Design of Fault Diagnosis Method Based on Circuit Frequency Domain Characteristics Fault Dictionary Method

In recent years, with the rapid development of integrated circuits and the widespread application of smart chips, the internal structure of complex electronic systems have become increasingly complex, which greatly increases the difficulty of board-level circuit testing and fault diagnosis of complex electronic systems. At present, traditional manual testing methods are no longer able to meet the maintenance requirements of modern complex electronic system board-level circuit. Therefore, the development of intelligent and universal automatic testing systems has become an important issue in fault detection of complex electronic systems. Based on the testing requirements of the automatic test system(ATS), this paper designs the corresponding overall scheme, software scheme and hardware circuit, and studies the fault diagnosis and test of board-level circuit of complex electronic system, and proposes to apply the traditional Bode diagram to the engineering practice of ATS. Fault detection is realized by utilizing the frequency domain characteristics of the object under test.


Introduction
The ATS of board level circuit can complete the index test, function test and fault diagnosis of board level circuit by controlling various instruments, and provides an important hardware platform for the development of automatic test program, the realization of fault diagnosis algorithm, signal processing and analysis.With the development of science and technology, the structure of board-level circuit becomes more and more complex.Compared with traditional manual test methods, ATS has faster test speed and higher measurement accuracy.Fault diagnosis method is an important part of the automatic test system [1].It is the technical basis for fault diagnosis and positioning.It is also a difficulty in the development of automatic test technology.In recent years, it has gradually become one of the focuses of automatic test system research [2].
At present, if ATS wants to realize effective fault diagnosis and location, a difficult and key point that needs to be solved urgently is the need to build powerful signal processing tools for the test system.According to the literature analysis, the current ATS mainly relies on the way of joint programming and the use of professional data processing software to solve the above problems [3].However, when performing joint programming, professional signal processing often requires a large amount of computer resources.Therefore, when large-scale automatic test systems perform complex circuit detection, it is often difficult for hardware equipment to meet the real-time and stability requirements of the test.
The frequency domain characteristics of the board-level circuit of a complex electronic system can reflect its circuit structure and component parameters [4].By testing the frequency domain characteristics of circuit and drawing frequency domain characteristics diagram, the working state of circuit can be studied.When there is a fault in the circuit, the system structure of the circuit will change, and then the frequency test of the response signal spectrum is bound to have a certain degree of variability.Therefore, by using appropriate excitation signals to input the circuit, the changes in the circuit output response can be studied, and then the changes in the internal structure of the circuit can be analyzed to achieve fault diagnosis of the circuit under test.

Principle of Fault Dictionary Method
The basic principle of fault dictionary method is: firstly, apply excitation signal to the circuit under test, obtain the circuit characteristics under various fault conditions, such as current, voltage, port impedance and amplitude-frequency characteristics, etc., and then make one-to-one correspondence between the obtained characteristics and faults, and organize them into fault dictionary [5].When performing fault diagnosis, the same excitation signal can be applied to the circuit, the response signal of the circuit under test is collected, and then compared with the data saved in the fault dictionary to assist in finding and determining the fault in the circuit.
By using frequency domain test method, a lot of circuit internal information can be obtained by testing multiple test points of board-level circuit.Input an AC signal of a certain frequency to the circuit network under test, and analyze the response signal to obtain the network gain of the circuit under test for this frequency signal.The amplitude-frequency characteristic curve and phase-frequency characteristic curve of the circuit can be obtained by changing the frequency of the input signal for many times within the working frequency band of the circuit under test.The Bode diagram of the system can be drawn by expressing the amplitude-frequency and phase-frequency characteristics of the system in logarithmic form.The graphical features of the Bode diagram such as the asymptote inflection point, slope, and zero point are related to the typical link parameters that make up the circuit.By studying the Bode diagram, the parameters of each link that make up the circuit can be analyzed.Therefore, the amplitude-frequency characteristics of each test point of the circuit can be selected as the circuit characteristics.Researchers can obtain the frequency domain characteristics of each measuring point under normal working state and each fault state of the tested circuit through circuit simulation in advance, establish a fault dictionary, and save it in the system database.

Circuit Linearization Analysis
Frequency-domain test methods are often used in the study of linear systems, but there are some limitations in the application of nonlinear systems.In the actual test process, it is often found that there are various nonlinear characteristics in the circuit.Therefore, this paper needs to study the nonlinear characteristics of the system in order to achieve the linearization of the nonlinear part.
After a strict inspection of the system component characteristics, especially the static characteristics, it is not difficult to find: Almost all components have non-linear relationships to varying degrees.Taking the iron core coil shown in figure 1 as an example, the relationship between voltage   and current  at both ends is: Wherein: / is constant when the coil is hollow, but / is a variable value when the core is inserted into the coil, and   has a nonlinear relationship with  When studying system characteristics, it is inevitable to encounter the problem of solving nonlinear differential equations.However, solving high-order nonlinear equations is quite difficult, and there is no universal solution.If we narrow the scope of study and approximate only a certain section of the nonlinear motion equation, then the input-output relationship of the system can be approximately regarded as linear and described by linear differential equations with constant coefficients in the scope of study.This method, which approximately transforms nonlinear differential equations into linear differential equations under certain conditions, is called linearization of nonlinear differential equations.
Usually control systems have an equilibrium state, and its corresponding position in the generalized space is called the equilibrium operating point.If the variable has derivative or partial derivative at the equilibrium point, then the nonlinear function describing nonlinear characteristics can be expanded into Taylor series by the deviation form of the variable near the equilibrium point.If the deviation of the variable from the equilibrium operating point is very small, then the influence of the higher-order terms of the deviation in the series can be ignored, and only its first-order term is considered.Finally, a linear function with the deviation of the variable as the independent variable can be obtained.Assuming that the function () is continuously differentiable near point A, the function can be expanded into a Taylor series in the domain of  0 : When the transformation quantity ∆ =  −  0 is very small, only the first-order term in Taylor series can be retained, and the higher-order term of second order and above can be omitted, then there is: order Then there is ∆ = ∆, which can be regarded as a linear equation of the nonlinear function () near the static operating point.In a geometric sense, this linear equation can be seen as approximately replacing the curve with the tangent of the curve at the equilibrium point near the equilibrium point.
Even nonlinear systems with more than two independent variables can be linearized using similar methods.For example, the nonlinear function  = (, ) with two independent variables  ,  is expanded into a Taylor series in the field of equilibrium points ( 0 ,  0 ): All partial derivatives of nonlinear function  = (, ) near the equilibrium operating point have corresponding finite values, ∆ =  −  0 , ∆ =  −  0 are the deviations of variables from the original operating point respectively.If the offsets of (, ) from the equilibrium operating point ( 0 ,  0 ) are very small, that is, when ∆ and ∆ are both trace amounts, the higher-order terms of ∆ and ∆ can be ignored and only their linear terms are considered, and we get This linearization method is based on the assumption that the deviation of variables from the predetermined operating point is very small, so it is called small deviation linearization method.It should be noted that the model constructed using small deviation linearization uses the deviation signal of variable x from the equilibrium operating point  0 as the independent variable.

Frequency Domain Test Solution Design
When various components in the circuit work near the static operating point, they can be regarded as a combination of linear devices.By referring to the data of the circuit under test and carrying out experimental tests, the static operating point voltage of the circuit under test is obtained, and then the superposition signal of the DC voltage signal at the static operating point and the AC signal with small amplitude is input as the excitation signal.The circuit under test can be regarded as a linear circuit and can be analyzed by frequency domain analysis method.
In order to obtain the frequency domain characteristics of a circuit, it is necessary to test the response signals of the circuit to multiple frequencies.When sinusoidal signal is used as excitation signal, only one frequency response can be obtained in a single test.In order to obtain enough fault characteristic information, it is necessary to repeat multiple rounds of experiments, which requires long test time and heavy test workload.To reduce the workload, it is necessary to redesign the excitation signal, which can directly input the superimposed signal of multiple frequency components, and then perform multifrequency analysis on the response signal to reduce duplication of work.
The square wave signal contains rich frequency domain components.The Fourier series expansion of the square wave is composed of odd harmonics.The square wave signal   () with frequency  0 and amplitude  0 can be regarded as superposition of multiple frequency signals.Since the overall structure of the circuit is relatively complex, in order to facilitate the analysis of circuit performance, the circuit can be divided into functional modules according to module functions and signal transmission relationships.The specific circuit division and signal flow are shown in figure 4. Because the input of each mocircuit module comes from the upper module, the low-pass filter amplifier circuit, the single-ended differential circuit and the differential filter circuit can be tested in turn during the test process.For C1, R1, C2 and R2, since the element is connected between two signal paths, when there is a short circuit fault, the circuit has no signal output, and the short circuit between the two channels is easy to cause damage to the test instrument.Therefore, when starting the test, it is necessary to test the input port impedance and output port impedance to determine whether there is a short circuit.
When C1, R1, C2 and R2 do not have short circuit phenomenon, the response signal of the circuit to each frequency signal can be tested and the circuit Bode diagram can be drawn.Bode diagram can be obtained by testing the circuit under normal working state and fault state of each element [7].
By analyzing Bode diagram and setting different fault states, the amplitude-frequency characteristics of the circuit under the action of specific frequency signals are tested, and the appropriate frequency points are found as test frequency points, and the amplitude-frequency characteristics of the circuit are obtained by frequency domain testing.As shown in figure 6 and figure 7, the abscissa in the figure is the frequency, and the ordinate refers to the noise power.Figure 6 is the normal working state of the circuit, and figure 7 is the short circuit state of L1 or L2.L1 and L2 are two coils on the common mode choke.The two coils are wound in the same direction.When the common mode interference enters the L1 and L2 coils, the two coils will generate magnetic fields in the same direction and superimpose each other.At this time, the flow coil will show larger impedance, thereby suppressing common mode current.For differential mode signals, the magnetic fields generated by the two coils are in opposite directions, so the impedance of the coils is small and has less impact on the original signal.Each fault state used as a waveform variable name, and the frequency point and its corresponding gain are used as waveform variable data and stored in user variables and saved as a signal_filter_ref.csvfile to establish a fault dictionary.

Summary
This paper conducts research on board-level circuit fault diagnosis testing of complex electronic systems, and proposes to apply the traditional Bode diagram to the engineering practice of automatic test systems, using the frequency domain characteristics of the object under test to achieve fault detection.First, the implementation principle of the fault diagnosis algorithm of the automatic test system, that is, the principle of the fault dictionary method, is introduced.It is proposed to apply the traditional Bode diagram to the engineering practice of the automatic test system and use the frequency domain characteristics of the measured object to establish a fault dictionary.Then the small deviation linearization theory was used to prove the feasibility of using frequency domain features, and a frequency domain test method was designed.Finally, each function of the fault diagnosis system was verified, and the design of the fault diagnosis scheme was completed.

Figure 1 .
Figure 1.Schematic diagram of iron core coil.

Figure 2 .
Figure 2. f(x) schematic diagram.As shown in figure 2, in the function  = (), x and y have a nonlinear relationship, and point A ( 0 ,  0 ) is in its equilibrium state.Take point A as the research object to carry out the research.When  =  0 + ∆, that is, point B in the figure,  =  0 + ∆.Assuming that the function () is continuously differentiable near point A, the function can be expanded into a Taylor series in the domain of  0 :

Figure 4 .
Figure 4. Circuit module division.Due to space limitations, this article only analyzes the differential filter circuit module, and the differential filter circuit is shown in figure 5. Since there are various kinds of interference in the differential signal transmission process, in order to reduce the interference caused by noise signals, it is necessary to use the differential filter circuit as shown in the figure to suppress noise and reduce interference [6].