals, and factors such as temperature, humidity, noise, electromagnetism and the like of a working environment can generate noise, so that collected fault data are more difficult to distinguish.
(4) And the number of actually measurable nodes is small. With the increase of the integration level of the circuit, a large number of inaccessible test nodes exist in the actual circuit, data for fault diagnosis can be collected only in a small number of accessible test nodes, and meanwhile, compared with a normal state, the fault state time is short, so that fault data samples are more rare.
The inventor finds that the early-stage fault of the analog circuit has the characteristics of strong coupling, few samples, fuzzy fault boundary and the like, fault characteristics extracted by the traditional characteristic extraction methods such as principal component analysis, wavelet packet decomposition and the like can not effectively distinguish various faults, and the characteristic self-learning method based on the neural network usually only extracts original one-dimensional time domain data or distinguishing information in transformed two-dimensional image data; because early failure samples are few, the features extracted by the feature self-learning method are often not rich enough.
The Support Vector Machine (SVM) is widely applied to the field of analog circuit fault diagnosis because of being prominent in the problems of small samples, nonlinearity and high-dimensional mode recognition, but the performance of the SVM is greatly influenced by model parameters; the parameters obtained by the traditional grid optimizing method are too coarse, and the performance of a support vector machine cannot be fully exerted. In order to improve the classification accuracy of the SVM, the particle swarm optimization is often used for parameter optimization of the SVM, but in a complex optimization problem, the original particle swarm optimization is prone to fall into local optimization. The existing improved particle swarm algorithm for SVM parameter optimization still has the problems of being susceptible to too fast loss of diversity, being prone to falling into local optimization, poor balance between global exploration and local development and the like in the field of early faults of analog circuits with strong coupling.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides the analog circuit fault diagnosis method based on feature fusion and improved particle swarm optimization, and the analog circuit fault diagnosis precision is greatly improved.
In order to achieve the purpose, the invention adopts the following technical scheme:
the invention provides a method for diagnosing faults of an analog circuit based on feature fusion and improved particle swarm optimization.
A fault diagnosis method for an analog circuit based on feature fusion and improved particle swarm optimization comprises the following steps:
acquiring fault one-dimensional time domain data of an analog circuit, and converting the one-dimensional time domain data into two-dimensional time frequency image data;
obtaining a first characteristic according to the one-dimensional time domain data and the one-dimensional convolutional neural network; obtaining a second characteristic according to the two-dimensional time-frequency image data and the depth residual error network; splicing the first characteristic and the second characteristic to obtain a final fault characteristic;
obtaining a fault type according to the fault characteristics and a pre-trained support vector machine;
the method comprises the steps of optimizing punishment parameters and kernel parameters of a support vector machine by adopting an improved particle swarm algorithm, calculating the comprehensive fraction of each particle according to a balance factor, the distance between the particle and the particle with the optimal population fitness and a fitness value in the improved particle swarm algorithm, and selecting the particle with the highest comprehensive fraction as the comprehensive optimal particle according to the order of the fractions of the particles from high to low.
As an optional implementation manner, the one-dimensional time domain data is converted into the two-dimensional time-frequency image data by using cross wavelet transform.
As an optional implementation manner, the distance between the ith particle and the particle Gbest with the optimal population fitness is as follows:
as an optional implementation manner, calculating a composite score of each particle according to the balance factor, the distance between the particle and the particle with the optimal population fitness, and the fitness value includes:
FDBscore i ＝α(1-normF i )+(1-α)normDG i
wherein α is a balance factor, norm F i For normalization of the fitness value, DG i And the normalized value of the distance between the ith particle and the particle Gbest with the optimal population fitness is obtained.
Further, in the above-mentioned case,
wherein, t max And t is the current iteration time.
As an optional implementation manner, in the improved particle swarm algorithm, the method further includes:
replacing the population fitness optimal particle Gbest by utilizing the comprehensive optimal particle FDBbest, and replacing a second speed formula by utilizing a first speed formula to update the particle speed; wherein the first velocity formula is:
the second velocity equation is:
c 1 and c 2 Is an empirical weight, r 1 And r 2 Is a discount factor;
performing cross operation on the comprehensive optimal particle FDBbest and the particles with poor adaptability;
introducing a probability factor, wherein the value of the probability factor is gradually linearly reduced from 0.5 to 0 along with the increase of the iteration times, and if the random number Rand is smaller than the probability factor, the particle carries out speed updating according to a first speed formula; otherwise, the speed is updated according to the second speed formula.
As an optional implementation manner, each particle develops a surrounding solution space according to its current position, including:
generating N random numbers which are uniformly distributed according to a U (0, 1) standard;
n random numbers rho i (i =1,2,..,. N) are ordered from large to small, while the N particles are ordered from small to large according to the fitness value;
the N random numbers correspond to the N particles one by one, so that the particles with small fitness value are paired with the large random numbers, and the particles with large fitness value are paired with the small random numbers;
in each iteration, if the particle corresponds to ρ i Greater than a threshold value P s Then the particle carries out particle individual self-development operation; otherwise, not carrying out the individual self-development operation of the particles;
all rho i Greater than the current threshold P s The particles of (2) are developed into their own surrounding area according to the following formula:
wherein norrndC id Means that a obedient N (0, C) is randomly generated id ) Normally distributed random number, C id The d-th dimension of the ith particle.
The invention provides an analog circuit fault diagnosis system based on feature fusion and improved particle swarm optimization.
An analog circuit fault diagnosis system based on feature fusion and improved particle swarm optimization comprises the following steps:
a data acquisition module configured to: acquiring fault one-dimensional time domain data of an analog circuit, and converting the one-dimensional time domain data into two-dimensional time frequency image data;
a feature extraction module configured to: obtaining a first characteristic according to the one-dimensional time domain data and the one-dimensional convolution neural network; obtaining a second characteristic according to the two-dimensional time-frequency image data and the depth residual error network; splicing the first characteristic and the second characteristic to obtain a final fault characteristic;
a fault diagnosis module configured to: obtaining a fault type according to the fault characteristics and a pre-trained support vector machine;
the method comprises the steps of optimizing punishment parameters and kernel parameters of a support vector machine by adopting an improved particle swarm algorithm, calculating the comprehensive fraction of each particle according to a balance factor, the distance between the particle and the particle with the optimal population fitness and a fitness value in the improved particle swarm algorithm, and selecting the particle with the highest comprehensive fraction as the comprehensive optimal particle according to the order of the fractions of the particles from high to low.
A third aspect of the present invention provides a computer-readable storage medium on which a program is stored, which when executed by a processor, implements the steps in the method for diagnosing a fault in an analog circuit based on feature fusion and improved particle swarm optimization as set forth in the first aspect of the present invention.
A fourth aspect of the present invention provides an electronic device, which includes a memory, a processor and a program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method for diagnosing faults of analog circuits based on feature fusion and improved particle swarm optimization as described in the first aspect of the present invention.
Compared with the prior art, the invention has the beneficial effects that:
1. the analog circuit fault diagnosis method based on the feature fusion and the improved particle swarm algorithm obtains a first feature according to one-dimensional time domain data and a one-dimensional convolution neural network; obtaining a second characteristic according to the two-dimensional time-frequency image data and the depth residual error network; and after the first characteristic and the second characteristic are spliced, the final fault characteristic is obtained, more accurate characteristic extraction is realized, and the fault diagnosis precision is improved.
2. According to the analog circuit fault diagnosis method based on the feature fusion and the improved particle swarm algorithm, in the improved particle swarm algorithm, the comprehensive score of each particle is calculated according to the balance factor, the distance between the particle and the particle with the optimal population fitness and the fitness value, the particles with the highest comprehensive score are selected as the comprehensive optimal particles according to the ranking of the particle scores from high to low, and the fault diagnosis precision is further improved.
3. The inventionAccording to the analog circuit fault diagnosis method based on the feature fusion and improved particle swarm optimization, the more likely the particles with better fitness develop the surrounding area of the particles; in the early stage of iteration, P s The value of (A) is small, the number of particles capable of carrying out ISE operation is large, the exploration range is also large, and the exploration capability of the population is strong; at the end of the iteration, P s The value of (A) is larger, only a few particles with better fitness can be developed in a small range by using an ISE strategy, and the development capability of the population is enhanced.
4. The invention relates to an analog circuit fault diagnosis method based on feature fusion and improved particle swarm optimization, wherein a probability factor P is introduced when a particle selects a speed updating mode f The value of which gradually decreases linearly from 0.5 to 0 with increasing number of iterations; if the random number Rand is less than P f Then the particle is updated according to the first formula; otherwise, updating the speed according to a second formula; the updating mode can not only reduce the probability of the population falling into the local optimum, but also keep the characteristic of the particle swarm algorithm of fast convergence.
Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the invention and together with the description serve to explain the invention and not to limit the invention.
Fig. 1 is a schematic flow chart of the HPSO-FDB-ISE algorithm provided in embodiment 1 of the present invention.
Fig. 2 is a schematic diagram of a Chebyshev filter circuit according to embodiment 1 of the present invention.
Fig. 3 is a schematic diagram of a fourier-op-amp filter circuit according to embodiment 1 of the present invention.
Fig. 4 is a diagram of an original feature distribution provided in embodiment 1 of the present invention.
Fig. 5 is a feature distribution diagram of deep feature fusion network extraction provided in embodiment 1 of the present invention.
FIG. 6 is a graph showing the convergence of HPSO-FDB-ISE and its variants provided in example 1 of the present invention.
Fig. 7 is a flowchart of analog circuit fault diagnosis provided in embodiment 1 of the present invention.
FIG. 8 is a comparison of the diagnostic accuracy provided in example 1 of the present invention.
Fig. 9 is a graph showing the convergence of the two algorithms provided in embodiment 1 of the present invention on a Chebyshev filter circuit.
Fig. 10 is a schematic distribution diagram of an original data set and a fault feature data set extracted by a deep feature fusion network of an actual Chebyshev filter circuit in a three-dimensional space according to embodiment 1 of the present invention.
Fig. 11 is a schematic distribution diagram of an original data set and a fault feature data set extracted by a deep feature fusion network of an actual fourier-op-amp filter circuit in a three-dimensional space according to embodiment 1 of the present invention.
Detailed Description
The invention is further described with reference to the following figures and examples.
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the invention. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the invention. As used herein, the singular forms "a", "an", and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
The embodiments and features of the embodiments of the present invention may be combined with each other without conflict.
Example 1:
in order to solve the problems existing in the prior art, namely, the problem that the fault diagnosis precision of an SVM is affected due to the fact that the feature extraction capability of an early fault diagnosis model of an analog circuit is insufficient and a particle swarm algorithm is prone to fall into local optimization in a complex optimization problem, embodiment 1 of the present invention provides an analog circuit fault diagnosis method based on feature fusion and improved particle swarm algorithm, which comprises the following processes:
acquiring fault one-dimensional time domain data of an analog circuit, and converting the one-dimensional time domain data into two-dimensional time frequency image data;
obtaining a first characteristic according to the one-dimensional time domain data and the one-dimensional convolutional neural network; obtaining a second characteristic according to the two-dimensional time-frequency image data and the depth residual error network; splicing the first characteristic and the second characteristic to obtain a final fault characteristic;
and obtaining the fault type according to the fault characteristics and a pre-trained support vector machine.
Specifically, the method comprises the following steps:
s1: collecting fault data and constructing an original fault data set: firstly, applying an excitation signal to an input end of a circuit to be tested; then, collecting the voltage signal of the circuit output end as an original data sample, and setting the original fault data acquired in the fault state k
Wherein the content of the first and second substances,
the ith sample acquired in the fault state k; finally, the circuit data samples (including the data samples in the normal state) in different circuit states are sorted to form an original data set.
S2: original time domain data is converted into time-frequency image data by using cross wavelet transformation to increase the information content of the original data, and the main formula of the cross wavelet transformation is as follows:
assuming that x (t) and y (t) are two time series, the cross wavelet transform between the two can be expressed as:
CWT(a,τ)＝WT x (a,τ)WT y * (a,τ) (1)
in the formula, WT x (a, τ) and WT y (a, τ) are the continuous wavelet transforms of x (t) and y (t), respectively, representing complex conjugates.
Selecting circuit output data of which all elements are at a nominal value as a reference signal, and selecting sample data of other states of the circuit as another signal; the original one-dimensional time domain data is converted into time-frequency image data by performing cross wavelet transformation on circuit output signals in different states and reference signals, so that a time-frequency image data set is formed.
S3: performing feature extraction by using a deep feature fusion network based on a one-dimensional convolutional neural network (1 DCNN) and a deep residual error network (ResNet): firstly, inputting original one-dimensional time domain data into a one-dimensional convolution neural network, wherein the detailed parameter setting of the neural network is shown in table 1:
table 1:1DCNN parameter settings
Wherein, the convolution kernel with the size of 7 x 1 is selected as the first layer convolution layer so as to obtain more comprehensive characteristic information. And after the model is trained, outputting the extracted features of the first full connection layer for subsequent feature fusion. Then, the two-dimensional time-frequency image data is input into a depth residual error network, the residual error network selected in this embodiment is ResNet18, and the parameter settings are shown in table 2. And outputting the extracted features of the ResNet full connection layer as 1 DCNN. And finally, splicing the features extracted by the 1DCNN and the ResNet, and outputting the feature as a final feature fault data set.
Table 2: resNet18 network parameter setting
S4: and (3) dividing the feature fault data set extracted by the deep feature fusion network into a training set and a testing set according to the ratio of 7.
S5: the flow chart of inputting the training set into the support vector machine and optimizing the penalty parameter and the kernel parameter of the SVM by using the improved particle swarm optimization algorithm is shown in FIG. 1.
S6: setting initial parameters of the particle swarm, e.g. maximum number of iterations t max Number of particles N, dimension D of search space, and range of search space [ X ] min ,X max ]Range of particle velocity [ V ] min ,V max ]And maximum and minimum inertia weights ω min And omega max 。
S7: initializing the population, wherein an initialization formula is as follows:
in the formula (I), the compound is shown in the specification,
is a value of d-th dimension of the ith particle, and Rand is a random number uniformly distributed following the U (0, 1) standard.
S8: and (5) starting iteration, calculating the fitness value of each particle and selecting the comprehensive optimal particle FDBbest. The selection of FDBbest depends not only on the fitness value but also on the distance between each particle and Gbest (the particle with the best population fitness), and the calculation formula of the distance between the ith particle and Gbest is as follows:
in order to select FDBbest, a scoring function is constructed based on fitness and distance-based optimal particle selection strategy (FDB). The argument of this function is DG i And a fitness value F i 。
First, DG is added i And F i Normalization is performed according to equations (4) and (5):
wherein DG and F are respectively DG of all particles in the current iteration i And F i A collection of (a).
Then, the integrated fraction FDBcore of each particle is calculated according to the formula (7) i . And finally, sorting the particles with the highest comprehensive score according to the order of the scores of the particles from high to low to select the particles with the highest comprehensive score as FDBbest.
FDBscore i ＝α(1-normF i )+(1-α)normDG i (7)
In the formula, t max Is the maximum number of iterations, alpha is the balance factor, a smaller alpha indicates norm DG i For FDBscore i The larger the influence of (A), the more beneficial to the global exploration of the population; a larger alpha indicates norm F i For FDBscore i The larger the influence of (b), the more favorable the local development of the population.
FDBbest selected by the strategy replaces Gbest to guide the particles to finish speed updating. The new velocity update formula is as follows:
wherein, c 1 And c 2 Is an empirical weight, r 1 And r 2 Is a discount factor.
S9: updating Pbest and Gbest if the fitness value of a certain particle is better than that of Pbest or GbestGbest, then the position X of the particle is used i Replacing Pbest or Gbest.
S10: and performing cross operation on the FDBbest and the particles with poor adaptability, wherein the strategy mainly adopts the idea of an elite strategy, and the overall searching efficiency of the population is improved by enhancing the information exchange between the particles with poor performance and the FDBbest. The update formula of the strategy is as follows:
in the formula, r 3 Are random numbers uniformly distributed following the U (0, 1) standard; n is a radical of hydrogen c Is the number of particles undergoing crossover operation.
S11: and calculating the inertia weight of the iteration, wherein the formula is as follows.
In the formula, ω max ＝0.9、ω min =0.2, the maximum and minimum of the inertial weight, respectively; p =0.25.
S12: the FDB strategy and the conventional speed update strategy are used alternately: when each particle is updated in speed, the particle is randomly updated according to the formula (8) or the basic speed updating formula (11), wherein the basic speed updating formula is as follows:
wherein, c 1 And c 2 Is an empirical weight, r 1 And r 2 Is a discount factor.
The location update is performed according to equation (12), which is as follows:
considering particle swarm earlier stagePaying attention to the characteristics of global exploration and local development in the later stage, and introducing a probability factor P when a particle selection speed updating mode is adopted f The value of which gradually decreases linearly from 0.5 to 0 as the number of iterations increases. If the random number Rand is less than P f Then the particle is updated according to equation (8); otherwise, the speed update is performed according to equation (11). The updating mode can not only reduce the probability of the population falling into the local optimum, but also keep the characteristic of the particle swarm algorithm of fast convergence.
S13: and each particle develops a surrounding solution space according to the current position of the particle, and the particle with good adaptability has higher development efficiency according to the characteristics of the particle swarm. Therefore, in the particle individual self-development (ISE) strategy, each particle has a different probability of developing its own surrounding space according to the fitness value of the particle.
First, N random numbers uniformly distributed according to the U (0, 1) standard are generated. Secondly, N random numbers rho are added i (i =1,2.., N) is ordered from large to small. Meanwhile, aiming at the minimization problem, the N particles are sorted from small to large according to the fitness value. Then, the N random numbers correspond to the N particles one by one, so that the particles with small fitness value and the large random numbers are paired, and the particles with large fitness value and the small random numbers are paired. Finally, in each iteration, if the particle corresponds to ρ i Greater than a threshold value P s If so, ISE operation is carried out on the particle; otherwise, ISE operation is not carried out.
P s Is defined as follows:
it is considered that the development of particles is getting finer and finer in the later stages of the iteration. Therefore, the range of the particle pair around itself should be gradually reduced with the number of iterations.
Will develop a range coefficient C id The definition is as follows:
in the formula, C id Is the development range of the d-th dimension of the i-th particle.
In this iteration, all ρ i Greater than the current threshold P s The particles (2) are developed in the region around the particles (1) according to the formula (15).
In the formula, norm dC id Means that a obey N (0, C) is randomly generated id ) Normally distributed random numbers.
As can be seen from the above, in the ISE policy proposed in this section, the more suitable the particles are, the more likely the particles are to develop their own surrounding area. In the early stage of iteration, P s The value of (A) is small, the number of particles capable of carrying out ISE operation is large, the exploration range is also large, and the exploration capability of the population is strong; at the end of the iteration, P s The value of (A) is larger, only a few particles with better fitness can be developed in a small range by using an ISE strategy, and the development capability of the population is enhanced.
If probability value ρ of ith particle i >P s If so, the particle uses the ISE strategy to develop the surrounding space of the particle; otherwise, the development is not carried out, and the S9 is returned.
S14: judging whether the maximum iteration times is reached, if so, stopping iteration and finishing model training; otherwise, return to S9.
S5: and after the model training is finished, inputting the test set into the trained SVM to finish the test of the fault diagnosis model.
The test circuits used in this embodiment are a Chebyshev (Chebyshev) filter circuit and a Four-operational amplifier biquad two-order high-pass (Four-op-amp) filter circuit, respectively. In both circuits, the tolerance of the resistor is set to 5% of the nominal value and the tolerance of the capacitor is set to 10% of the nominal value. The parameter values of the elements are all considered to be normal within normal tolerances, and the circuit state within 30% of the nominal value is considered to be an early failure. The schematic circuit diagrams of the Chebyshev filter circuit and the Fourier-op-amp filter circuit are shown in FIGS. 2 and 3, respectively, and the original feature distribution diagram is shown in FIG. 4. A feature profile for feature extraction of raw fault data using a deep feature fusion network is shown in fig. 5. As can be seen from the figure, the features extracted by the deep feature fusion network can basically separate different types of faults, have better intra-class cohesion and inter-class distance and show better distinguishability, which indicates that the deep feature fusion network can excavate deeper fault features with better distinguishability.
An improved particle swarm algorithm is named as HPSO-FDB-ISE in the experiment, and in order to verify the performance of each strategy in the HPSO-FDB-ISE, a CEC2017 benchmark test set proposed by the P.N. Suganthan team is adopted for carrying out a comparison experiment. The convergence curves of HPSO-FDB-ISE and its variants on the basis functions F1, F3, F4, F11-F14, F18-F20, F28, F30 are shown in FIG. 6. As can be seen from (b), (g) and (j) in fig. 6, although the PSO converges faster than the PSO-FDB in the early stage, it falls into local optima in the later stage. The PSO-FDB avoids falling into local optimum to a certain extent due to the introduction of an FDB strategy, so that a better optimization result is obtained, but the PSO-FDB development capability is slightly insufficient. In the whole iterative process, the adaptability value of the PSO-FDB-ISE is superior to that of the PSO-FDB, and the difference between the adaptability value and the PSO-FDB-ISE is continuously expanded, which shows that the ISE strategy plays a positive role in the local development process of the PSO-FDB-ISE.
In each subgraph of FIG. 6, the accuracy and convergence rate of HPSO-FDB-ISE are all stronger than those of the PSO, PSO-FDB and PSO-FDB-ISE algorithms. Although the PSO-FDB-ISE performed better than the HPSO-FDB-ISE in the middle of the iteration in (a), (e), (l) of FIG. 6, the PSO-FDB-ISE failed to balance the exploration and development capabilities of the population in this period, thereby causing the population to fall into local optima in the later period of the iteration. The HPSO-FDB-ISE uses the nonlinear time-varying inertia weight, so that the global exploration and local development capability of the population is better balanced in the process, the population is prevented from falling into local optimum, the fitness value is continuously reduced, and the optimization capability of the population is improved.
In conclusion, the optimal particle selection strategy with the repeated fitness and distance improves the global exploration capacity of the population, reduces the probability that the population is trapped in local optimal, the particle individual self-development strategy improves the local development capacity of the population, and the nonlinear time-varying inertia weight well balances the global and local search capacities of the population. Under the action of the three, the comprehensive performance of the HPSO-FDB-ISE algorithm is obviously improved.
As shown in fig. 7, the deep feature fusion network and the SVM optimized by the HPSO-FDB-ISE algorithm are integrated to form a complete intelligent fault diagnosis model of the analog circuit, and the performance of the model is verified on the simulation circuit and the actual circuit. In the simulation circuit, the accuracy rates of the SVM, the PSO-SVM and the HPSO-FDB-ISE-SVM on the Chebyshev filter circuit and the Fourier-op-amp filter circuit are shown in fig. 8, and as can be seen from fig. 8, the accuracy rates of the provided fault diagnosis model on the Chebyshev filter circuit and the Fourier-op-amp filter circuit respectively reach 100% and 99.70%. This shows that the SVM can accurately classify fault samples that are overlapped together through optimization of the HPSO-FDB-ISE algorithm. The convergence curve graph of the PSO and the HPSO-FDB-ISE in the SVM optimization process is shown in fig. 9, where the optimal fitness in the graph is the fitness of the globally optimal particle, and the average fitness is the average of the fitness of all particles in the current iteration. As can be seen from (a) in fig. 9, the fitness value of the PSO algorithm falls into local optimum when reaching around 98%; as can be seen from the average fitness, in the early stage of iteration, the population diversity of the original PSO algorithm is lost too fast, and the optimization process of all particles is stopped. As can be seen from (b) in fig. 9, although the HPSO-FDB-ISE algorithm falls into local optima in the pre-iteration period, it quickly jumps out of the local optima, thereby achieving higher fault diagnosis accuracy; according to the average fitness, the whole particle swarm exploration process is very active in the early and middle stages of iteration, and in the later stage of iteration, the fitness of all particles gradually gets close to the global optimum value in order to accelerate convergence.
The whole process of the intelligent fault diagnosis model applied to the actual circuit is as follows: taking the Chebyshev filter circuit as an example, the original data acquisition device comprises three parts, namely a signal generator, a circuit to be tested and an oscilloscope. Firstly, the signal generator sends out a pulse voltage with the amplitude of 10V, and the pulse voltage is input into a circuit to be tested. And then, a voltage signal at the output end of the circuit to be tested is led into an oscilloscope, and the oscilloscope in the figure displays an output waveform of the Chebyshev filter circuit. And finally, exporting the original data information from the oscilloscope into a csv file, and integrating to form an original data set.
After an original data set of an actual circuit is obtained, feature extraction is firstly carried out on the original data set by using a deep feature fusion network. Then, the extracted fault feature data set is divided into a training set and a test set. The training set is used for training the SVM model, and the penalty parameter and the kernel parameter of the SVM are optimized by using an HPSO-FDB-ISE algorithm. And finally, inputting the test set into the trained model to carry out fault diagnosis and outputting a diagnosis result.
The distribution of the original data set and the fault feature data set extracted by the deep feature fusion network of the two actual circuits in the three-dimensional space is shown in fig. 10 and fig. 11, respectively. As can be seen, the original faults are mixed and dispersed. After feature extraction is performed through a deep feature fusion network, distinctiveness among fault features is obvious, but intra-class cohesion is general, and possible reasons are as follows: firstly, the parameter distribution of the fault element is not uniform during data acquisition, and secondly, some noise interference (such as noise introduced by connecting lines or joints) exists in the actual circuit.
The feature data set extracted by the deep feature fusion network is input into an HPSO-FDB-ISE optimized SVM for fault diagnosis, and the diagnosis accuracy of the two circuits is shown in Table 3. As can be seen from the table, the intelligent fault diagnosis model for the analog circuit provided herein maintains high fault diagnosis accuracy on the Chebyshev filter circuit with a relatively simple network structure. Although the diagnosis precision of the HPSO-FDB-ISE-SVM on the actual Fourier-op-amp filter circuit is reduced by about 3% compared with a simulation circuit, the intelligent fault diagnosis model still has better implementability and generalization capability on the whole.
Table 3: intelligent fault diagnosis model diagnosis accuracy statistical table (%)
Example 2:
the embodiment 2 of the present invention provides an analog circuit fault diagnosis system based on feature fusion and improved particle swarm optimization, including:
a data acquisition module configured to: acquiring fault one-dimensional time domain data of an analog circuit, and converting the one-dimensional time domain data into two-dimensional time frequency image data;
a feature extraction module configured to: obtaining a first characteristic according to the one-dimensional time domain data and the one-dimensional convolutional neural network; obtaining a second characteristic according to the two-dimensional time-frequency image data and the depth residual error network; splicing the first characteristic and the second characteristic to obtain a final fault characteristic;
a fault diagnosis module configured to: obtaining a fault type according to the fault characteristics and a pre-trained support vector machine;
the improved particle swarm optimization method comprises the steps of optimizing punishment parameters and nuclear parameters of a support vector machine by adopting an improved particle swarm optimization, calculating comprehensive scores of all particles according to balance factors, distances between the particles and particles with the optimal population fitness and fitness values in the improved particle swarm optimization, and selecting the particles with the highest comprehensive scores as the comprehensive optimal particles according to the ranking of the scores of the particles from high to low.
The working method of the system is the same as the method for diagnosing the fault of the analog circuit based on the feature fusion and improved particle swarm optimization algorithm provided in the embodiment 1, and details are not repeated here.
Example 3:
embodiment 3 of the present invention provides a computer-readable storage medium on which a program is stored, which when executed by a processor, implements the steps in the analog circuit fault diagnosis method based on feature fusion and improved particle swarm optimization as described in embodiment 1 of the present invention.
Example 4:
embodiment 4 of the present invention provides an electronic device, which includes a memory, a processor, and a program stored in the memory and executable on the processor, and when the processor executes the program, the steps in the analog circuit fault diagnosis method based on feature fusion and improved particle swarm optimization described in embodiment 1 of the present invention are implemented.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above may be implemented by a computer program, which may be stored in a computer readable storage medium and executed by a computer to implement the processes of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A method for diagnosing faults of an analog circuit based on feature fusion and improved particle swarm optimization is characterized by comprising the following steps:
the method comprises the following steps:
acquiring fault one-dimensional time domain data of an analog circuit, and converting the one-dimensional time domain data into two-dimensional time frequency image data;
obtaining a first characteristic according to the one-dimensional time domain data and the one-dimensional convolution neural network; obtaining a second characteristic according to the two-dimensional time-frequency image data and the depth residual error network; splicing the first characteristic and the second characteristic to obtain a final fault characteristic;
obtaining a fault type according to the fault characteristics and a pre-trained support vector machine;
the improved particle swarm optimization method comprises the steps of optimizing punishment parameters and nuclear parameters of a support vector machine by adopting an improved particle swarm optimization, calculating comprehensive scores of all particles according to balance factors, distances between the particles and particles with the optimal population fitness and fitness values in the improved particle swarm optimization, and selecting the particles with the highest comprehensive scores as the comprehensive optimal particles according to the ranking of the scores of the particles from high to low.
2. The method for diagnosing faults of an analog circuit based on feature fusion and improved particle swarm optimization as claimed in claim 1, wherein:
and converting the one-dimensional time domain data into two-dimensional time-frequency image data by using cross wavelet transformation.
3. The method for diagnosing faults of an analog circuit based on feature fusion and improved particle swarm optimization as claimed in claim 1, wherein:
the distance between the ith particle and the optimal population fitness particle Gbest is as follows:
4. the method for diagnosing faults of an analog circuit based on feature fusion and improved particle swarm optimization as claimed in claim 1, wherein:
calculating the comprehensive score of each particle according to the balance factor, the distance between the particle and the particle with the optimal population fitness and the fitness value, wherein the calculation comprises the following steps:
FDBscore i ＝α(1-normF i )+(1-α)normDG i
wherein α is a balance factor, norm F i For normalization of the fitness value, DG i And obtaining a normalized value of the distance between the ith particle and the particle Gbest with the optimal population fitness.
5. The method for diagnosing faults of an analog circuit based on feature fusion and improved particle swarm optimization as claimed in claim 4, wherein:
wherein, t max Is the maximum number of iterationsAnd t is the current iteration number.
6. The method for diagnosing faults of an analog circuit based on feature fusion and improved particle swarm optimization as claimed in claim 1, wherein:
in the improved particle swarm optimization, the method further comprises the following steps:
replacing the optimal population fitness particle Gtest by the comprehensive optimal particle FDBbest, and replacing a second speed formula by a first speed formula to update the particle speed;
wherein the first velocity formula is:
the second velocity formula is:
c 1 and c 2 Is an empirical weight, r 1 And r 2 Is a discount factor;
performing cross operation on the comprehensive optimal particle FDBbest and the particles with poor adaptability;
introducing a probability factor, wherein the value of the probability factor is gradually linearly reduced from 0.5 to 0 along with the increase of the iteration times, and if the random number Rand is smaller than the probability factor, the particle carries out speed updating according to a first speed formula; otherwise, the speed is updated according to the second speed formula.
7. The method for diagnosing faults of an analog circuit based on feature fusion and improved particle swarm optimization as claimed in claim 1, wherein:
each particle develops a surrounding solution space according to the current position of the particle, and the method comprises the following steps:
generating N random numbers uniformly distributed according to a U (0, 1) standard;
n random numbers rho i (i =1, 2...., N) are ordered from large to small, while the N particles are ordered from small to large according to the fitness value;
the N random numbers correspond to the N particles one by one, so that the particles with small fitness value are paired with the large random numbers, and the particles with large fitness value are paired with the small random numbers;
in each iteration, if the particle corresponds to ρ i Greater than a threshold value P s Then the particle carries out particle individual self-development operation; otherwise, not carrying out individual self-development operation of the particles;
all rho i Greater than the current threshold P s The particles of (2) are developed into their own surrounding area according to the following formula:
wherein norrndC id Means that a obedient N (0, C) is randomly generated id ) Normally distributed random number, C id The development range of the d-th dimension of the i-th particle.
8. An analog circuit fault diagnosis system based on feature fusion and improved particle swarm optimization is characterized in that:
the method comprises the following steps:
a data acquisition module configured to: acquiring fault one-dimensional time domain data of an analog circuit, and converting the one-dimensional time domain data into two-dimensional time frequency image data;
a feature extraction module configured to: obtaining a first characteristic according to the one-dimensional time domain data and the one-dimensional convolutional neural network; obtaining a second characteristic according to the two-dimensional time-frequency image data and the depth residual error network; splicing the first characteristic and the second characteristic to obtain a final fault characteristic;
a fault diagnosis module configured to: obtaining a fault type according to the fault characteristics and a pre-trained support vector machine;
the improved particle swarm optimization method comprises the steps of optimizing punishment parameters and nuclear parameters of a support vector machine by adopting an improved particle swarm optimization, calculating comprehensive scores of all particles according to balance factors, distances between the particles and particles with the optimal population fitness and fitness values in the improved particle swarm optimization, and selecting the particles with the highest comprehensive scores as the comprehensive optimal particles according to the ranking of the scores of the particles from high to low.
9. A computer-readable storage medium on which a program is stored, the program, when being executed by a processor, implementing the steps in the method for diagnosing faults in an analog circuit based on feature fusion and improved particle swarm optimization as claimed in any one of claims 1 to 7.
10. An electronic device comprising a memory, a processor and a program stored on the memory and executable on the processor, wherein the processor implements the steps of the method for diagnosing faults of analog circuits based on feature fusion and improved particle swarm optimization as claimed in any one of claims 1 to 7 when executing the program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210515698.5A CN115186564A (en) | 2022-05-12 | 2022-05-12 | Analog circuit fault diagnosis method based on feature fusion and improved particle swarm algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210515698.5A CN115186564A (en) | 2022-05-12 | 2022-05-12 | Analog circuit fault diagnosis method based on feature fusion and improved particle swarm algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115186564A true CN115186564A (en) | 2022-10-14 |
Family
ID=83513720
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210515698.5A Pending CN115186564A (en) | 2022-05-12 | 2022-05-12 | Analog circuit fault diagnosis method based on feature fusion and improved particle swarm algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115186564A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114355173A (en) * | 2022-01-04 | 2022-04-15 | 电子科技大学 | Analog filter circuit fault diagnosis method based on multi-input residual error network |
| CN117991082A (en) * | 2024-04-07 | 2024-05-07 | 垣矽技术(青岛)有限公司 | Fault diagnosis supervision system suitable for current frequency conversion chip |
-
2022
- 2022-05-12 CN CN202210515698.5A patent/CN115186564A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114355173A (en) * | 2022-01-04 | 2022-04-15 | 电子科技大学 | Analog filter circuit fault diagnosis method based on multi-input residual error network |
| CN114355173B (en) * | 2022-01-04 | 2023-05-30 | 电子科技大学 | Analog filter circuit fault diagnosis method based on multi-input residual error network |
| CN117991082A (en) * | 2024-04-07 | 2024-05-07 | 垣矽技术(青岛)有限公司 | Fault diagnosis supervision system suitable for current frequency conversion chip |
| CN117991082B (en) * | 2024-04-07 | 2024-06-11 | 垣矽技术(青岛)有限公司 | Fault diagnosis supervision system suitable for current frequency conversion chip |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108303253B (en) | Bearing initial failure recognition methods based on long short-term memory Recognition with Recurrent Neural Network | |
| CN115186564A (en) | Analog circuit fault diagnosis method based on feature fusion and improved particle swarm algorithm | |
| CN109000930B (en) | Turbine engine performance degradation evaluation method based on stacking denoising autoencoder | |
| DE202017106532U1 (en) | Search for a neural architecture | |
| CN110516305B (en) | Intelligent fault diagnosis method under small sample based on attention mechanism meta-learning model | |
| CN110334580A (en) | The equipment fault classification method of changeable weight combination based on integrated increment | |
| KR101964412B1 (en) | Method for diagnosing anomaly log of mobile commmunication data processing system and system thereof | |
| CN110427893A (en) | A kind of specific emitter identification method, apparatus and computer storage medium based on convolutional neural networks | |
| CN108051660A (en) | A kind of transformer fault combined diagnosis method for establishing model and diagnostic method | |
| CN108106844A (en) | A kind of Method for Bearing Fault Diagnosis of the automatic coding machine of self-tuning parameter adjustment | |
| CN105572572A (en) | WKNN-LSSVM-based analog circuit fault diagnosis method | |
| CN110879351A (en) | Fault diagnosis method for non-linear analog circuit based on RCCA-SVM | |
| CN110568483A (en) | Automatic evaluation method for seismic linear noise suppression effect based on convolutional neural network | |
| CN110543832A (en) | Electroencephalogram data classification method based on random forest and convolutional neural network | |
| CN115587543A (en) | Federal learning and LSTM-based tool residual life prediction method and system | |
| CN111539612B (en) | Training method and system of risk classification model | |
| CN110782041B (en) | Structural modal parameter identification method based on machine learning | |
| CN111767277A (en) | Data processing method and device | |
| CN116561517A (en) | Unbalanced intelligent fault quantitative diagnosis method based on reward optimization deep reinforcement learning | |
| CN115130514A (en) | Construction method and system for health index of engineering equipment | |
| CN112380932B (en) | Vibration signal characteristic value selection method and elevator health state evaluation or fault diagnosis method | |
| CN114330486A (en) | Power system bad data identification method based on improved Wasserstein GAN | |
| CN103714385B (en) | Analogue circuit fault diagnosis method based on improved type clone selection algorithm | |
| CN110347579B (en) | Deep learning test case selection method based on neuron output behavior pattern | |
| CN112146880A (en) | Intelligent diagnosis method for internal structure faults of rolling bearing at different rotating speeds |
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 |