A Near-Field Scan Method Based on Active Learning for Obtaining Conductive Coupling Path Detection

Abstract

This invention discloses a near-field scanning detection method based on active learning to obtain conducted coupling paths. The method includes: constructing a near-field scanning detection platform; placing the circuit board under test (PCB) in a microwave anechoic chamber; and using the near-field scanning detection platform directly above the PCB to perform detection; obtaining the detection results using an active learning-based near-field scanning method; and obtaining a near-field current distribution map on the PCB at continuous time intervals based on the detection results, thereby realizing the detection of conducted coupling paths. This invention utilizes active machine learning methods to achieve more efficient near-field scanning detection, concentrating scanning points in areas with strong currents and accurately reconstructing the actual current changes of the PCB using the scan results of sparse points, greatly improving detection efficiency. By using near-field scanning detection to reconstruct the current changes on the circuit board, the coupling paths of conducted interference currents on the PCB can be quickly analyzed, meeting the requirements for detecting electromagnetic interference in systems in practical engineering.

Description

Technical Field

[0001] This invention relates to a coupling path detection method, specifically a near-field scanning method for obtaining conductive coupling path detection based on active learning. Background Technology

[0002] The electronics and semiconductor industries have developed rapidly. However, as the complexity of components and circuit designs in electronic devices increases, electromagnetic interference (EMI) problems are becoming increasingly serious. At best, it affects the working state of electronic devices, reducing performance and causing malfunctions; at worst, it can severely damage electronic systems and even endanger personal safety. Therefore, EMI has always been a key focus and research area during the research and development phase of electronic products.

[0003] Near-field scanning uses electromagnetic field probes to measure the electromagnetic field distribution above the device under test (DUT). It is frequently used for the precise detection and location of electromagnetic interference (EMI) sources at the board or system level, and has received increasing attention in recent years. Near-field scanning technology helps electronic engineers detect and quickly locate EMI sources in electronic devices, enabling them to implement corresponding EMI suppression measures and significantly shorten the development cycle for EMC (electromagnetic compatibility) issues. Furthermore, near-field scanning can be used to reconstruct the time-domain waveform of transient currents and capture the expansion of the entire circuit over time, aiding in the analysis of conducted interference problems in complex electronic systems.

[0004] Currently, near-field scanning detection technology is only used to study the propagation process of electrostatic discharge current in a system, without further investigation of propagation details or its application to the coupling path analysis of conducted interference currents. Traditional near-field scanning methods obtain complete near-field information by uniformly sampling on the scanning plane. Due to the large number of sampling points and the long data storage time at each location, this requires a significant amount of time. To accelerate the near-field scanning process and ensure detection accuracy, current methods employ interpolation using sparse samples selected in the scanning matrix to reduce scanning time. Therefore, the selection of sparse samples is particularly important. Current AI algorithms for selecting sparse scanning samples are complex in terms of hyperparameter selection, and most studies focus on single-frequency detection, which cannot be directly applied to time-domain current waveform distribution measurement. Summary of the Invention

[0005] To address the problems existing in the background technology, the present invention provides a near-field scanning method for detecting conductive coupling paths based on active learning.

[0006] The technical solution adopted in this invention is:

[0007] The near-field scanning method for detecting conductive coupling paths of the present invention includes the following steps:

[0008] Step 1: Set up a near-field scanning detection platform. Place the circuit board under test horizontally in a small microwave anechoic chamber. When the circuit board under test is working, use the near-field scanning detection platform to detect it from directly above the circuit board.

[0009] Step 2: During the near-field scanning detection platform, the detection results are obtained using a near-field scanning method based on active learning. Based on the detection results, the near-field current distribution map of the current on the circuit board under test changes with time in continuous moments is obtained. Based on the near-field current distribution map of continuous moments, the conduction coupling path of the current on the circuit board under test is obtained, thus realizing the detection of the conduction coupling path.

[0010] In step one, the near-field scanning detection platform includes a computer, a microcontroller, a robotic arm, a near-field probe, and an oscilloscope. The near-field probe is mounted at the end of the robotic arm, which is positioned at a fixed height directly above the circuit board under test. The computer is electrically connected to the microcontroller, the robotic arm, and the oscilloscope. The microcontroller is electrically connected to the circuit board under test, and the oscilloscope is electrically connected to the near-field probe. The robotic arm, the near-field probe, and the circuit board under test are all located in a small microwave anechoic chamber. The detection process is conducted within a small microwave anechoic chamber to avoid external electromagnetic interference. During the detection process, the computer automatically controls all instruments and processes the data. Various types of test probes can be selected based on different test environments.

[0011] In step one, when the circuit board under test is working, a near-field scanning detection platform is used to detect the circuit board directly above it. Specifically, the computer first initializes the connection devices to ensure control of each device and sets the working status of each device. Then, the computer controls the microcontroller to generate a trigger signal and inject it into the circuit board under test to drive it to work normally. Then, the computer controls the robotic arm to move the near-field probe to a fixed height above the circuit board under test for detection. The robotic arm is used to fix and accurately position the near-field probe. Under the control of the computer, it moves continuously to the detection position. The near-field probe generates an induced voltage and transmits it to the computer via an oscilloscope to obtain the detection result.

[0012] The detection result is specifically the time-domain waveform of the induced voltage generated by the near-field probe.

[0013] In step two, during the near-field scanning detection platform operation, a near-field scanning method based on active learning is used to obtain the detection results, as detailed below:

[0014] First, a planar area at a fixed height directly above the circuit board under test is designated as the scanning detection area. Based on the robotic arm's step size, N uniformly spaced N values ​​are obtained within this scanning detection area. a The number of scan points is N. aN0 scan points are selected at uniform intervals from the given scan point locations to construct a scan point dataset. Then, a near-field probe is used to scan the N0 scan point locations in the dataset. A radial basis function (RBF) interpolation method is then used to select a scan point location to be scanned and add it to the scan point dataset. This scanning process is repeated until the number of scan point locations added to the dataset reaches the preset algorithm scan point number N. q Then stop scanning; use an oscilloscope to measure the scanned N0+N values ​​generated by the near-field probe. q The detection results at each scan point location are input into the computer, and the computer, based on the scanned N0+N... q The detection results at each scan point location are interpolated to obtain N within the scanned detection area for each unscanned scan point location. a The detection results of each scan point location. The total number of scan point locations N in the scan detection area. a Greater than the preset number of scan points N q The sum of the number of scan point positions N0.

[0015] The scanning and detection process described above involves using a near-field probe to scan and detect the N0 scanning point positions in the scanning point dataset, as detailed below:

[0016] For each scan point in the scan point dataset, a near-field probe is used to detect the induced voltage at consecutive time points. Then, a computer constructs an initial scan distribution map based on the distribution of the induced voltages at each scan point at the same time within the scan detection area. Several initial scan distribution maps are obtained within each consecutive time period. The variance of each induced voltage in each initial scan distribution map is calculated. Simultaneously, each initial scan distribution map is uniformly divided into N distribution sets. The initial scan distribution map with the largest variance is selected from each distribution set. Then, three initial scan distribution maps with the largest variance from the selected N initial scan distribution maps are selected as the predicted scan distribution map. These three frames represent three time points. Then, the radial basis function (RBF) interpolation method is used to continue cyclic scanning. Specifically, four prediction models—linear interpolation, cubic spline interpolation, thin plate spline interpolation, and quintic spline interpolation—are used to predict the N values ​​in each frame of the predicted scan distribution map. aThe induced voltages at unscanned scan point locations (excluding those in the scan point dataset) are interpolated to obtain prediction results. Then, the uncertainty of each prediction result for the induced voltage at each unscanned scan point location is calculated. The unscanned scan point location with the largest uncertainty is selected as the next scan point location to be scanned and added to the scan point dataset. This scanning and detection process continues cyclically until the number of scan point locations added to the scan point dataset reaches the preset algorithm scan point number N. q The scanning was then stopped, and N0+N generated by the near-field probe was finally obtained. q The induced voltage at each scan point location.

[0017] The uncertainty of each predicted result for the induced voltage at unscanned scan point locations is calculated specifically by using four prediction models of the radial basis function (RBF) interpolation method for each predicted scan distribution map. These models calculate the predicted induced voltage at each unscanned scan point location using the induced voltage at each scanned scan point location in the predicted scan distribution map. For each unscanned scan point location, each prediction model generates a prediction result. The variance of the four prediction results is calculated, and then the average of the variances obtained from the three predicted scan distribution maps is used as the uncertainty of each predicted result for the induced voltage at the unscanned scan point location. The higher the degree of inconsistency between the prediction results of the various models, the higher the uncertainty at that location, and therefore the higher its search value.

[0018] In step two, the final obtained N0+N q The induced voltage at each scanned point is input into the computer via an oscilloscope. Linear interpolation is then performed on the unscanned points within the scanned detection area to obtain the N values ​​within the scanned detection area. a The induced voltage at each scan point location.

[0019] In step two, a near-field current distribution map showing the change of current on the circuit board under test over time is obtained based on the detection results, i.e., based on the N values ​​within the scanning detection area over time. a The induced voltage at each scanning point is used to obtain the current through the mathematical conversion relationship between induced voltage and current. Finally, the near-field distribution map of the current on the circuit board under test is obtained in continuous time. The current flow video is generated based on the near-field distribution map of the current on the circuit board under test in continuous time. The conduction coupling path of the current on the circuit board under test is obtained based on the current flow in the current flow video.

[0020] The beneficial effects of this invention are:

[0021] 1. This invention utilizes an active machine learning method to achieve more efficient near-field scanning detection. The algorithm calculates the more valuable scanning positions among the test points and concentrates the scanning points in areas with strong current. Thus, the actual current changes of the test device are accurately restored using the scanning results of sparse points, which greatly improves the detection efficiency.

[0022] 2. Coupling path analysis has always been a crucial part of electromagnetic interference (EMI) analysis in circuit design, especially in cases of extremely complex circuit structures. Circuit analysis and manual single-point probing alone are insufficient to determine the EMI coupling path. Near-field scanning detection reconstructs current changes on the circuit board, enabling rapid analysis of the coupling path of conducted interference currents on the device under test (DUT), thus meeting the EMI detection requirements of practical engineering projects. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the near-field scanning detection platform based on active learning according to the present invention;

[0024] Figure 2 This is a complete scan of Hx and Hy, as well as a synthesized near-field distribution map, in an embodiment of the present invention.

[0025] Figure 3 The algorithm scans the distribution of Hx and Hy points and the uniformly selected point distribution in this embodiment of the invention.

[0026] Figure 4 This is a composite near-field distribution map of all scans, algorithm scans, and uniform scans in an embodiment of the present invention;

[0027] Figure 5 This is a schematic diagram illustrating the mean square error of the algorithm scanning compared to uniform scanning and the entire scanning in this embodiment of the invention.

[0028] Figure 6 The images show Hx, Hy, and the synthesized near-field distribution map obtained by scanning using an algorithm after severing the coupling path in this embodiment of the invention. Detailed Implementation

[0029] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0030] like Figure 1 As shown, the near-field scanning method for detecting conductive coupling paths according to the present invention includes the following steps:

[0031] Step 1: Set up a near-field scanning detection platform. Place the circuit board under test horizontally in a small microwave anechoic chamber. When the circuit board under test is working, use the near-field scanning detection platform to detect it from directly above the circuit board.

[0032] In step one, the near-field scanning detection platform includes a computer, a microcontroller, a robotic arm, a near-field probe, and an oscilloscope. The near-field probe is mounted at the end of the robotic arm, positioned at a fixed height directly above the circuit board under test. The computer is electrically connected to the microcontroller, robotic arm, and oscilloscope. The microcontroller is electrically connected to the circuit board under test, and the oscilloscope is electrically connected to the near-field probe. The robotic arm, near-field probe, and circuit board under test are all located in a small microwave anechoic chamber. The detection process is conducted within a small microwave anechoic chamber to avoid external electromagnetic interference. During the detection process, the computer automatically controls all instruments and processes the data. Various types of test probes can be selected based on the different test environments.

[0033] In step one, while the circuit board under test (PCB) is operating, a near-field scanning detection platform is used to detect the PCB directly above it. Specifically, the computer first initializes the connected devices, ensuring control over each device and setting its operating status. Then, the computer controls the microcontroller to generate a trigger signal and injects it into the PCB to drive it to operate normally. Next, the computer controls a robotic arm to move the near-field probe to a fixed height above the PCB for detection. The robotic arm is used to fix and accurately position the near-field probe. Under computer control, it continuously moves to the detection position. The near-field probe generates an induced voltage, which is transmitted to the computer via an oscilloscope to obtain the detection result.

[0034] The detection result is specifically the time-domain waveform of the induced voltage generated by the near-field probe.

[0035] Step 2: During the near-field scanning detection platform, the detection results are obtained using a near-field scanning method based on active learning. Based on the detection results, the near-field current distribution map of the current on the circuit board under test changes with time in continuous moments is obtained. Based on the near-field current distribution map of continuous moments, the conduction coupling path of the current on the circuit board under test is obtained, thus realizing the detection of the conduction coupling path.

[0036] In step two, during the near-field scanning detection platform's detection, an active learning-based near-field scanning method is used to obtain the detection results, as detailed below:

[0037] First, a planar area at a fixed height directly above the circuit board under test is designated as the scanning detection area. Based on the robotic arm's step size, N uniformly spaced N values ​​are obtained within this scanning detection area. a The number of scan points is N. aN0 scan points are selected at uniform intervals from the given scan point locations to construct a scan point dataset. Then, a near-field probe is used to scan the N0 scan point locations in the dataset. A radial basis function (RBF) interpolation method is then used to select a scan point location to be scanned and add it to the scan point dataset. This scanning process is repeated until the number of scan point locations added to the dataset reaches the preset algorithm scan point number N. q Then stop scanning; use an oscilloscope to measure the scanned N0+N values ​​generated by the near-field probe. q The detection results at each scan point location are input into the computer, and the computer, based on the scanned N0+N... q The detection results at each scan point location are interpolated to obtain N within the scanned detection area for each unscanned scan point location. a The detection results of each scan point location. The total number of scan point locations N in the scan detection area. a Greater than the preset number of scan points N q The sum of the number of scan point positions N0.

[0038] The scanning and detection process is performed by using a near-field probe to scan and detect the N0 scanning point locations in the scanning point dataset, as detailed below:

[0039] For each scan point in the scan point dataset, a near-field probe is used to detect the induced voltage at consecutive time points. Then, a computer constructs an initial scan distribution map based on the distribution of the induced voltages at each scan point at the same time within the scan detection area. Several initial scan distribution maps are obtained within each consecutive time period. The variance of each induced voltage in each initial scan distribution map is calculated. Simultaneously, each initial scan distribution map is uniformly divided into N distribution sets. The initial scan distribution map with the largest variance is selected from each distribution set. Then, three initial scan distribution maps with the largest variance from the selected N initial scan distribution maps are selected as the predicted scan distribution map. These three frames represent three time points. Then, the radial basis function (RBF) interpolation method is used to continue cyclic scanning. Specifically, four prediction models—linear interpolation, cubic spline interpolation, thin plate spline interpolation, and quintic spline interpolation—are used to predict the N values ​​in each frame of the predicted scan distribution map. aThe induced voltages at unscanned scan point locations (excluding those in the scan point dataset) are interpolated to obtain prediction results. Then, the uncertainty of each prediction result for the induced voltage at each unscanned scan point location is calculated. The unscanned scan point location with the largest uncertainty is selected as the next scan point location to be scanned and added to the scan point dataset. This scanning and detection process continues cyclically until the number of scan point locations added to the scan point dataset reaches the preset algorithm scan point number N. q The scanning was then stopped, and N0+N generated by the near-field probe was finally obtained. q The induced voltage at each scan point location.

[0040] The uncertainty of each predicted result for the induced voltage at unscanned scan point locations is calculated. Specifically, for each predicted scan distribution map, four prediction models using the radial basis function (RBF) interpolation method are used. The predicted induced voltage at each scanned scan point location in the predicted scan distribution map is used to calculate the predicted induced voltage at each unscanned scan point location. For each unscanned scan point location, each prediction model generates a prediction result. The variance of the four prediction results is calculated, and then the average of the variances obtained from the three predicted scan distribution maps is used as the uncertainty of each predicted result for the induced voltage at the unscanned scan point location. The higher the degree of inconsistency between the prediction results of the various models, the higher the uncertainty at that location, and therefore the higher its query value.

[0041] In step two, the final obtained N0+N q The induced voltage at each scanned point is input into the computer via an oscilloscope. Linear interpolation is then performed on the unscanned points within the scanned detection area to obtain the N values ​​within the scanned detection area. a The induced voltage at each scan point location.

[0042] In step two, a near-field current distribution map showing the change of current on the circuit board under test over time is obtained based on the detection results. This map is derived from the N values ​​within the scanned detection area over consecutive time intervals. a The induced voltage at each scanning point is used to obtain the current through the mathematical conversion relationship between induced voltage and current. Finally, the near-field distribution map of the current on the circuit board under test is obtained in continuous time. The current flow video is generated based on the near-field distribution map of the current on the circuit board under test in continuous time. The conduction coupling path of the current on the circuit board under test is obtained based on the current flow in the current flow video.

[0043] A specific detection embodiment of the present invention is a dual-channel high-voltage discharge board. The embodiment includes two discharge circuits, each consisting of a trigger circuit, a control switch, and an energy storage capacitor. The high-voltage converter module outputs high voltage to charge the two energy storage capacitors. Subsequently, an external microcontroller injects a trigger signal into the trigger circuit, causing the control switch to open and allowing the capacitor to complete a transient discharge. During normal operation, the two discharge circuits are controlled independently by two trigger signals spaced a certain time apart, thus the second circuit discharges with a delay. However, experiments have shown that when the circuit operating voltage is high, the first discharge circuit generates severe electromagnetic interference to the trigger circuit of the second circuit during discharge, causing the energy storage capacitor in the second circuit to discharge prematurely. At the moment of discharge in the first discharge circuit, the electromagnetic interference affects the normal operation of the electronic components in the second circuit through conducted interference, leading to abnormal discharge system conditions and threatening the safety of the control circuit. Therefore, it is necessary to analyze the electromagnetic interference coupling path of the circuit board and take corresponding measures to suppress the interference.

[0044] This active learning-based near-field scanning method was used to scan a dual-channel high-voltage discharge board. A magnetic field probe was selected for detection based on the testing requirements. The probe was placed 2.5mm above the circuit board. Due to the complex structure of the circuit board under test, and preliminary assessment confirming that the conduction of electromagnetic interference might be related to the ground plane, the scanning area was set at a local location on the ground plane. The scanning area was 165mm*95mm. The coordinates of the area under test were obtained and set as the scanning range of the robotic arm. The interval between two points within the scanning range, i.e., the scanning step size, was set to 2.5mm, resulting in a total of 2613 scanning positions. The initial number of points was set to 50, and the algorithm's scanning point count was set to 160. First, 50 scanning points are uniformly selected within the scanning area for detection. Then, the three frames with the largest test variance are calculated from all frames. Next, four interpolation models are used to predict the values ​​of the unscanned points, and the variance of the prediction results of the four models is calculated. The position with the largest average variance of the three frames is selected as the next scanning point. This process is repeated until the number of points selected by the algorithm reaches 160, at which point the scanning stops. The interpolation function is then used to predict the values ​​of the remaining unscanned points, which reconstructs the temporal detection results within the entire scanning area and generates a video of the current propagation over time. Finally, the conduction coupling path is determined based on the detection results.

[0045] To verify the accuracy of the algorithm's scanning, all points in the scanned area were probed, such as... Figure 2As shown, from left to right, the images represent the Hx, Hy, and synthesized near-field distribution of the 15000th frame of the entire scan. Hx refers to the horizontal magnetic field amplitude, and Hy refers to the vertical magnetic field amplitude. The synthesized near-field distribution compares Hx and Hy at the same location, selecting the value with the larger absolute value as the near-field value at that location. The propagation path of the coupled current is clearly visible in the synthesized near-field distribution. Furthermore, to verify the superiority of the algorithmic scanning, a uniform scan was used for comparison, selecting the same number of points as the algorithmic scan and performing interpolation calculations in the same manner. Figure 3 As shown, the difference between the distribution of points selected by the algorithm for scanning Hx and Hy and that of uniform scanning points is that the scanning points selected by the algorithm have a clear tendency, and the scanning points are concentrated in the location of strong current. This is the detection area that should be the focus of the detection process. In contrast, uniform scanning selects points sparsely and evenly across the entire scanning area without any tendency.

[0046] Detection results as follows Figure 4 As shown, to demonstrate the universality of the near-field detection comparison results of different scanning methods at different times, frames 14100 and 15000 were selected for detection comparison. It is clearly visible from the figure that the detection results of the algorithmic scanning at both times are almost consistent with the full scan, while the results of the uniform scanning are not very clear, and the detected current coupling path is not complete. The mean square error of the algorithmic scanning and uniform scanning compared to the full scan is calculated, and the results are as follows... Figure 5 As shown, it can be seen that with the increase of the number of detection points, the mean square error of both algorithm scanning and uniform scanning decreases, but the error of algorithm scanning is almost always better than that of mean square error. This is also related to... Figure 4 The detection results are consistent, and the error between the detection results of the algorithm scan and the actual detection is smaller.

[0047] To demonstrate the superior detection efficiency of this method, the times for full scan, uniform scan, and algorithmic scan were statistically analyzed, as shown in Table 1. Detecting the coupling path requires scanning in both the Hx and Hy directions. For full scan, the number of detection points and scan time are the same for both scans. For uniform scan, the detection time is significantly reduced due to the decrease in the number of sampling points. For algorithmic scan, although some computation time is consumed in selecting scan points, the total detection time is much shorter than full scan, only 1 / 12 of the full scan time, given the significantly reduced number of scan points. Furthermore, the detection accuracy of algorithmic scan is higher than that of uniform scan. While maintaining a very small difference in detection accuracy compared to full scan, this method greatly improves detection efficiency.

[0048] Table 1

[0049]

[0050] After determining the coupling path of the dual-channel high-voltage discharge board, the coupling path can be cut off to suppress the interference of the conducted coupling current on the second discharge circuit. After cutting off, the active learning-based near-field scanning method is used again to detect it in both the Hx and Hy directions. The detection results of the 15000th frame and the synthesized near-field distribution map are shown below. Figure 6 As shown in the figure, the original conduction current was effectively suppressed after the coupling path was cut off, which proves the reliability of this method for detecting the coupling path of a dual-channel high-voltage discharge board.

[0051] The above example illustrates the detection implementation of this invention on a dual-channel high-voltage discharge board. As can be seen from this implementation, this invention enables efficient detection of board-level current coupling paths. It is a near-field scanning method based on active learning. By selecting only a small number of detection points within the scanning range, the time-domain near-field distribution across the entire area can be quickly acquired, allowing analysis of the current coupling path. This enables the implementation of corresponding electromagnetic interference suppression measures to improve circuit design, thus meeting the research needs of practical electronics fields.

Claims

1. A detection method for near-field scanning to obtain conductive coupling paths based on active learning, characterized in that: The method includes the following steps: Step 1: Set up a near-field scanning detection platform. Place the circuit board under test horizontally in the microwave anechoic chamber. When the circuit board under test is working, use the near-field scanning detection platform to detect it from directly above the circuit board under test. Step 2: During the near-field scanning detection platform, the detection results are obtained using a near-field scanning method based on active learning. Based on the detection results, the near-field current distribution map of the current on the circuit board under test changes with time in continuous time intervals is obtained. Based on the near-field current distribution map of continuous time intervals, the conduction coupling path of the current on the circuit board under test is obtained, thus realizing the detection of the conduction coupling path. In step one, the near-field scanning detection platform includes a computer, a microcontroller, a robotic arm, a near-field probe, and an oscilloscope. The near-field probe is installed at the end of the robotic arm, and the end of the near-field probe is located at a fixed height directly above the circuit board under test. The computer is electrically connected to the microcontroller, the robotic arm, and the oscilloscope. The microcontroller is electrically connected to the circuit board under test, and the oscilloscope is electrically connected to the near-field probe. The robotic arm, near-field probe, and circuit board under test are all located in a microwave anechoic chamber. In step two, during the near-field scanning detection platform operation, a near-field scanning method based on active learning is used to obtain the detection results, as detailed below: First, a planar area at a fixed height directly above the circuit board under test is designated as the scanning detection area. Based on the robotic arm's step size, N uniformly spaced N values ​​are obtained within this scanning detection area. a The number of scan points is N. a N0 scan points are selected at uniform intervals from the given scan point locations to construct a scan point dataset. Then, a near-field probe is used to scan the N0 scan point locations in the dataset. A radial basis function (RBF) interpolation method is then used to select a scan point location to be scanned and add it to the scan point dataset. This scanning process is repeated until the number of scan point locations added to the dataset reaches the preset algorithm scan point number N. q Then stop scanning; use an oscilloscope to measure the scanned N0+N values ​​generated by the near-field probe. q The detection results at each scan point location are input into the computer, and the computer, based on the scanned N0+N... q The detection results at each scan point location are interpolated to obtain N within the scanned detection area for each unscanned scan point location. a Detection results of each scan point location; The scanning and detection process described above involves using a near-field probe to scan and detect the N0 scanning point positions in the scanning point dataset, as detailed below: For each scan point in the scan point dataset, a near-field probe is used to detect the induced voltage at each scan point location at consecutive time intervals. Then, a computer constructs an initial scan distribution map based on the distribution of the induced voltages obtained at each scan point location at the same time interval within the scan detection area. Several initial scan distribution maps are obtained within each consecutive time interval. The variance of each induced voltage in each initial scan distribution map is calculated. Simultaneously, each initial scan distribution map is uniformly divided into N distribution map sets. The initial scan distribution map with the largest variance is selected from each distribution map set. Then, the three initial scan distribution maps with the largest variance from the selected N initial scan distribution maps are selected as the predicted scan distribution maps. Then, the radial basis function (RBF) interpolation method is used to continue cyclic scanning. Specifically, four prediction models—linear interpolation, cubic spline interpolation, thin plate spline interpolation, and quintic spline interpolation—are used to predict the N values ​​in each predicted scan distribution map. a The induced voltages at unscanned scan point locations (excluding those in the scan point dataset) are interpolated to obtain prediction results. Then, the uncertainty of each prediction result for the induced voltage at each unscanned scan point location is calculated. The unscanned scan point location with the largest uncertainty is selected as the next scan point location to be scanned and added to the scan point dataset. This scanning and detection process continues cyclically until the number of scan point locations added to the scan point dataset reaches the preset algorithm scan point number N. q The scanning was then stopped, and N0+N generated by the near-field probe was finally obtained. q The induced voltage at each scan point position; The uncertainty of each prediction result for the induced voltage at the unscanned scan point position is calculated as follows: For each predicted scan distribution map, four prediction models using the radial basis function (RBF) interpolation method are used. The induced voltage at each scanned scan point position in the predicted scan distribution map is used to calculate the prediction result for the induced voltage at each unscanned scan point position in the predicted scan distribution map. For each unscanned scan point position, each prediction model generates a prediction result. The variance of the four prediction results is calculated, and then the average of the variances obtained from the three predicted scan distribution maps is calculated as the uncertainty of each prediction result for the induced voltage at the unscanned scan point position. Finally, the uncertainty of each prediction result for the induced voltage at each unscanned scan point position is obtained.

2. The detection method for obtaining conductive coupling paths based on active learning in near-field scanning according to claim 1, characterized in that: In step one, when the circuit board under test is working, a near-field scanning detection platform is used to detect the circuit board directly above it. Specifically, a computer-controlled microcontroller generates a trigger signal and injects it into the circuit board under test to drive it to work normally. Then, the computer controls a robotic arm to move the near-field probe to a fixed height above the circuit board under test for detection. The near-field probe generates an induced voltage and transmits it to the computer via an oscilloscope to obtain the detection result.

3. The detection method for obtaining conductive coupling paths based on active learning in near-field scanning according to claim 2, characterized in that: The detection result is specifically the time-domain waveform of the induced voltage generated by the near-field probe.

4. The detection method for obtaining conductive coupling paths based on active learning in near-field scanning according to claim 1, characterized in that: In step two, the final obtained N0+N q The induced voltage at each scanned point is input into the computer via an oscilloscope. Linear interpolation is then performed on each unscanned point in the scanned detection area to obtain the N values ​​within the scanned detection area. a The induced voltage at each scan point location.

5. The detection method for obtaining conductive coupling paths based on active learning in near-field scanning according to claim 4, characterized in that: In step two, a near-field current distribution map showing the change of current on the circuit board under test over time is obtained based on the detection results, i.e., based on the N values ​​within the scanning detection area over time. a The induced voltage at each scanning point is used to obtain the current, and finally the current near-field distribution map on the circuit board under test is obtained in a continuous time. The current flow video is generated based on the current near-field distribution map on the circuit board under test in a continuous time. The conduction coupling path of the current on the circuit board under test is obtained based on the current flow in the current flow video.

Images