Research On Matrix Model Method For Analog Circuit Testability And Fault Diagnosis

With the increase of circuit scale,the number of feedback loops increases,and the structure is highly complex,the difficulty of circuit testability analysis and fault diagnosis is greatly increased.If the traditional iterative method is used to solve the feedback problem,the amount of computation will be too large.However,the emerging artificial intelligence algorithms,affected by the parameters of circuit components,can not improve the fault diagnosis rate of analog circuits,and some algorithms require too much computation,resulting in a long diagnosis time.In order to solve the problem of the selection of test components and the low fault diagnosis rate in the analog circuit.In this paper,the transfer function of the circuit is used to obtain the partial derivative of the coefficients of the fault diagnosis equations of the component parameters that may fail.And the data after the partial derivative is used to establish a testability matrix,simplify the testability matrix,and find out the testability matrix.The linear correlation column of the property matrix yields the testability and fuzzy sets of the circuit.Select some testable components,and then adopts an algebraic-based method to form a response matrix of the output voltage timing of the circuit under test(CUT)using the matrix principle.Different parameters change to obtain different response matrices,and different matrices correspond to different matrix characteristics.Therefore,this paper uses the main eigenvalue and the largest singular value of the matrix to reflect the parameter changes.In this way,fault diagnosis and fault location can be realized,so as to find a set of optimal testable components for each circuit.The CUTs studied are Sallen-Key,CTSV,TowThomas and four opamp high-pass filter circuits.When the CUT structure is more complex,the data volume of the output voltage response matrix increases,so it can be known that the order of the matrix will affect the effect of fault diagnosis.When there are second-order canonical fuzzy groups in the circuit,the circuit is not completely testable.The innovations and research results of this paper are as follows:The first innovation of this research work is the testability analysis of CUT by an improved testability matrix method.And the simplification of the testability matrix to obtain its rank,which leads to the testability of the circuit,the fuzzy group,and the certain testable components.The second innovation is to combine the results of the circuit testability analysis with the matrix test optimization method and apply them to the fault diagnosis process of the CUT.Select some testable components,combine the matrix principle,and use the main eigenvalue and maximum singular value of the matrix to perform curve fitting respectively,so as to realize fault diagnosis and fault location.And the method in this paper obtains a high diagnostic rate and a smaller maximum recognition error.It can be seen from the fault diagnosis results of the Sallen-Key circuit that after the testability analysis of the circuit is carried out,some testable components are selected.And the quadratic curve fitting of the main eigenvalues and maximum singular values of the components is performed.The diagnosis rate in the case of the third-order output voltage response matrix is only 93%,while the diagnosis rate in the case of the fourth-order matrix is as high as 99%.It is shown that an increase in the order of the matrix leads to a more substantial increase in the diagnosis rate and that the maximum identification error in both cases is much less than 1%.Compared with the same CUT in the literature,the diagnostic rate of this paper has improved,and the maximum recognition error has a large decrease,and the diagnostic rate of the CTSV circuit under this method has also reached 99%.It shows that the selection of the main eigenvalue and the largest singular value of the matrix can accurately locate the fault of the selected testable components,and the method in this paper can obtain a high diagnosis rate without training samples.And a relatively optimal set of testable components in the two circuits can be determined.In order to verify the wide applicability of the method in this paper,two more complex circuits: the Tow-Thomas and the four opamp high-pass filter circuits for testability analysis and fault diagnosis.And their fault diagnosis rates are both above 98.75%.The diagnosis of the four opamp high-pass filter circuit is up to 100%.Compared to the same matrix model diagnostic method in the literature,the diagnostic rate of this method increased by 0.62 percentage points,indicating that performing a circuit testability analysis first helps to improve the fault diagnosis rate.