CN112731098A - Radio frequency low-noise discharge circuit fault diagnosis method, system, medium and application - Google Patents
Radio frequency low-noise discharge circuit fault diagnosis method, system, medium and application Download PDFInfo
- Publication number
- CN112731098A CN112731098A CN202011395989.2A CN202011395989A CN112731098A CN 112731098 A CN112731098 A CN 112731098A CN 202011395989 A CN202011395989 A CN 202011395989A CN 112731098 A CN112731098 A CN 112731098A
- Authority
- CN
- China
- Prior art keywords
- model
- fault
- radio frequency
- initial
- iteration
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- 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
- G01R31/2832—Specific tests of electronic circuits not provided for elsewhere
- G01R31/2836—Fault-finding or characterising
- G01R31/2843—In-circuit-testing
Landscapes
- Engineering & Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Complex Calculations (AREA)
Abstract
The invention belongs to the technical field of radio frequency circuits, and discloses a radio frequency low-noise discharge circuit fault diagnosis method, a system, a medium and application, wherein the method comprises the steps of inputting the state number, initial probability distribution, an initial state transition matrix, an iteration error, a maximum iteration step number and an observation vector sequence of a model; segmenting training sample data by a K-means algorithm according to the model state number, and initially estimating a Gaussian mixture density parameter by a GMM (Gaussian mixture model) to obtain an initial model of the CHMM; carrying out parameter model training through a Baum-Welch model to obtain an iterative reestimation model in the ith step; calculating the output probability under the reestimated model by a Viterbi algorithm, and calculating the increment error of the reestimated model output probability; and sequentially iterating until the error condition is satisfied and the convergence or the iteration step number is exceeded, and taking the reestimated model as a final result model. The rate of identifying the open-circuit and short-circuit faults of the elements in the radio frequency circuit can reach 100 percent of accuracy.
Description
Technical Field
The invention belongs to the technical field of radio frequency circuits, and particularly relates to a radio frequency low-noise discharge circuit fault diagnosis method, a radio frequency low-noise discharge circuit fault diagnosis system, a radio frequency low-noise discharge circuit fault diagnosis medium and application.
Background
At present: the current society is an information-based society, and especially the development of 5G technology has become the focus of the world in recent years. Significant scientific research costs invested in various countries around the world are attempting to dominate the fight against modern communications, with semiconductor device development undoubtedly being the focus. Analog circuits in which a radio frequency circuit is based on a semiconductor device and is combined with other components to convert certain energy into high-frequency signals and operate in high-frequency bands are rapidly popularized in the fields of communication, military, medical treatment and the like. Because the radio frequency circuit works in a high-frequency band, the switching speed of a semiconductor device of the radio frequency circuit is high, and some semiconductor devices usually need to work under a high-power condition such as a high-power amplifier, and the failure probability of the radio frequency circuit is greatly increased compared with that of a low-frequency analog circuit by considering factors such as working environment and the like. The failure rate of the semiconductor and welding failure in the energy conversion system accounts for 34%, and when a circuit fails, a failure source is timely positioned, so that the circuit can be timely maintained. In the early stage, the circuit fault diagnosis method mainly adopts manual maintenance, but with the rise of circuit integration and complexity, the requirement is far from met only by manual diagnosis.
The fault diagnosis refers to locating the fault position and the reason of a circuit which has a fault. In the early 60 s of the 20 th century, the failure diagnosis of circuits was first studied in the military industry as the third major branch of network theory. The fault diagnosis of the circuit can be divided into digital circuit fault diagnosis and analog circuit fault diagnosis, and the digital circuit fault diagnosis is developed to be mature at present and has good engineering application. And because excitation and response of the analog circuit are continuous quantities, fault parameters are difficult to extract, and the like, so that fault diagnosis of the analog circuit is slow. The problem of fault diagnosis of the analog circuit can be regarded as a pattern classification problem, and if a fault occurs in the operation process of the analog circuit, the characteristic parameters of the circuit deviate from a normal state, and the characteristic vectors also change. Therefore, such fault information is represented by the characteristic parameters as long as the fault source exists. The difference between the radio frequency circuit and the analog circuit lies in the working frequency, and the analysis theory of the traditional analog circuit is not applicable due to the working high-frequency band of the radio frequency circuit. Because the radio frequency circuit transmits energy by an electromagnetic field, the significance of analyzing the working state of the circuit by extracting input excitation and output response curves is not great. In addition, since the electromagnetic field is very sensitive to the change of the working environment, it also brings great difficulty to the test point selection of the radio frequency circuit. At present, few fault diagnosis researches on radio frequency circuits at home and abroad are carried out, and the research direction mainly takes physical failure of semiconductor devices as main research. The radio frequency circuit fault diagnosis research in foreign countries is mainly started from the interior of a transistor, and an IGBT physical failure model in a power supply of a power electronic conversion system is researched. The problem of thermal damage of the IGBT in a short-circuit state is researched through a simulation method. In China, a plurality of radio frequency circuit fault diagnosis researches are carried out, and fault diagnosis and service life prediction are carried out on radio frequency circuit fault modeling through an HSMM algorithm. And calculating the Mahalanobis distance by using an HMM model to evaluate the fault state of the circuit.
Through the above analysis, the problems and defects of the prior art are as follows:
(1) at present, radio frequency semiconductor devices tend to be miniaturized, have low power consumption, gradually increase working frequency, and greatly increase the failure rate of radio frequency circuits due to factors such as complex working environment and the like.
(2) At present, the fault parameters of the actual radio frequency circuit are difficult to extract, and the normal work of the circuit can be influenced by selecting the test points.
The difficulty in solving the above problems and defects is: due to the great development of radio frequency semiconductor devices, the cause of radio frequency circuit failure is more complicated. Multiple causes of failure act together, making the extracted data difficult to diagnose, including multiple sources of failure. The multiple faults act together to increase the complexity of extracting data, which also has higher requirements on the algorithm for establishing the fault model. On the other hand, the radio frequency circuit works mostly by transmitting signals through electromagnetic fields, and the addition of test points may affect the normal operation of the circuit. In combination with the above, the main difficulties in radio frequency circuit fault diagnosis are extraction of fault data and how to solve the problem of the combined action of fault sources to increase the data complexity.
The significance of solving the problems and the defects is as follows: whether the radio frequency circuit fault diagnosis is successful or not lies in the extraction of fault data, and a reasonable data extraction scheme is also beneficial to the selection of a radio frequency circuit fault diagnosis algorithm.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a method, a system, a medium and an application for diagnosing faults of a radio frequency low-noise discharge circuit.
The invention mainly relates to a radio frequency low-noise discharge circuit fault diagnosis method, which comprises the following steps:
a low-noise amplifying circuit is selected as a research object, and the low-noise amplifying circuit is simulated in the ADS.
And designing a fault data extraction scheme, performing fault injection on R1, R2 and R3, changing the ambient temperature to accelerate circuit degradation, and performing fault data extraction.
The method comprises the steps that a GMM-HMM model is used as a fault diagnosis method, and the state number, initial probability distribution, an initial state transition matrix, an iteration error, the maximum iteration step number and an observation vector sequence of the model are input;
carrying out sample segmentation on training sample data according to the model state number by using a K-means algorithm to obtain a Gaussian mixture density parameter so as to obtain an initial model of the CHMM;
further, the probability density function of the gaussian mixture model for the jth state in the HMM model, which produces the observation vector O, is written as:
in the formula, M represents the number of the j-th state Gaussian elements, cjlDenotes the number of Gaussian elements in the j-th state, N denotes the normal distribution, and μjlRepresenting the mean vector of the ith Gaussian element of the jth state and the covariance matrix of the ith Gaussian element of the jth state; the probability density function is reevaluated as follows, γt(j, l) represents the output probability of the ith Gaussian element of state j at time t:
further, model training is carried out through a Baum-Welch algorithm to obtain an iterative reestimation model in the step i;
calculating the output probability under the i-th reestimated model by a Viterbi algorithm, calculating the growth error of the reestimated model output probability, not meeting the error condition and keeping the number of iteration steps less than the maximum number of iteration steps, and continuing model training by taking the i-th reestimated model as the i + 1-th initial condition; and (4) meeting the error condition convergence or exceeding the maximum iteration step number, and taking the model reestimated in the ith step as a training result of the fault model.
And establishing a fault diagnosis model for each fault type according to the above, and sequentially bringing all the test data into each fault model. And acquiring all test data output probabilities by using a Viterbi algorithm for the fault model of each fault, and diagnosing the test data corresponding to the maximum test data output probability as the fault type corresponding to the fault model.
It is a further object of the invention to provide a computer device comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to perform the steps of:
inputting the state number, initial probability distribution, initial state transition matrix, iteration error, iteration maximum step number and observation vector sequence of the model;
carrying out sample segmentation on training sample data according to the model state number by using a K-means algorithm to obtain a Gaussian mixture density parameter so as to obtain an initial model of the CHMM; calculating each parameter of the CHMM according to a reestimation formula to obtain an iterative reestimation model;
calculating the output probability of the observation sample sequence under the reestimation model by a Viterbi algorithm
Calculating the output probability of the i-th reestimated model and the output probability error of the i-1-th reestimated model, wherein the error condition epsilon is not satisfied and i is<Continuing model training for L, i + 1; model for reestimating meeting error condition convergence or exceeding maximum iteration step number L
As a final result model λ;
it is another object of the present invention to provide a computer-readable storage medium storing a computer program which, when executed by a processor, causes the processor to perform the steps of:
inputting the state number, initial probability distribution, initial state transition matrix, iteration error, iteration maximum step number and observation vector sequence of the model;
carrying out sample segmentation on training sample data according to the model state number by using a K-means algorithm to obtain a Gaussian mixture density parameter so as to obtain an initial model of the CHMM; calculating each parameter of the CHMM according to a reestimation formula to obtain an iterative reestimation model;
calculating the output probability of the observation sample sequence under the reestimation model by a Viterbi algorithm
Calculating the output probability of the i-th reestimated model and the output probability error of the i-1-th reestimated model, wherein the error condition epsilon is not satisfied and i is<Continuing model training for L, i + 1; model for reestimating meeting error condition convergence or exceeding maximum iteration step number L
As a final result model λ;
another object of the present invention is to provide an information data processing terminal, which is used for implementing the radio frequency low noise discharging circuit fault diagnosis method.
Another object of the present invention is to provide a radio frequency low noise discharging circuit fault diagnosis system for implementing the radio frequency low noise discharging circuit fault diagnosis method, wherein the radio frequency low noise discharging circuit fault diagnosis system includes:
inputting the state number, initial probability distribution, initial state transition matrix, iteration error, iteration maximum step number and observation vector sequence of the model;
carrying out sample segmentation on training sample data according to the model state number by using a K-means algorithm to obtain a Gaussian mixture density parameter so as to obtain an initial model of the CHMM; calculating each parameter of the CHMM according to a reestimation formula to obtain an iterative reestimation model;
calculating the output probability of the observation sample sequence under the reestimation model by a Viterbi algorithm
Calculating the output probability of the i-th reestimated model and the output probability error of the i-1-th reestimated model, wherein the error condition epsilon is not satisfied and i is<Continuing model training for L, i + 1; model for reestimating meeting error condition convergence or exceeding maximum iteration step number L
As a final result model λ;
another object of the present invention is to provide a circuit fault diagnosis method using the radio frequency low noise circuit fault diagnosis method.
By combining all the technical schemes, the invention has the advantages and positive effects that: specifically, a radio frequency low-noise amplification circuit is taken as a research basis, and a Gaussian-hidden Markov model (GMM-HMM) is introduced to be taken as a fault diagnosis model. Inputting the state number, initial probability distribution, initial state transition matrix, iteration error, maximum iteration step number and observation vector sequence of the model; segmenting training sample data by a K-means algorithm according to the model state number, and initially estimating a Gaussian mixture density parameter by a GMM (Gaussian mixture model) to obtain an initial model of the CHMM; carrying out parameter model training through a Baum-Welch model to obtain a reestimation model of the ith (i < ═ maximum iteration step number) step iteration; calculating the output probability under the reestimated model by a Viterbi algorithm, and calculating the increment error of the reestimated model output probability; and sequentially iterating until the error condition is satisfied and the convergence or the iteration step number is exceeded, and taking the reestimated model as a final result model. Experimental results show that the fault diagnosis method based on the GMM + HMM (Gaussian mixture model + hidden Markov model) is provided. Experimental results show that the method can achieve 100% of accuracy rate for identifying open-circuit and short-circuit faults of components in a radio frequency circuit, and can achieve more than 83% of fault identification rate for components which deviate from a threshold range due to faults.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments of the present application will be briefly described below, and it is obvious that the drawings described below are only some embodiments of the present application, and it is obvious for those skilled in the art that other drawings can be obtained from the drawings without creative efforts.
Fig. 1 is a flowchart of a method for diagnosing a fault of a radio frequency low noise discharging circuit according to an embodiment of the present invention.
Fig. 2 is a schematic structural diagram of a radio frequency low noise discharge circuit fault diagnosis system provided in an embodiment of the present invention;
in fig. 2: 1. a parameter input module; 2. an iterative reestimation model obtaining module; 3. and a result model obtaining module.
Fig. 3 is a schematic diagram of an HMM model provided by an embodiment of the present invention.
Fig. 4 is a schematic diagram of an ATF54143 low-noise amplifier circuit according to an embodiment of the present invention.
Fig. 5 is a flowchart of a radio frequency low noise amplifier fault diagnosis provided in an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail with reference to the following embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
In view of the problems in the prior art, the present invention provides a method, a system, a medium and an application for diagnosing a fault of a radio frequency low noise discharge circuit, and the present invention is described in detail with reference to the accompanying drawings.
The radio frequency low noise amplifier is determined as a research object. The invention aims to find a fault diagnosis method suitable for a radio frequency analog circuit, and in order to enable the method to have better universality, a radio frequency low noise amplifier is selected for research, and the radio frequency low noise amplifier is often used in a radio frequency front-end circuit in the field of wireless communication due to the extremely low noise characteristic and is very widely applied.
The radio frequency circuit, regardless of its operating frequency or its characteristic of using electromagnetic field to propagate signals, will bring certain difficulties to the extraction of fault parameters. In addition, circuit fault inducement usually acts simultaneously, so that difficulty in extracting fault parameters in real objects is increased. For effective research, the fault data extraction work is carried out in a software simulation mode.
Circuit simulation was performed in ADS, as in fig. 4. In actual circuit design, software simulation is indispensable, and the difference between the physical performance of the welding through the template and the circuit performance calculated by simulation is not large.
And determining the type of the fault parameter. Parameters that can be used to reflect a fault for a radio frequency circuit include noise figure, stability figure, S-parameter, etc. In the process of research, the S parameter is finally determined as the final fault parameter type, and other parameters such as noise coefficient and stability coefficient are small in fluctuation range along with temperature, so that the fluctuation of the coefficient caused by the fault of a circuit is difficult to reflect.
And designing a fault parameter extraction scheme. As can be seen from the principle diagram of the radio frequency low noise amplifier circuit shown in fig. 4, R1, R2, and R3 are important elements of the static operating point of the circuit, and the fault parameters are obtained by changing R1, R2, and R3. For the convenience of research, the value of one component is changed in the fault injection process, and the parameter values of the other two components are kept unchanged or slightly changed.
And determining an extraction scheme for changing S parameter variation caused by simulation environment temperature variation. The actual circuit operation can cause energy conversion into heat energy to be released to a working environment due to aging, and the scheme of changing the environment temperature to accelerate the aging of the analog circuit accords with physical reality.
And extracting and recording fault parameters, and separating data according to a test set and a training set.
And selecting a left-right CHMM to establish a fault model. Left-right type CHMM is often used for fault diagnosis due to its characteristics, and the observation sequence of CHMM is usually assumed to be generated by gaussian probability density function simulation.
And establishing a GMM-HMM mixed model. In practical applications, gaussian probability density functions are often used to simulate the generation of observation sequences, each gaussian probability density function has its own mean and covariance matrix, and these parameters can be obtained by observing the characteristics of a sample.
The training samples are segmented according to the HMM state number by the K-means algorithm, then the mean value and the covariance matrix of each group of segmented sample data are calculated by a Gaussian mixture model, and the emission probability matrix B is simulated by the parameters.
As shown in fig. 1, the method for diagnosing a fault of a radio frequency low noise discharge circuit provided by the present invention includes the following steps:
s101: inputting the state number, initial probability distribution, initial state transition matrix, iteration error, iteration maximum step number and observation vector sequence of the model;
s102: obtaining initial model parameters by a K-means algorithm and a Gaussian mixture model to obtain an initial model lambda of the CHMM0；
S103, calculating each parameter of the CHMM according to a reestimation formula to obtain an iterative reestimation model;
s104: calculating the output probability of the observation sample sequence under the reestimation model by a Viterbi algorithm
Calculating the output probability of the i-th reestimated model and the output probability error of the i-1-th reestimated model, wherein the error condition epsilon is not satisfied and i is<Return to S103 is made to L, i + 1. Model for reestimating meeting error condition convergence or exceeding maximum iteration step number L
As a final result model λ;
s105: and (4) bringing all the test data into a fault model, obtaining the output probability by a Viterbi algorithm, and bringing the output probability into a classifier for fault diagnosis.
Those skilled in the art can also implement the method of diagnosing a fault in a radio frequency low noise discharging circuit by using other steps, and the method of diagnosing a fault in a radio frequency low noise discharging circuit provided by the present invention shown in fig. 1 is only a specific embodiment.
As shown in fig. 2, the radio frequency low noise discharging circuit fault diagnosis system provided by the present invention includes:
the parameter input module 1 is used for inputting the state number, the initial probability distribution, the initial state transition matrix, the iteration error, the iteration maximum step number and the observation vector sequence of the model;
an iterative reestimation model obtaining module 2 obtains initial model parameters by a K-means algorithm and a Gaussian mixture model to obtain an initial model lambda of the CHMM0(ii) a Calculating each parameter of the CHMM according to a reestimation formula to obtain an iterative reestimation model;
a result model obtaining module 3, which calculates the output probability of the observation sample sequence under the reestimation model by the Viterbi algorithm
Calculating the output probability of the i-th reestimated model and the output probability error of the i-1-th reestimated model, wherein the error condition epsilon is not satisfied and i is<Continuing model training for L, i + 1; satisfy the error condition convergence or exceed the maximum iteration step number L, model lambda reestimated0As a final result model λ;
the technical solution of the present invention is further described below with reference to the accompanying drawings.
The method simulates the low-noise amplifying circuit which is widely applied at present, establishes a fault model by using an HMM algorithm in a fault injection mode, and has certain reference significance for the research of radio frequency circuit fault diagnosis and positioning through experimental results.
1 hidden Markov model
1.1 Markov chain
A markov chain is a type of random variable sequence with a discrete set of state spaces that is finite or can be listed in time parameters.
Let random process X ═ Xn| n ═ 1,2, …, n }, the state space set is S, called XnWhen i is n, the random process is in the state i, and if the random process at the next time is in the state j, there are:
P(Xn+1＝j∣Xn＝i,Xn-1＝in-1,…,X1＝i1)＝P(Xn+1＝j∣Xn＝i)；
then we call the stochastic process X discrete time markov chain, majeldahl chain.
From the above definition, it is clear that the state of the Markov chain (Markov) at the next moment is only related to the current state, i.e. the process is memoryless, and the probability of the distribution of the next state is determined only from the current state, and not from its previous state.
1.2 hidden Markov chain (HMM)
HMMs have been developed based on markov chains, and unlike markov chains, HMM states are hidden, and hidden transition states and observable sequences satisfy a certain probability distribution. Thus, the HMM is a double stochastic process. A Markov chain that is hidden from transition states is modeled by the relationship of observable sequences to hidden states, referred to as a Hidden Markov Model (HMM), and the probabilistic relationship between observable sequences and hidden states is referred to as emission probability. One HMM model is shown in fig. 3, with Markov states hidden and observable variables represented by O.
The main parameters of the HMM model are:
(1) n: in the model, the model is divided into a plurality of models,number of states of Markov chain. Recording N states as S1,S2,···,SN. Recording the time t, wherein the Markov chain is in a state of qtIs apparent that qt∈(S1,S2,···,SN)。
(2) M: the number of possible observations for each state. Recording M observed values as o1,o2,···,oMRecording: observed value at time t is otWherein o ist∈(v1,v2,···,vM)。
(3) Pi: the initial probability distribution is pi ═ pi (pi)1,π2,···,πN). Wherein. Pii＝P(qt＝Si)，1≤i≤N。
(4) A: observation value transfer matrix, a ═ aij)N×NWherein a isij＝P(qi+1＝Sj|qi＝Si)，1≤i，j≤N。
(5) B: observed value probability matrix B ═ Bjk)N×MWherein b isjk＝P(oi＝vk|qi＝Sj)，1≤j≤N，1≤k≤M。
The HMM is noted as: λ ═ N, M, pi, a, B, abbreviated λ ═ N, a, B.
2 Gauss Mixed Model (GMM)
HMMs can be classified into continuous hidden markov models (DHMMs) and Discrete Hidden Markov Models (DHMMs) by distinguishing between the observation matrix B either discrete or continuous. The DHMM needs to be discrete in observation variable during modeling, the calculated amount is small due to the fact that discrete data are processed, and the model training speed is high. The method has the disadvantages that when continuous observation variables are processed, the continuous variables need to be discretized. The discretization is usually a vector or scalar quantization method, the information loss is inevitably caused by the discretization of continuous variables, and the accuracy of the established model is lower under the condition that the data quantity of observable variables is small. Unlike DHMM, CHMM has certain advantages in dealing with continuously observed variables, and CHMM is used for modeling in conjunction with the present invention.
The observation vectors of discrete HMM models can be represented by distributed probabilities, and the observation sequences of continuous HMM models are usually assumed to be generated by gaussian probability density function simulations.
2.1GMM theory
The gaussian mixture model refers to a probability distribution model having the form:
wherein alpha iskIs a coefficient, αk≥0，Σαk＝1；φ(y|θk) Is a Gaussian distribution density thetak＝(μk,σk)：
Referred to as the kth distribution model.
2.2GMM-HMM Algorithm
The GMM is developed on the basis of a single Gaussian probability distribution density function, can be used for simulating training sample data, and has good applicability to some continuity probability distribution problems, so that the GMM can be used together with the HMM, the defect of solving the continuity problem of the HMM is overcome, and the GMM is used for fitting the emission probability of the HMM to form a GMM-HMM algorithm.
For the jth state in the HMM model, the probability density function that produces the observation vector O can be written as:
wherein, the number of the Gaussian elements of the jth state of M, cjlThe number of Gaussian elements in the j-th state, N represents the standard normal distribution, mujlAnd the covariance matrix represents the mean vector of the ith Gaussian element of the jth state and the ith Gaussian element of the jth state. The probability density function is reevaluated as follows, γt(j, l) represents the output probability of the ith Gaussian element of state j at time t:
3 protocol
The experiment can divide the fault prediction of the radio frequency circuit into two parts. And part is the extraction and preprocessing of circuit fault parameters. And the other part is the establishment of a parameter model and the fault prediction. Based on the ATF54143 low noise discharge circuit design schematic diagram as shown in FIG. 5, the design simulation is performed in ADS software.
3.1 Fault parameter extraction
Considering the composition of a capacitor, an inductor and an active component in a circuit, faults can be manually injected into the circuit respectively. The main reason for this phenomenon, which is the principle that the performance of electronic circuits gradually degrades with increasing operating time, is that the ambient temperature of electronic circuits continuously rises with increasing operating time during operation. Thus, the degradation performance of the electronic circuit is known to be related to the operating time and the ambient temperature. According to the law of conservation of energy, energy can only be transferred from one object to another object, and can also be converted into mechanical energy or other energy, but the total value of the energy is kept unchanged in the conversion process. By combining the above, the aging process of the electronic circuit can be accelerated by changing the environmental temperature of the electronic circuit, and the extraction work of the fault characteristic parameters of the electronic circuit can be completed. The method has the advantages of better reflecting the normal degradation process of the electronic circuit and better conforming to the physical characteristics of the electronic circuit. The partial extraction data are shown in tables 1 and 2.
TABLE 1 partial failure data
TABLE 2 partial failure data
3.2 Fault model training and recognition
Based on the extracted fault parameters, the fault model training may be divided into element-level fault model training and system-level fault model training. And (3) element-level fault model training, wherein fault parameters are extracted by injecting faults (open circuit, short circuit, and +/-5% exceeding the normal working range of the element) into the element in the circuit, and then corresponding HMM fault models are respectively trained. The transmit probability matrix B may be obtained by fitting the transmit probability through the GMM algorithm. And then combining with the Baum-Welch algorithm to train the model. As shown in fig. 5, the complete CHMM model training procedure is as follows:
1) firstly, inputting the state number, the initial probability distribution pi, an initial state transition matrix A, an iteration error epsilon, an iteration maximum step number L and an observation vector sequence O of a model; (ii) a
2) Carrying out sample segmentation on training sample data according to the model state number by using a K-means algorithm to obtain a Gaussian mixture density parameter so as to obtain an initial model of the CHMM; calculating each parameter of the CHMM according to a reestimation formula to obtain an iterative reestimation model;
3) calculating the output probability of the observation sample sequence under the reestimation model by a Viterbi algorithm
Calculating the output probability of the i-th reestimated model and the output probability error of the i-1-th reestimated model, wherein the error condition epsilon is not satisfied and i is<Continuing model training for L, i + 1; model for reestimating meeting error condition convergence or exceeding maximum iteration step number L
As a final result model λ;
and (4) carrying out training by bringing the trained fault model into a test set, analyzing and identifying accuracy, and analyzing an experimental result.
4 results of the experiment
The invention selects ATF54143 to design the low noise amplifier, and simulates in ADS software. The correlation of various parameters along with the circuit degradation is analyzed, and finally the S21 parameter is selected as a fault parameter for modeling and identification.
TABLE 3 Fault diagnosis results
TABLE 4 Fault diagnosis results
Table 3 the fault types correspond in sequence to R1 open, R1 short, R2 open, R2 short, R3 open, R3 short. Table 4 fault types correspond in sequence to R1, R2, R3 deviating from the normal threshold range by-10%, 5%, + 10%. From experimental results, the method has good diagnosis accuracy for R1, R2 and R3 open-circuit short-circuit faults. For R1, R2, R3, the diagnosis accuracy is reduced from the normal threshold range, and as can be seen from table 4, the failure type 1 and the failure type two have a set of data that is misdiagnosed as failure type 5, and the failure type 3 is all misdiagnosed as failure type 9.
The invention provides a fault diagnosis research method based on an HMM algorithm. And establishing a low-noise amplification circuit by taking the ATF54143 transistor as a model, and providing a method for extracting fault parameters. On the basis of the previous research, the low-noise amplifier circuit is subjected to element-level and system-level fault diagnosis research respectively.
From experimental results, the accuracy rate of identifying the open-circuit short-circuit fault is 100%. This is because current fault data injection is dominated by open-circuit shorts. As can be seen from Table 3, the two types of fault data are distinguished more obviously, and the fault model trained in the way is also higher in distinguishing degree, so that the identification accuracy is high. From table 4, it can be seen that the deviation degree of the components from the threshold value is small, which indicates that the circuit is in the critical fault state, the fault identification accuracy at this time is reduced, and the result is reflected in table 4, and the accuracy for identifying such faults is 84%.
It should be noted that the embodiments of the present invention can be realized by hardware, software, or a combination of software and hardware. The hardware portion may be implemented using dedicated logic; the software portions may be stored in a memory and executed by a suitable instruction execution system, such as a microprocessor or specially designed hardware. Those skilled in the art will appreciate that the apparatus and methods described above may be implemented using computer executable instructions and/or embodied in processor control code, such code being provided on a carrier medium such as a disk, CD-or DVD-ROM, programmable memory such as read only memory (firmware), or a data carrier such as an optical or electronic signal carrier, for example. The apparatus and its modules of the present invention may be implemented by hardware circuits such as very large scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., or by software executed by various types of processors, or by a combination of hardware circuits and software, e.g., firmware.
The above description is only for the purpose of illustrating the present invention and the appended claims are not to be construed as limiting the scope of the invention, which is intended to cover all modifications, equivalents and improvements that are within the spirit and scope of the invention as defined by the appended claims.
Claims (10)
1. A radio frequency low noise discharge circuit fault diagnosis method is characterized by comprising the following steps:
selecting a low-noise amplifying circuit as a research object, and simulating the low-noise amplifying circuit in the ADS;
designing a fault data extraction scheme, carrying out fault injection on R1, R2 and R3, changing the ambient temperature to accelerate circuit degradation, and carrying out fault data extraction;
the method comprises the steps that GMM-HMM is used as a fault diagnosis method, and the state number, initial probability distribution, an initial state transition matrix, an iteration error, an iteration maximum step number and an observation vector sequence of a model are input;
carrying out sample segmentation on training sample data according to the model state number by using a K-means algorithm to obtain a Gaussian mixture density parameter so as to obtain an initial model of the CHMM;
carrying out model training through a Baum-Welch algorithm to obtain an iterative reestimation model in the step i;
calculating the output probability under the i-th reestimated model by a Viterbi algorithm, calculating the growth error of the reestimated model output probability, not meeting the error condition and keeping the number of iteration steps less than the maximum number of iteration steps, and continuing model training by taking the i-th reestimated model as the i + 1-th initial condition; satisfying the error condition convergence or exceeding the maximum iteration step number, and taking the model reestimated in the ith step as the training result of the fault model;
and establishing a fault diagnosis model for each fault type, and bringing all test data into each fault model in sequence. And obtaining all test data output probabilities by using a Viterbi algorithm for the fault model of each fault, and diagnosing the test data corresponding to the maximum likelihood probability of all the test data as the fault type corresponding to the fault model.
2. The radio frequency low noise discharging circuit fault diagnosing method as claimed in claim 1, wherein the probability density function of the radio frequency low noise discharging circuit fault diagnosing method for the jth state in the HMM model, which generates the observation vector O, is written as:
wherein M represents the number of the j-th state Gaussian elements, cjlDenotes the number of Gaussian elements in the j-th state, N denotes the normal distribution, and μjlRepresenting the mean vector of the ith Gaussian element of the jth state and the covariance matrix of the ith Gaussian element of the jth state; the probability density function is reevaluated as follows, γt(j, l) represents the output probability of the ith Gaussian element of state j at time t:
3. the radio frequency low noise discharging circuit fault diagnosis method according to claim 1, wherein the radio frequency low noise discharging circuit fault diagnosis method includes: extracting and preprocessing circuit fault parameters; establishing a parameter model and predicting faults.
4. The radio frequency low noise discharging circuit fault diagnosis method according to claim 3, wherein the fault parameter extraction of the radio frequency low noise discharging circuit fault diagnosis method accelerates the aging process of the electronic circuit by changing the ambient temperature of the electronic circuit, and completes the electronic circuit fault characteristic parameter extraction work.
5. The radio frequency low noise discharging circuit fault diagnosis method according to claim 3, wherein the radio frequency low noise discharging circuit fault diagnosis method performs fault model training and identification, and performs element-level fault model training according to the extracted fault parameters; fault parameters are extracted by injecting faults into components in the circuit, and HMM fault models corresponding to the fault parameters are trained respectively; fitting the emission probability through a GMM algorithm to obtain an emission probability matrix B, and then performing model training by combining a Baum-Welch algorithm, wherein the complete CHMM model training step is as follows:
1) firstly, inputting the state number, the initial probability distribution pi, an initial state transition matrix A, an iteration error epsilon, an iteration maximum step number L and an observation vector sequence O of a model;
2) obtaining initial model parameters by a K-means algorithm and a Gaussian mixture model to obtain an initial model lambda of the CHMM0；
3) Using Baum-Welch algorithm can be represented byi-1Deriving the ith CHMM model and λi；
4) Calculating the output probability of the observation sample sequence under the reestimation model by a Viterbi algorithm
Calculating the output probability of the i-th reestimated model and the output probability error of the i-1-th reestimated model, wherein the error condition epsilon is not satisfied and i is<L, i +1 returns to 3); model for reestimating meeting error condition convergence or exceeding maximum iteration step number L
As a final result model λ.
6. A computer device, characterized in that the computer device comprises a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to carry out the steps of:
inputting the state number, initial probability distribution, initial state transition matrix, iteration error, iteration maximum step number and observation vector sequence of the model;
obtaining initial model parameters by a K-means algorithm and a Gaussian mixture model to obtain an initial model lambda of the CHMM0(ii) a Calculating each parameter of the CHMM according to a reestimation formula to obtain an iterative reestimation model lambdai；
Calculating the output probability of the observation sample sequence under the reestimation model by a Viterbi algorithm
Calculating the output probability of the i-th reestimated model and the output probability error of the i-1-th reestimated model, wherein the error condition epsilon is not satisfied and i is<Continuing model training for L, i + 1; model for reestimating meeting error condition convergence or exceeding maximum iteration step number L
As a final result model λ.
7. A computer-readable storage medium storing a computer program which, when executed by a processor, causes the processor to perform the steps of:
inputting the state number, initial probability distribution, initial state transition matrix, iteration error, iteration maximum step number and observation vector sequence of the model;
obtaining initial model parameters by a K-means algorithm and a Gaussian mixture model to obtain an initial model lambda of the CHMM0(ii) a Calculating each parameter of the CHMM according to a reestimation formula to obtain an iterative reestimation model lambdai；
Calculating the output probability of the observation sample sequence under the reestimation model by a Viterbi algorithm
Calculating the output probability of the i-th reestimated model and the output probability error of the i-1-th reestimated model, wherein the error condition epsilon is not satisfied and i is<Continuing model training for L, i + 1; model for reestimating meeting error condition convergence or exceeding maximum iteration step number L
As a final result model λ.
8. An information data processing terminal, characterized in that the information data processing terminal is used for implementing the fault diagnosis method of the radio frequency low noise discharging circuit of any one of claims 1 to 5.
9. A radio frequency low noise discharging circuit fault diagnosis system for implementing the radio frequency low noise discharging circuit fault diagnosis method of any one of claims 1 to 5, wherein the radio frequency low noise discharging circuit fault diagnosis system comprises:
the parameter input module is used for inputting the state number, the initial probability distribution, the initial state transition matrix, the iteration error, the iteration maximum step number and the observation vector sequence of the model;
an iterative reestimation model obtaining module, which obtains initial model parameters by a K-means algorithm and a Gaussian mixture model to obtain an initial model lambda of the CHMM0(ii) a Calculating each parameter of the CHMM according to a reestimation formula to obtain an iterative reestimation model lambdai；
And the result model obtaining module is used for calculating the output probability of the observation sample sequence under the reestimation model by a Viterbi algorithm, calculating the increment error of the reestimation model output probability, meeting the error condition convergence and taking the reestimated model as a final result model.
10. A circuit fault diagnosis method, characterized in that the radio frequency low noise discharge circuit fault diagnosis method of any claim 1-5 is used.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011395989.2A CN112731098B (en) | 2020-12-03 | 2020-12-03 | Radio frequency low-noise discharge circuit fault diagnosis method, system, medium and application |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011395989.2A CN112731098B (en) | 2020-12-03 | 2020-12-03 | Radio frequency low-noise discharge circuit fault diagnosis method, system, medium and application |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112731098A true CN112731098A (en) | 2021-04-30 |
| CN112731098B CN112731098B (en) | 2022-04-29 |
Family
ID=75598419
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011395989.2A Active CN112731098B (en) | 2020-12-03 | 2020-12-03 | Radio frequency low-noise discharge circuit fault diagnosis method, system, medium and application |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112731098B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113377595A (en) * | 2021-06-11 | 2021-09-10 | 上海壁仞智能科技有限公司 | Fault diagnosis method, fault diagnosis device, electronic equipment and storage medium |
| CN113625146A (en) * | 2021-08-16 | 2021-11-09 | 长春理工大学 | Semiconductor device 1/f noise S alpha S model parameter estimation method |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090180695A1 (en) * | 2008-01-15 | 2009-07-16 | Xerox Corporation | Asymmetric score normalization for handwritten word spotting system |
| US20130124438A1 (en) * | 2011-11-11 | 2013-05-16 | Sung-young Lee | Method of recognizing patterns based on markov chain hidden conditional random field model |
| CN104504296A (en) * | 2015-01-16 | 2015-04-08 | 湖南科技大学 | Gaussian mixture hidden Markov model and regression analysis remaining life prediction method |
| CN105137328A (en) * | 2015-07-24 | 2015-12-09 | 四川航天***工程研究所 | Analog integrated circuit early-stage soft fault diagnosis method and system based on HMM |
| CN108563874A (en) * | 2018-04-17 | 2018-09-21 | 重庆大学 | A kind of analog circuit intermittent fault diagnostic method |
| CN109029992A (en) * | 2018-06-15 | 2018-12-18 | 深圳大学 | The method and system of the bearing failure diagnosis of machine tool chief axis |
| CN111444988A (en) * | 2020-05-11 | 2020-07-24 | 北华大学 | Rolling bearing fault diagnosis system |
| CN111751714A (en) * | 2020-06-11 | 2020-10-09 | 西安电子科技大学 | Radio frequency analog circuit fault diagnosis method based on SVM and HMM |
-
2020
- 2020-12-03 CN CN202011395989.2A patent/CN112731098B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090180695A1 (en) * | 2008-01-15 | 2009-07-16 | Xerox Corporation | Asymmetric score normalization for handwritten word spotting system |
| US20130124438A1 (en) * | 2011-11-11 | 2013-05-16 | Sung-young Lee | Method of recognizing patterns based on markov chain hidden conditional random field model |
| CN104504296A (en) * | 2015-01-16 | 2015-04-08 | 湖南科技大学 | Gaussian mixture hidden Markov model and regression analysis remaining life prediction method |
| CN105137328A (en) * | 2015-07-24 | 2015-12-09 | 四川航天***工程研究所 | Analog integrated circuit early-stage soft fault diagnosis method and system based on HMM |
| CN108563874A (en) * | 2018-04-17 | 2018-09-21 | 重庆大学 | A kind of analog circuit intermittent fault diagnostic method |
| CN109029992A (en) * | 2018-06-15 | 2018-12-18 | 深圳大学 | The method and system of the bearing failure diagnosis of machine tool chief axis |
| CN111444988A (en) * | 2020-05-11 | 2020-07-24 | 北华大学 | Rolling bearing fault diagnosis system |
| CN111751714A (en) * | 2020-06-11 | 2020-10-09 | 西安电子科技大学 | Radio frequency analog circuit fault diagnosis method based on SVM and HMM |
Non-Patent Citations (2)
| Title |
|---|
| 李丽敏等: "基于多状态ＭＯＧ－ＨＭＭ 和Ｖｉｔｅｒｂｉ的航空发动机突发故障预测", 《振动、测试与诊断》 * |
| 陆汝华等: "基于CGHMM的轴承故障音频信号诊断方法", 《基于CGHMM的轴承故障音频信号诊断方法》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113377595A (en) * | 2021-06-11 | 2021-09-10 | 上海壁仞智能科技有限公司 | Fault diagnosis method, fault diagnosis device, electronic equipment and storage medium |
| CN113377595B (en) * | 2021-06-11 | 2023-06-23 | 上海壁仞智能科技有限公司 | Fault diagnosis method, device, electronic equipment and storage medium |
| CN113625146A (en) * | 2021-08-16 | 2021-11-09 | 长春理工大学 | Semiconductor device 1/f noise S alpha S model parameter estimation method |
| CN113625146B (en) * | 2021-08-16 | 2022-09-30 | 长春理工大学 | Semiconductor device 1/f noise S alpha S model parameter estimation method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112731098B (en) | 2022-04-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109931678B (en) | Air conditioner fault diagnosis method based on deep learning LSTM | |
| CN112731098B (en) | Radio frequency low-noise discharge circuit fault diagnosis method, system, medium and application | |
| CN110334726A (en) | A kind of identification of the electric load abnormal data based on Density Clustering and LSTM and restorative procedure | |
| CN104778298A (en) | Gaussian process regression soft measurement modeling method based on EGMM (Error Gaussian Mixture Model) | |
| CN111931983B (en) | Precipitation prediction method and system | |
| CN111814956A (en) | Multi-task learning air quality prediction method based on multi-dimensional secondary feature extraction | |
| CN111751714A (en) | Radio frequency analog circuit fault diagnosis method based on SVM and HMM | |
| CN115809624B (en) | Automatic analysis design method for integrated circuit microstrip line transmission line | |
| CN110750948A (en) | Analog circuit optimization algorithm based on multi-objective acquisition function integrated parallel Bayesian optimization | |
| CN113988415B (en) | Medium-and-long-term power load prediction method | |
| CN113657030A (en) | Method for predicting remaining service life of lithium battery based on Gaussian process regression | |
| Somayaji et al. | Prioritized reinforcement learning for analog circuit optimization with design knowledge | |
| Zhang et al. | A novel analog circuit fault diagnosis approach | |
| CN111291820B (en) | Target detection method combining positioning information and classification information | |
| Xiang et al. | An improved multiple imputation method based on chained equations for distributed photovoltaic systems | |
| CN114329805A (en) | Connecting piece multidisciplinary collaborative design optimization method based on self-adaptive agent model | |
| CN117561502A (en) | Method and device for determining failure reason | |
| CN113433409A (en) | Electric automobile IGBT type common direct current bus charging equipment fault diagnosis method based on deep learning | |
| CN111709192A (en) | Plane inverted F-shaped antenna resonant frequency prediction method based on semi-supervised learning | |
| Sun et al. | Efficient Mask Attention-Based NARMAX (MAB-NARMAX) Model Identification | |
| Dhanasekar et al. | Analysis of Machine Learning Techniques for Time Domain Waveform Prediction in Analog and Mixed Signal Integrated Circuit Verification | |
| Masamba et al. | Supervised learning for coverage-directed test selection in simulation-based verification | |
| CN116595862B (en) | Self-adaptive modeling method based on Gaussian process regression | |
| CN117520802B (en) | Method, system and equipment for identifying lightweight audit event of electric power engineering site | |
| CN114298487B (en) | Reliability evaluation method and system for ship equipment system |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |