CN110673017B - Analog circuit fault element parameter identification method based on genetic algorithm - Google Patents

Abstract

The invention discloses a method for identifying fault components of an analog circuit based on a genetic algorithm. First, the transfer functions of the analog circuit at different measuring points are obtained by analysis, and the output voltages of the analog circuit at the measuring points under a preset excitation signal are obtained by measuring, The element parameter value vector is used as the individual in the genetic algorithm. When the individual after crossover and mutation is optimized, the population individuals are firstly grouped according to the failure type, and the optimal individual of each failure type is selected to join the next generation population, and then from The optimal individual among the remaining individuals is added to the next generation population, and the parameter value of the optimal individual in the last generation population is used as the element parameter identification result. The invention realizes the identification of the fault element parameters of the single fault and double fault of the analog circuit through the genetic algorithm.

Description

A Genetic Algorithm-Based Method for Parameter Identification of Fault Component in Analog Circuits

technical field

The invention belongs to the technical field of fault diagnosis of analog circuits, and more particularly relates to a genetic algorithm-based identification of fault components of analog circuits.

Background technique

With the high integration of circuits, the area of the chip is constantly decreasing, and the number of analog components integrated on the chip is increasing, and the diagnostic cost of the analog part in the chip remains high. The reason lies in the rapid development of integrated circuits, while the corresponding analog circuit fault diagnosis technology is stagnant. With the development of machine learning and artificial intelligence, theoretical methods for fault diagnosis of analog circuits based on machine learning and artificial intelligence have sprung up. For example, analog circuit fault diagnosis methods based on neural network, SVM, etc. combined with wavelet transform, measurement point selection, etc., need to extract a large number of fault features to train neural networks, support vector machines, etc. How to extract effective fault features in the case of tolerance is the problem to be solved by this kind of method.

SUMMARY OF THE INVENTION

The purpose of the present invention is to overcome the deficiencies of the prior art, and to provide a genetic algorithm-based method for identifying the parameters of faulty components of an analog circuit.

In order to achieve the above-mentioned purpose of the invention, the method for identifying the parameters of faulty components of an analog circuit based on a genetic algorithm of the present invention comprises the following steps:

S1: Obtain the transfer function of the analog circuit at D measuring points t _d , d=1,2,...,D;

S2: Carry out the fuzzy group analysis of fault diagnosis through the measurement point t in the analog circuit, select a representative fault element for each fuzzy group, record the number of representative fault elements as N, and record the number of other non-representative fault elements is M;

S3: Setting no fault means that all faulty components in the analog circuit are not faulty, single fault means that only one representative faulty component fails when the analog circuit fails, and double faults means that there are two representative faulty components when the analog circuit fails If faults occur at the same time, record the number of fault types

分别表示输出电压

的实部和虚部，j为虚数单位；S4: When it is necessary to identify the fault parameters of the analog circuit, measure the output voltages at D measuring points t _d under the preset excitation signal

Respectively represent the output voltage

The real and imaginary parts of , j is the imaginary unit;

S5: Take X={x ₁ ,...,x _N ,x' ₁ ,...,x' _M } as the individuals in the genetic algorithm, where x _n represents the parameter value of the nth representative fault element, n=1, _{2 ,} . ,2,...,K, in each individual in the classification group p _k , take the parameter value x _n of the representative fault element corresponding to the kth fault type within the fault range of the representative fault element, and other The parameter value of the fault element is within the tolerance range; then the K subgroups p _k are merged to form a population P, and the number of individuals in the population P is recorded as G;

S6: judge whether the iteration end condition of the genetic algorithm is reached, if yes, go to step S11, otherwise go to step S7;

S7: Perform crossover and mutation operations on population P to obtain subpopulation Q; when performing crossover and mutation operations, it is necessary to ensure that the parameter values of at most two representative fault elements in each individual in subpopulation Q are located in the representative fault element The fault range of the other faulty components is within its tolerance range;

S8: Combine population P and population Q to form population S, that is, S=P∪Q;

S9: Substitute each individual in the population S into the transfer function to obtain the output voltages U _g,d =α _g,d +jβ _g,d ,α _g, at D measuring points t _d under the preset excitation signal _d and β _g,d represent the real part and imaginary part of the output voltage U _g, d respectively, g=1,2,...,2G, and then use the following formula to calculate the difference between the output voltage of the gth individual and the output voltage of the current analog circuit. The Euclidean distance D _g between :

S10: Divide the individuals in the population S into K subgroups _sk according to their corresponding fault types, screen out the individual with the smallest Euclidean distance from each subgroup _sk , add it to the next generation population P', and add it from Delete from the population S to obtain the population S'; then select GK individuals from the population S' and add the next generation population P'; then set the population P=P', and return to step S6;

S11: Select the individual with the smallest Euclidean distance from the current population. The representative fault element whose parameter value is within the fault range in the individual is the fault diagnosis result, and the corresponding parameter value is the parameter identification result of the fault element.

The invention is based on the genetic algorithm-based method for identifying the parameters of faulty components of an analog circuit. First, the transfer functions of the analog circuit at different measuring points are obtained by analysis, and the output voltages of the analog circuit at these measuring points under the preset excitation signal are measured and obtained, and the parameter values of the components are calculated. The vector is used as the individual in the genetic algorithm. When the individuals after crossover and mutation are optimized, the population individuals are firstly grouped according to the failure type, and the optimal individual of each failure type is selected to join the next generation population, and then the remaining individuals are selected. Individuals are added to the next generation of populations, and the parameter values of the optimal individuals in the last generation of populations are used as the result of element parameter identification.

By using the genetic algorithm, the present invention can effectively avoid the extraction of fault features and avoid a lot of simulation work. Parameter identification of faulty components in single-fault and double-fault analog circuits.

Description of drawings

Fig. 1 is the specific implementation flow chart of the method for identifying the parameters of analog circuit fault components based on genetic algorithm of the present invention;

FIG. 2 is a topology diagram of the second-order Thomas analog filter circuit of this embodiment.

Detailed ways

The specific embodiments of the present invention are described below with reference to the accompanying drawings, so that those skilled in the art can better understand the present invention. It should be noted that, in the following description, when the detailed description of known functions and designs may dilute the main content of the present invention, these descriptions will be omitted here.

FIG. 1 is a flow chart of a specific embodiment of a method for identifying parameters of faulty components in an analog circuit based on a genetic algorithm of the present invention. As shown in Figure 1, the specific steps of the genetic algorithm-based analog circuit fault component parameter identification method include:

S101: Get the transfer function:

Obtain the transfer function of the analog circuit at D measuring points t _d , d=1,2,...,D.

Since the present invention is aimed at single fault and double fault, and the double fault model in which two different components fail at the same time has aliasing on the two-dimensional plane and cannot be isolated, the number of measuring points is preferably more than 2, so that Troubleshooting.

S102: Fuzzy group analysis:

For the fuzzy group analysis of fault diagnosis through the measurement point t in the analog circuit, select a representative fault element for each fuzzy group, and record the number of representative fault elements as N. Obviously, N also represents the number of fuzzy groups. The number of representative faulty elements is M.

S103: Determine the fault type:

In the present invention, single fault and double fault are considered, that is, setting no fault means that all faulty components in the analog circuit are not faulty, single fault means that only one representative faulty component fails when the analog circuit fails, and double faults means that the analog circuit occurs. At the time of failure, there are two representative failure components that fail at the same time, so the number of failure types can be recorded

S104: Determine the current output of the analog circuit:

分别表示输出电压

的实部和虚部，j为虚数单位。为了使故障状态下的输出电压更加准确，可以多次测量各个测点处的输出电压后进行平均，将平均值作为输出电压

When it is necessary to identify the fault parameters of the analog circuit, the output voltages at the D measuring points t _d are measured under the preset excitation signal

Respectively represent the output voltage

The real and imaginary parts of , and j is the imaginary unit. In order to make the output voltage in the fault state more accurate, the output voltage at each measuring point can be measured multiple times and averaged, and the average value is taken as the output voltage

S105: Initialize the genetic algorithm population:

Take X={x ₁ ,...,x _N ,x' ₁ ,...,x' _M } as the individuals in the genetic algorithm, where x _n represents the parameter value of the nth representative fault element, n=1, 2, ...,N,x' _m denotes the parameter value of the mth non-representative fault element, m=1,2,...,M. For each fault type, an initial classification group p _k is generated, k=1,2,...,K, in each individual in the classification group p _k , the representative fault element corresponding to the kth fault type is The parameter value x _n takes values within the fault range of the representative faulty element, and the parameter values of other faulty elements (ie, other representative faulty elements and M non-representative faulty elements) take values within the tolerance range. Then the K subgroups p _k are merged to form a population P, and the number of individuals in the population P is recorded as G. In general, in order to make the diagnosis probability of each fault type equal at the beginning, the number of individuals in each subgroup _pk is the same when the population is initialized.

S106: Determine whether the iteration end condition of the genetic algorithm is reached, if yes, go to step S110, otherwise go to step S107. There are generally two conditions for the end of iteration of the genetic algorithm, one is to reach the maximum number of iterations, and the other is that the value of the objective function reaches a preset threshold, which can be set according to actual needs.

S107: Generate subpopulations:

Crossover and mutation operations are performed on population P to obtain subpopulation Q. When performing crossover and mutation operations, it is necessary to ensure that the parameter values of at most two representative faulty elements in each individual in the subgroup Q are within the fault range of the representative faulty element, and the parameter values of other faulty elements are within its tolerance range. In this embodiment, the crossover operation adopts simulated binary crossover (SBX), and the mutation operation is polynomial mutation (POL). At the same time, the individual gene value is limited to vary from 0 to positive infinity in the crossover and mutation, so as to prevent the individual gene value from being negative. After the crossover and mutation are completed, if the number of representative faulty elements in the individual whose parameter values are within the fault range exceeds two, the parameter values of the exceeded representative faulty elements are randomly limited back to the tolerance range.

S108: Combined population:

The population P and the population Q are combined to form the population S, that is, S=P∪Q. Obviously, the number of individuals in the combined population is 2G.

S109: Calculate the individual objective function value:

Next, the objective function value needs to be calculated for each individual in the population S. For the present invention, the Euclidean distance between the output voltage obtained by each individual under different frequency excitation signals and the output voltage of the current analog circuit is used. As the objective function, the specific calculation method is as follows:

Substitute each individual in the population S into the transfer function to obtain the output voltage U _{g,d =} α _g,d +jβ _g,d , α _g,d , β _g,d represent the real part and imaginary part of the output voltage U _g, d respectively, g=1,2,…,2G, and then use the following formula to calculate the difference between the gth individual output voltage and the output voltage of the current analog circuit: Euclidean distance D _g :

Obviously, in terms of fault diagnosis, the smaller the Euclidean distance, the closer the output voltage is to the current analog circuit and output voltage, and the better the individual is.

S110: Generate the next generation population:

The individuals in the population S are divided into K subgroups _sk according to their corresponding fault types. According to the specific methods of population initialization and crossover and mutation operations, there will be at most two representative fault elements in each individual whose parameter values are located in the fault range, and the parameter values of other representative fault elements and non-representative fault elements are located in the fault range. Within the tolerance range, subgroups can be divided according to this. Select the individual with the smallest Euclidean distance from each subgroup _sk , add it to the next generation population P', and delete it from the population S to obtain the population S'. Then, GK individuals are selected from the population S' and added to the next-generation population P', so that the number of individuals in the next-generation population P' is G. Then set the population P=P', and return to step S106.

个优选个体，

表示向下取整，然后将这些优选个体按照欧式距离进行升序排序，选择前G-K个个体加入下一代种群P′。In this embodiment, when performing individual optimization on the population S', the tournament optimization method is used to perform individual optimization on the current population S, and obtain

a preferred individual,

Represents rounded down, and then sorts these preferred individuals in ascending order according to Euclidean distance, and selects the first GK individuals to join the next generation population P'.

According to the above process, the individuals in the next generation population in the present invention are obtained in two ways. First, the population is grouped and optimized to ensure that at least one individual exists in each fault type after the iteration, and then the remaining population is optimized and supplemented total group number.

S111: Determine the fault parameter identification result:

The individual with the smallest Euclidean distance is selected from the current population, and the representative fault element whose parameter value is within the fault range is the fault diagnosis result, and the corresponding parameter value is the fault parameter identification result. Obviously, if the analog circuit does not fail, the parameters of all representative failed components are within tolerance.

Example

In order to better illustrate the technical solution of the present invention, the present invention is described in detail by taking the second-order Thomas analog filter circuit as an example. FIG. 2 is a topology diagram of the second-order Thomas analog filter circuit of this embodiment. As shown in Figure 2, the second-order Thomas analog filter circuit of this embodiment includes 3 amplifiers, 6 resistors and 2 capacitors. The input of the _first resistor R1 is used as the input of the entire circuit, and the first amplifier The output of t 1 is the measuring point t ₁ , and the output of the third amplifier is the measuring point t ₂ . The transfer function of measuring point t ₁ in the circuit shown in Figure 2 is shown in the following formula:

where ω represents the angular frequency.

The transfer function of the measuring point t ₂ is as follows:

In this embodiment, the method disclosed in the patent titled "A method for identifying fuzzy groups of analog circuits" and the patent number "201410336727.7" is adopted, and the fuzzy group analysis is performed based on the circle model, and the fuzzy group conditions of the circuit are obtained as follows: {R ₁ }, {R ₂ }, {R ₃ , C ₁ }, {R ₄ , R ₅ , R ₆ , C ₂ }. The faults of the internal elements of the fuzzy group are indistinguishable, and the faults between the fuzzy groups can theoretically be distinguished. In this embodiment, the representative fault elements of the four fuzzy groups are R ₁ , R ₂ , R ₃ , and R ₄ respectively, then the single fault types include {R ₁ }, {R ₂ }, {R ₃ }, {R ₄ }, double fault types include {R ₁ ,R ₂ }, {R ₁ ,R ₃ },{R ₁ ,R ₄ },{R ₂ ,R ₃ },{R ₂ ,R ₄ },{R 3 ,{R ₃ , R ₄ }, plus the no-fault state, there are 11 fault types in total. Table 1 is the nominal parameter value of each element in this embodiment.

element R₁ R₂ R₃ R₄ R₅ R₆ C₁ C₂ parameter value 10kΩ 10kΩ 10kΩ 10kΩ 10kΩ 10kΩ 0.01μF 0.01μF

Table 1

In this embodiment, the tolerance parameter is set to 0.05. Taking the element R ₁ as an example, its tolerance range is [9500Ω, 10500Ω], and the fault range is [0,9500Ω)∪(10500Ω,+∞).

Assuming that {R ₂ , R ₂ } is faulty, the parameters of each component of the analog circuit are [R ₁ , R ₂ , R ₃ , R ₄ , R ₅ , R ₆ , C ₁ , C ₂ ]=[20kΩ, 20kΩ, 9.6kΩ , 1.014kΩ, 9.85kΩ, 1.04kΩ, 9.796nF, 9.965nF], the responses of measuring points t ₁ and t ₂ are [-0.0245,-0.1545j], [-0.5143,0.0801j] respectively.

There are 11 fault types in this embodiment, so when initializing the genetic algorithm population, 11 subgroups are initialized, each subgroup contains one fault type, and the number of individuals in each subgroup is 10. The maximum number of iterations of the genetic algorithm is set to 100, the crossover probability is 1, and the mutation probability is 0.5. Table 2 is the running result of the genetic algorithm in this embodiment.

Table 2

According to the running results, the fault type corresponding to the optimal individual is {R ₁ , R ₂ }, and the corresponding component parameter vector is [19864.47, 19853.68, 9509.14, 9655.31, 10409.9449, 10151.26, 9.818e-09, 1.005e-08] . It can be seen that the diagnosis result is consistent with the fault type set in the example, and the parameters of each element are also very similar, so the present invention can accurately diagnose single and double fault types and identify single and double fault element parameters.

Although illustrative specific embodiments of the present invention have been described above to facilitate the understanding of the present invention by those skilled in the art, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, As long as various changes are within the spirit and scope of the present invention as defined and determined by the appended claims, these changes are obvious, and all inventions and creations utilizing the inventive concept are included in the protection list.

Claims (2)

1. a kind of analog circuit fault element parameter identification method based on genetic algorithm, is characterized in that, comprises the following steps: S1: Obtain the transfer function of the analog circuit at D measuring points t _d , d=1,2,...,D; S2: Carry out the fuzzy group analysis of fault diagnosis through the measurement point t in the analog circuit, select a representative fault element for each fuzzy group, record the number of representative fault elements as N, and record the number of other non-representative fault elements is M; S3: Setting no fault means that all faulty components in the analog circuit are not faulty, single fault means that only one representative faulty component fails when the analog circuit fails, and double faults means that there are two representative faulty components when the analog circuit fails If faults occur at the same time, record the number of fault types

S4: When it is necessary to identify the fault parameters of the analog circuit, measure the output voltages at D measuring points t _d under the preset excitation signal

Respectively represent the output voltage

The real and imaginary parts of , j is the imaginary unit; S5: Take X={x ₁ ,...,x _N ,x' ₁ ,...,x' _M } as the individuals in the genetic algorithm, where x _n represents the parameter value of the nth representative fault element, n=1, _{2 ,} . ,2,...,K, in each individual in the classification group p _k , take the parameter value x _n of the representative fault element corresponding to the kth fault type within the fault range of the representative fault element, and other The parameter value of the fault element is within the tolerance range; then the K subgroups p _k are merged to form a population P, and the number of individuals in the population P is recorded as G; S6: judge whether the iteration end condition of the genetic algorithm is reached, if yes, go to step S11, otherwise go to step S7; S7: Perform crossover and mutation operations on population P to obtain subpopulation Q; when performing crossover and mutation operations, it is necessary to ensure that the parameter values of at most two representative fault elements in each individual in subpopulation Q are located in the representative fault element The fault range of the other faulty components is within its tolerance range; S8: Combine population P and population Q to form population S, that is, S=P∪Q; S9: Substitute each individual in the population S into the transfer function to obtain the output voltages U _g,d =α _g,d +jβ _g,d ,α _g, at D measuring points t _d under the preset excitation signal _d and β _g,d represent the real part and imaginary part of the output voltage U _g, d respectively, g=1,2,...,2G, and then use the following formula to calculate the difference between the output voltage of the gth individual and the output voltage of the current analog circuit. The Euclidean distance D _g between :

S10: Divide the individuals in the population S into K subgroups _sk according to their corresponding fault types, screen out the individual with the smallest Euclidean distance from each subgroup _sk , add it to the next generation population P', and add it from Delete from the population S to obtain the population S'; then select GK individuals from the population S' and add the next generation population P'; then set the population P=P', and return to step S6; S11: Select the individual with the smallest Euclidean distance from the current population. The representative fault element whose parameter value is within the fault range in the individual is the fault diagnosis result, and the corresponding parameter value is the parameter identification result of the fault element. 2 . The method for identifying parameters of faulty components of an analog circuit according to claim 1 , wherein the specific method for performing individual optimization on the population S′ in the step S10 is: using the tournament optimization method to perform individual optimization on the current population S 2 . ,get

These preferred individuals are then sorted in ascending order according to the Euclidean distance, and the first GK individuals are selected to join the next generation population P'.

Images