Fault judgment method, device, electronic device and storage medium of electronic components based on phase space reconstruction
Through phase space reconstruction and fractal dimension analysis, the problems of process diagnosis and fault prevention in massive electronic system performance measurement data are solved, and efficient judgment and fault prevention of electronic components are achieved.
Patent Information
- Application Number
- CN202110934576.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-08-16
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2041-08-16
AI Technical Summary
In modern chip research and development, the nonlinearity, non-balance and multi-characteristic characteristics of massive electronic system performance measurement data make the diagnosis of process processes and early failure prevention a problem.
The fault judgment method of electronic components based on phase space reconstruction is used to sort the test data matrix, calculate the phase space reconstruction trajectory and analyze the fractal dimensions to determine whether the components are on the normal trajectory, and then determine their fault status.
It realizes processing of all electronic components test data at one time, displays component status from the system level, improves testing efficiency, and can discover potential abnormalities and dynamic evolution laws in the production process, and realizes diagnosis of the process process and early failure prevention.
Smart Images

Figure CN113791328B_ABST
Abstract
Description
Technical Field
[0001] This technology involves fault diagnosis, fault tracing and predictive analysis of electronic components, especially quality analysis of the performance of electronic components and judgment of the overall quality when there is a large amount of electronic system performance measurement data. Background Art
[0002] In modern chip research and development, the produced chips are tested by various test instruments and test platforms, generating various performance test data. The number of test parameters is as high as 2,000, and the number of chip samples involved in the test is as high as hundreds, forming a massive amount of data. These data reflect the chip production process and embody the inherent evolution law and change process of the production process. Due to the complexity of the production process, the test data is nonlinear, unbalanced, atypical and massive, and has the characteristics of multiple features coexisting. Therefore, how to achieve the diagnosis of the process and early fault prevention is an important and unresolved issue. Summary of the invention
[0003] The purpose of the present invention is to provide a fault diagnosis method for electronic components based on phase space reconstruction, which can process all electronic component test data at one time and display the status and possible faults of electronic components from a system level.
[0004] To achieve the above object, the technical solution adopted by the present invention is: a fault judgment method for electronic components based on phase space reconstruction, which comprises the following steps:
[0005] a. Number the electronic components in sequence, and obtain a test data matrix of multiple parameters arranged according to the electronic component numbers. The test data matrix X is m×n dimensional, as follows:
[0006]
[0007] Each row of the test data matrix represents a sample of test data of a certain component for all available parameters, and each column represents the test data of a certain available parameter for all components; m is the number of samples, and n is the number of parameters;
[0008] b. Select one parameter from all the parameters tested as the independent variable parameter, and sort the column from small to large. At the same time, all other parameter columns in X are also arranged in this order to obtain the matrix Y:
[0009]
[0010] Assuming that the first parameter is taken as the independent variable parameter, then the first column x1 of the matrix X = [x 11 x 21 ... x m1 ]T Sort from small to large to get the first column of Y y1 = [y 11 y 21 ... y m1 ] T , the order of all columns except the first column of matrix X needs to be rearranged in this order to obtain all columns except the first column of matrix Y, and all other columns are the dependent variable parameter data sequence;
[0011] c. Calculate the phase space reconstruction trajectory of each dependent variable parameter data sequence;
[0012] d. Obtain new parameter test data. If the new parameter test data is not on the phase space reconstruction trajectory, it is determined that the component is faulty; otherwise, it is not faulty.
[0013] Optimized, step c: According to Takens embedding theorem, a single time series {x(t), t = 1, 2, ..., L} is embedded into the m-dimensional space, and the resulting phase space vector can be expressed as:
[0014] X(t)=[x(t),x(t+τ),…,x(t+(m-1)τ)]
[0015] m is the embedding dimension, τ is the delay time, and the space formed by m-dimensional vectors is called the m-dimensional reconstructed phase space;
[0016] Since there is no time vector in the test data matrix Y, we use the column vector of the selected independent variable in Y as the pseudo time vector. Except for the column vector of the independent variable in Y, all other columns are observed as dependent variable sequences. It is necessary to make the phase space reconstruction trajectory of each dependent variable parameter sequence.
[0017] Furthermore, when m = 2 and m = 3, that is, in the case of two or three dimensions, let τ = 100, and the dependent variable parameter data sequence y2 = [y 12 y 22 ... y m2 ] T , converted into a 3D phase space trajectory, the coordinates of the three dimensions can be obtained by the pseudo-time delay embedding method:
[0018]
[0019] The phase space trajectory is reconstructed according to the above parameter sequence. If the new parameter test data is not on the trajectory, the relevant test data of the component is faulty.
[0020] Optimized, in step c, phase space reconstruction is performed according to each dependent variable parameter sequence to obtain the phase space reconstruction trajectory, the fractal dimension Df of each parameter phase trajectory is calculated respectively, the phase trajectory with fractal dimension greater than A is retained and a new phase space reconstruction trajectory is obtained, which is determined as the seed parameter of such components and used for judging the fault data of the corresponding parameters of the components,
[0021] The calculation method of fractal dimension uses box dimension, which is defined as follows:
[0022]
[0023] Where: Df represents the box dimension, N E (ε) is the minimum number of boxes that can cover the geometric object to be measured with a small box of size ε (ε is the side length);
[0024] For each determined component seed parameter, the distance dist between the test data restored on the new phase space reconstruction trajectory and the original phase trajectory as a whole is calculated. If the distance is greater than B, it is judged to be a faulty component; otherwise, if it is less than B, it is judged to be a normal component.
[0025] Furthermore, A<2.
[0026] Furthermore, B is three times the standard deviation of the corresponding parameter value of the sample.
[0027] Optimized, in step a, it is necessary to remove blank rows of data in all test data, where blank rows of data are sample rows containing at least one parameter value that is a null value or non-numeric data.
[0028] The present invention also provides a fault judgment device for electronic components, the device comprising:
[0029] An acquisition module, for acquiring a test data matrix of multiple parameters arranged according to the numbers of electronic components;
[0030] The first calculation module selects one parameter from all the parameters tested as the independent variable parameter, and sorts the column from small to large. At the same time, all other parameter columns in X are also arranged in this order to obtain the matrix Y;
[0031] The second calculation module is used to calculate the phase space reconstruction trajectory of each dependent variable parameter data sequence;
[0032] The judgment module is used to obtain new parameter test data and compare the new parameter test data with the new phase space reconstruction trajectory. If the new parameter test data is not on the phase space reconstruction trajectory, the relevant test data of the component is faulty; otherwise, there is no fault.
[0033] The present invention also provides an electronic device, comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus;
[0034] Memory, used to store computer programs;
[0035] The processor is used to implement the above-mentioned fault judgment method when executing the program stored in the memory.
[0036] The present invention also provides a computer-readable storage medium, on which a program of the fault judgment method is stored, and when the program of the fault judgment method is executed by a processor, the steps of the above-mentioned fault judgment method are implemented.
[0037] Due to the application of the above technical scheme, the present invention has the following advantages compared with the prior art: since the present invention can process all electronic component test data at one time, and display the status and possible failures of electronic components from a system level, the test efficiency is greatly improved, and multi-feature data can be used to discover potential anomalies and dynamic evolution laws in the production process, thereby realizing the diagnosis of the process and early fault prevention. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] Attached Figure 1 Reconstruct the three-dimensional image of the trajectory for the parameter sequence phase space;
[0039] Attached Figure 2 It is a schematic diagram of the relative position relationship between the corresponding points of the parameters in the phase space and the reconstruction trajectory of the phase space when the parameters are wrong;
[0040] Attached Figure 3 The circuit diagram of the Chua circuit is composed of three energy storage elements and a piecewise linear resistor;
[0041] Attached Figure 4 Phase space reconstruction trajectory diagram for Chua's circuit test data analysis method;
[0042] Attached Figure 5 Schematic diagram of the positional relationship between the point [1.4759, 0.1878, -0.1677] in the reconstructed three-dimensional phase space and the phase space reconstruction trajectory;
[0043] Attached Figure 6 : The schematic diagram of the position relationship between the point whose median value is [2.1973, 0.5408, 1..6190] in the reconstructed three-dimensional phase space and the phase space reconstruction trajectory;
[0044] Attached Figure 7 Reconstruct the trajectory image of the phase space corresponding to the ADC conversion chip test data;
[0045] Attached Figure 8 It is a schematic diagram of the positional relationship between the test data points of the chip numbered 200 in the ADC conversion chip test data and the phase space reconstruction trajectory in the reconstructed three-dimensional phase space;
[0046] Attached Fig. 9 This is a schematic diagram of the positional relationship between the test data points of the chip numbered 201 in the ADC conversion chip test data and the phase space reconstruction trajectory in the reconstructed three-dimensional phase space. DETAILED DESCRIPTION
[0047] The present invention will be further described below in conjunction with the embodiments shown in the accompanying drawings.
[0048] By utilizing the correlation and coupling of the corresponding test data of electronic components when the external test conditions change, the two-dimensional or three-dimensional phase space evolution trajectory image is used to judge the quality and reliability of the device. The method includes the following steps:
[0049] (1) Obtain a matrix of multiple parameter test data arranged by electronic component number.
[0050] The large amount of data obtained from the test of electronic components can be represented as a test data matrix. Before subsequent processing, the blank row data should be removed, and the measurement data that is too large or too small for the parameter should be removed. The test data matrix X is m×n dimensional, as follows:
[0051]
[0052] Each row of the data matrix represents a sample of test data of a certain component for all available parameters, and each column represents the test data of a certain parameter for all components; m is the number of samples, n is the number of parameters, and the available parameters refer to: among all samples, the number of samples with null values or non-numeric data (NaN) for this parameter does not exceed W%, W<100, and W=20 is the best value to avoid the inability to draw the corresponding phase trajectory of the parameter due to insufficient sample data.
[0053] (2) Sort the test data according to a certain independent variable parameter.
[0054] Select one parameter from all the parameters tested as the independent variable parameter, and sort the column from small to large. At the same time, all other parameter columns in X are also arranged in this order to obtain the matrix Y:
[0055]
[0056] Assuming that the first parameter is taken as the independent variable parameter, the first column x1=[x11 x21 ...xm1]T of the matrix X is sorted from small to large to obtain the first column y1=[y11 y21 ... ym1]T of Y. The order of all columns except the first column of the matrix X needs to be rearranged in this order to obtain all columns except the first column of the matrix Y. Here, all other columns are considered to be dependent variable parameter columns.
[0057] (3) Calculate the phase space reconstruction trajectory of each dependent variable parameter data sequence.
[0058] For all other parameter columns, two-dimensional or three-dimensional phase space trajectory graphs are drawn according to the time delay embedding method. According to Takens embedding theorem, a single time series {x(t), t = 1, 2, ..., L} is embedded into the m-dimensional space, and the resulting phase space vector can be expressed as:
[0059] X(t)=[x(t),x(t+τ),…,x(t+(m-1)τ)]
[0060] Here m is the embedding dimension and τ is the delay time. The space formed by m-dimensional vectors is called the m-dimensional reconstructed phase space.
[0061] Since there is no time vector in the test data matrix Y, we use the column vector of the selected independent variable in Y as the (pseudo) time vector. Except for the column vector of the independent variable in Y, all other columns are observed as dependent variable sequences, and the phase space reconstruction trajectory of each dependent variable parameter sequence needs to be made.
[0062] For the convenience of intuitive observation, we only consider m = 2 and m = 3, that is, the two-dimensional or three-dimensional case, and assume τ = 100. For example, the dependent variable parameter data sequence y2 = [y 12 y 22 ...y m2 ] T , converted into a 3D phase space trajectory, the coordinates of the three dimensions can be obtained by the (pseudo) time delay embedding method:
[0063]
[0064] For example, the three-dimensional image of the phase space trajectory reconstructed by a certain parameter sequence is shown in the figure below. The evolution law of the parameter can be intuitively displayed through the image. If the test data of the parameter is not on the trajectory, the relevant test data of the component is faulty.
[0065] (4) Eliminate random phase space trajectories and obtain seed parameters and their corresponding phase space trajectories.
[0066] Not all phase space trajectory images have regular patterns that can be used for quality evaluation, and random phase space trajectories need to be eliminated. This method reconstructs the trajectory images of the phase space according to each dependent variable parameter sequence, calculates the fractal dimension Df of each parameter phase trajectory, retains the phase trajectory with a fractal dimension greater than a certain threshold (for example, Df>1.5), and determines it as the seed parameter of such components, which is used to judge the fault data of the corresponding parameters of the components. (The seed parameter here refers to the parameter that can be used for component fault judgment.)
[0067] The calculation method of fractal dimension uses box dimension, which is defined as follows:
[0068]
[0069] Here Df represents the box dimension, N E (ε) is the minimum number of boxes that can cover the geometric object to be measured using a small box of size ε (ε is the side length).
[0070] (5) Determine the fault status of the electronic component for each determined component seed parameter.
[0071] The distance dist between the test data restored on the phase trajectory diagram and the original phase trajectory as a whole is calculated. If the distance is greater than a certain threshold, it is judged as a faulty component; otherwise, if it is less than the threshold, it is judged as a normal component.
[0072] To calculate the distance dist from a test data point z to the phase trajectory, one method is to calculate the minimum value of the distance from the data point to all points on the phase trajectory, that is:
[0073] dist z =min y∈Y {dist(z,y)}<threshold
[0074] Where Y is the set of all points on the phase trajectory, dist(z, y) represents the Euclidean distance from the test data point to any point on the phase trajectory, such as Figure 2 shown.
[0075] Apply the above method to specific test data for analysis:
[0076] 1. Cai's circuit test data analysis method
[0077] Chua's circuit is a circuit designed by Chinese-American scientist Professor Shao-Tang Chua in 1983 to demonstrate chaotic phenomena. It is also a simple circuit that can produce complex dynamics. By changing the topological structure or circuit parameters of the circuit, it can produce a variety of chaotic phenomena such as period-doubling bifurcations, single scrolls, double scrolls, and multi-scroll attractors. Chua's circuit uses three energy storage elements and a piecewise linear resistor. The circuit is as follows: Figure 3 As shown:
[0078] A simplified Chua circuit corresponds to the following set of third-order differential equations:
[0079]
[0080] in, The values of the parameters are a=7, b=0.35, c=0.5, d=0.7, The initial value is [1.5, -4.4, 0.15]. To illustrate this method, the model is used to generate 201 sets of [x, y, z] data as normal data reflecting the evolution law of the model. Then, a new set of test data [xe, ye, ze] is generated by adding disturbances. It is calculated whether the test data is on the evolution trajectory to determine whether it is fault data. The first 10 sets of these 200 sets of data are as follows:
[0081]
[0082] This is the measurement data matrix X in step 1 of this method. Each column of the data can be regarded as a measurement sequence of a parameter. Since these data are generated in time order, the matrix Y in step 2 is X. In step 3, the phase space reconstruction method with time delay is used with a delay time of τ=1 to make the phase space reconstruction trajectory of the first parameter column as follows: Figure 4 As shown, the phase space reconstruction trajectory is an oblique spiral disk curve, reflecting the internal evolution law of the circuit system. Whether the test data is on the phase trajectory curve can be used as a basis for judging whether the components related to the test data are normal data.
[0083] The distance threshold for judging any test data set to the phase space reconstruction trajectory is set to 1.
[0084] The values of the first set of test data are selected as [1.4759, 0.1878, -0.1677]. When the position of the point (star point) and the phase trajectory are plotted on the reconstructed phase space trajectory, as shown in Figure 5 As shown, it is found that this group of data is basically on the phase space reconstruction trajectory, and the minimum Euclidean distance between this point and the phase space reconstruction trajectory is calculated to be 0.3347 (<1), so it is judged to be a normal data group.
[0085] The values of the second set of test data are selected as [2.1973, 0.5408, 1..6190]. When the position of the point (asterisk point) and the phase space reconstruction trajectory are plotted on the reconstructed phase space reconstruction trajectory, as shown in Figure 6 As shown in FIG. 1 , by rotating the coordinate axis, it can be observed that there is a gap between the test data set and the phase space reconstruction trajectory. The minimum Euclidean distance between the point and the phase space reconstruction trajectory is calculated to be 2.1156 (>1), so it is judged to be a faulty data set.
[0086] 2. Processing of test data of a certain ADC conversion chip
[0087] Ultra-high-speed analog-to-digital converters (ADCs) are core components in digital processing systems. After sorting out various test data in component development, the number of test parameters reached 544, and the number of samples involved in the test reached 1007, forming the test data matrix X shown below:
[0088]
[0089] Select parameter 1 as the independent variable parameter, sort the first column from small to large, adjust the order of all chips, and obtain matrix Y. Reconstruct the phase space trajectory of all other parameters according to the time-delayed phase space reconstruction trajectory into the phase space reconstruction trajectory of three-dimensional space. Calculate the fractal dimension of all phase space reconstruction trajectories, and only select the phase space reconstruction trajectories greater than 1.75. For example, the phase space reconstruction trajectory of parameter 213 is as follows Figure 7 As shown, its fractal box dimension is 1.82.
[0090] The distance threshold for judging any test data group to the phase space reconstruction trajectory is set to 2. After adding the test data of the chip numbered 200 to the phase space reconstruction trajectory, the minimum distance between it and the phase space reconstruction trajectory is calculated to be 0.0409 (<2). Figure 8 It is found that it is basically on the phase space reconstruction trajectory, which is determined to be normal component test data, that is, the component is normal; for the chip numbered 201, after adding its test data to the phase space reconstruction trajectory, it is calculated that its minimum distance from the phase space reconstruction trajectory is 4.7397 (>2). By rotating the phase space coordinate axis and observing Fig. 9 It is found that it is some distance away from the phase trajectory and is determined to be the faulty component test data, that is, component failure.
[0091] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable people familiar with the technology to understand the content of the present invention and implement it accordingly, and they cannot be used to limit the protection scope of the present invention. Any equivalent changes or modifications made according to the spirit of the present invention should be included in the protection scope of the present invention.
Claims
1. A fault diagnosis method for electronic components based on phase space reconstruction, characterized in that: It includes the following steps: a. Number the electronic components in sequence, and obtain a test data matrix of multiple parameters arranged according to the electronic component numbers. The test data matrix X is m×n dimensional, as follows: Each row of the test data matrix represents a sample of test data of a certain electronic component for all available parameters, and each column represents the test data of a certain available parameter for all electronic components; m is the number of samples, and n is the number of parameters; b. Select one parameter from all the parameters tested as the independent variable parameter, and sort the column from small to large. At the same time, all other parameter columns in X are also arranged in this order to obtain the matrix Y: Assuming that the first parameter is taken as the independent variable parameter, then the first column x1 of the matrix X = [x 11 x 21 … x m1 ] τ Sort from small to large to get the first column of Y y1 = [y 11 y 21 … y m1 ] τ , the order of all columns except the first column of matrix X needs to be rearranged in this order to obtain all columns except the first column of matrix Y, and all other columns are the dependent variable parameter data sequence; c. Calculate the phase space reconstruction trajectory of each dependent variable parameter data sequence; d. Obtain new parameter test data. If the new parameter test data is not on the phase space reconstruction trajectory, it is determined that the electronic component is faulty; otherwise, it is not faulty.
2. The method for fault diagnosis of electronic components based on phase space reconstruction according to claim 1, characterized in that: Step c: According to Takens embedding theorem, a single time series {x(t), t=1,2,…,L} is embedded into the M-dimensional space. The resulting phase space vector can be expressed as: X(t) = [x(t), x(t+τ), ..., x(t+(M-1)τ)], where M is the embedding dimension, τ is the delay time, and the space formed by M-dimensional vectors is called the M-dimensional reconstructed phase space; Since there is no time vector in the test data matrix Y, the column vector of the selected independent variable in Y is used as a pseudo-time vector. Except for the column vector of the independent variable in Y, the other columns are observed as dependent variable sequences, and the phase space reconstruction trajectory of each dependent variable parameter sequence needs to be made.
3. The method for fault diagnosis of electronic components based on phase space reconstruction according to claim 2, characterized in that: When M = 2 and M = 3, that is, two-dimensional or three-dimensional, let τ = 100, and replace the dependent variable parameter data sequence y2 = [y 12 y 22 …y m2 ] τ , converted into a 3D phase space trajectory, the coordinates of the three dimensions can be obtained by the pseudo-time delay embedding method: The phase space trajectory is reconstructed according to the above parameter sequence. If the new parameter test data is not on the trajectory, the relevant test data of the electronic component is faulty.
4. The method for fault diagnosis of electronic components based on phase space reconstruction according to claim 1, characterized in that: In step c, phase space reconstruction is performed according to each dependent variable parameter sequence to obtain a phase space reconstruction trajectory, and the fractal dimension Df of each parameter phase trajectory is calculated respectively. The phase trajectory with a fractal dimension greater than A is retained and a new phase space reconstruction trajectory is obtained, which is determined as the seed parameter of such electronic components and is used to judge the fault data of the corresponding parameters of the electronic components. The calculation method of fractal dimension uses box dimension, which is defined as follows: Where: Df represents the box dimension, N E (ε) is the number of boxes with the minimum data of the geometric object to be measured that can be covered by a small box with a scale of ε, and ε is the side length; For each determined seed parameter of an electronic component, calculate the distance dist between the test data restored on the new phase space reconstruction trajectory and the original phase trajectory as a whole. If the distance is greater than B, it is judged to be a faulty component; otherwise, if it is less than B, it is judged to be a normal component; A<2; B is 3 times the standard deviation of the corresponding parameter value of the sample.
5. The method for fault diagnosis of electronic components based on phase space reconstruction according to claim 1, characterized in that: In step a, it is necessary to remove blank rows of data from all test data. Blank rows of data are sample rows that contain at least one parameter value that is a null value or non-numeric data.
6. A fault diagnosis device for electronic components, characterized in that: The device comprises: An acquisition module, for acquiring a test data matrix of multiple parameters arranged according to the numbers of electronic components; The first calculation module selects one parameter from all the parameters tested as the independent variable parameter, and sorts the column from small to large. At the same time, all other parameter columns in the test data matrix X are also arranged in this order to obtain the matrix Y; The second calculation module is used to calculate the phase space reconstruction trajectory of each dependent variable parameter data sequence; The judgment module is used to obtain new parameter test data and compare the new parameter test data with the new phase space reconstruction trajectory. If the new parameter test data is not on the phase space reconstruction trajectory, the relevant test data of the electronic component is faulty; otherwise, there is no fault.
7. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; Memory, used to store computer programs; The processor is used to implement the fault judgment method described in any one of claims 1 to 5 when executing the program stored in the memory.
8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program of the fault judgment method, and when the program of the fault judgment method is executed by the processor, the steps of the fault judgment method described in any one of claims 1 to 5 are implemented.
