A Fast Near-Field Scanning Method for Emitters with Active Learning

Abstract

The present invention discloses a method for fast near-field scanning of radiation sources by active learning. First, the radiation field values of randomly scanned points are obtained, and then the radiation field values of the remaining scanned points are interpolated by various interpolation methods. The scanned point with the largest variance is selected, and the robotic arm is used to move the probe to obtain the radiation field value corresponding to this scanned point. By iterating the above steps multiple times, multiple scanned points can be obtained, quickly, accurately and effectively meeting the requirements of the user. The method of the present invention is simple to operate, has good real-time performance, high scanning efficiency and good robustness, and has obvious advantages in saving the number of scanned points and scanning time.

Description

Technical Field

[0001] The present invention relates to an electromagnetic near-field scanning method in the fields of electromagnetic compatibility and artificial intelligence, and more specifically to a fast near-field scanning method for radiation sources with active learning. Background Art

[0002] The high integration of electronic products leads to serious electromagnetic interference problems, which will affect the normal operation of electronic products. In the industrial and academic fields, near-field scanning is an effective electromagnetic diagnosis method.

[0003] In a typical near-field scanning system, a robotic arm controls a near-field probe to move above the radiation source to be measured along a certain path. The electric or magnetic field information above the radiation source to be measured is obtained by a receiver (usually a spectrum analyzer or a vector network analyzer) and saved in a computer. The time-consuming of the whole scanning process includes three parts: the time for the robotic arm to move, the time for the receiver to obtain the scanning field, and the time for the computer to store the field information. When the number of scanning points is large, the scanning time is very long, which greatly reduces the scanning efficiency. Summary of the Invention

[0004] In order to solve the problem of too long scanning time in the above-mentioned near-field scanning process, the present invention proposes a fast near-field scanning method for radiation sources based on active learning, which reduces the number of scanning points and directly improves the scanning efficiency.

[0005] The present invention adopts the following technical solutions:

[0006] It includes step S1 of detecting the radiation field value at the initial scanning points on the scanning plane with a probe moved by a robotic arm for a radiation source;

[0007] It includes step S2 of continuously using the robotic arm to move the probe to perform near-field scanning detection on the remaining scanning points in an active learning manner according to the radiation field values of the initial scanning points.

[0008] The specific content of step S1 is as follows:

[0009] For a given radiation source, a probe is installed at the end of the robotic arm. The robotic arm is moved to scan and move the probe on the scanning plane. There are scanning points on the scanning plane. The robotic arm is moved to place the probe at a scanning point and detect the radiation field value; a group of initial scanning points are randomly obtained by a computer. The robotic arm is moved to place the probe at each initial scanning point for detection, and the radiation field values of each initial scanning point are obtained. The initial scanning points are added to the scanned scanning points.

[0010] The radiation source is specifically an active radiation component such as an electronic product or an electronic system. The radiation field value is an electric field or a magnetic field. The scanning plane is usually a plane.

[0011] Specifically, in step S2, the probe is detected by each round of mobile robotic arm in the following manner:

[0012] S21. In the implementation case, in the current round, taking the currently undetected scan points as the unscanned points, different preliminary radiation field values of the unscanned points are obtained by respectively interpolating the radiation field values of the detected scan points through different interpolation functions. One radiation field value is obtained by processing with each interpolation function. The variance of the respective preliminary radiation field values of the unscanned points is calculated, and the unscanned point with the largest variance is selected.

[0013] By extracting the unscanned point with the largest variance, it is possible to quickly find the scan points with high uncertainty of radiation field values and high importance. The radiation field values of these unscanned points contribute greatly to the richness of the entire radiation field, and the radiation field values contain a lot of information about the radiation source.

[0014] S22. The robotic arm moves the probe to the position of the unscanned point with the largest variance for scanning detection to obtain the radiation field value of the unscanned point, and adds the unscanned point to the detected scan points.

[0015] S23. Steps S21 - S22 are continuously repeated for multiple rounds until the number of scan points detected by moving the probe with the robotic arm reaches the number required by the user. Finally, using the detected scan points and interpolation functions, a scan image of the complete scanning plane is obtained. After multiple cycles, new scan points are continuously selected until the scan image meets the accuracy requirements.

[0016] The different interpolation functions mentioned above are different types of interpolation functions, or the same type of interpolation functions with different function parameters.

[0017] The different interpolation functions are radial basis functions (RBFs) with different kernel functions. It can also be replaced with other fast interpolation functions, not limited to this.

[0018] For different radiation sources, the number of different interpolation functions can be adjusted, and the radiation field values of other scan points are obtained from the radiation field values of the detected scan points using different interpolation functions.

[0019] The active learning method provided by the present invention first randomly scans the radiation field values of points, then interpolates the radiation field values of the remaining scan points through multiple interpolation methods, selects the scan points with a relatively large variance in the results of multiple interpolation methods, and the robotic arm moves the probe to obtain the radiation field value of this point; iterating the above steps multiple times quickly, accurately, and effectively obtains the scan points required by the user.

[0020] The present invention calculates the variance of unscanned points through different interpolation functions, directly and in real time judges and updates the control scanning based on the variance of unscanned points, achieving the purpose of efficient near-field scanning through the active learning technology in artificial intelligence, and having the advantages of good real-time performance, high scanning efficiency, good robustness, and simple operation.

[0021] The present invention screens out most of the scanned points that do not carry important information of the radiation source in the complete scanned field map. Scanning only the points carrying important field information is sufficient to reflect the characteristics of the radiation source, and the scanning time can be greatly reduced.

[0022] Moreover, the method of the present invention is actively learning. Through artificial intelligence, the probe of the robotic arm calculates the positions of the points to be scanned next while scanning (most points do not need to be scanned), and all independently decides to perform scanning, forming a method that efficiently saves scanning time and ensures that the field map of some points can accurately reflect the characteristics of the radiation source.

[0023] Compared with the existing near-field scanning technologies, the beneficial effects of the present invention are as follows:

[0024] The method of the present invention can obtain accurate radiation field values at positions where the radiation field values change significantly using a near-field scanning system, and obtain radiation field values using an interpolation method at other positions. Compared with the traditional complete scanning method, the near-field scanning efficiency can be greatly improved.

[0025] The present invention has high processing efficiency, can be used for real-time near-field scanning, and the method has good robustness and does not require users to adjust the setting parameters too much.

[0026] Compared with the traditional full-scanning method or random point selection scanning, the method proposed by the present invention has obvious advantages in saving the number of scanned points and scanning time. Brief Description of the Drawings

[0027] Figure 1 is the flowchart of the embodiment of the present invention;

[0028] Figure 2 is an example of the radiation source for near-field scanning in the embodiment of the present invention;

[0029] Figure 3 is the complete magnetic field amplitude value of the radiation source obtained by the robotic arm moving the probe for scanning in the embodiment of the present invention;

[0030] Figure 4 are the initially selected 100 scanned points in the embodiment of the present invention;

[0031] Figure 5 is the variance distribution diagram of each scanned point on the entire scanned surface calculated according to the initially scanned points and 4 interpolation functions in the embodiment of the present invention;

[0032] Figure 6is the un-scanned point with the largest variance selected in the embodiment of the present invention;

[0033] Figure 7 are 510 scanned points selected by the method proposed in the embodiment of the present invention;

[0034] Figure 8 is the amplitude distribution diagram of the complete scanned plane radiation field after interpolation through 510 scanned points in the embodiment of the present invention;

[0035] Figure 9 are 510 scanned points randomly selected in the embodiment of the present invention;

[0036] Figure 10 is the amplitude distribution diagram of the complete scanned plane radiation field after interpolation through 510 randomly selected scanned points in the embodiment of the present invention;

[0037] Figure 11 is the comparison diagram of the mean square error between this method and random point selection. The mean square error is calculated from the mean square error between the complete scanned plane radiation field after interpolation through the selected scanned points and Figure 3 the complete radiation field in. Detailed implementation manners

[0038] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. Therefore, the detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but only represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0039] As Figure 1 shown, the embodiments of the present invention and their implementation processes are as follows:

[0040] As Figure 2 shown, in this embodiment, the radiation source for near-field scanning is the "C"-shaped metal drawn on the PCB board. The scanning surface is 3 mm away from the upper surface of the "C"-shaped metal, the size of the scanning surface is 180 mm × 180 mm, and the scanning interval is 3 mm. Therefore, the total number of scanned points is 61×61 = 3721. The amplitude value distribution diagram of the complete magnetic field of this radiation source is as Figure 3 shown.

[0041] First, the computer randomly generates 100 scanning point positions. The robotic arm controls the near-field probe to move to these 100 scanning point positions for detection, and a spectrometer is used to obtain the magnetic field amplitude values of the corresponding 100 scanning points. These 100 random scanning point positions are as shown in Figure 4 shown.

[0042] Based on the detected magnetic field amplitude values of the 100 scanning points, for the remaining 3621 scanning points, interpolation is performed using four kernel functions of the radial basis function (RBF), namely linear, cubic, thin_plate_spline, and quintic, according to the magnetic field amplitude values of the 100 scanning points to obtain the interpolation results of the four magnetic field amplitude values, and the variance of the four interpolation results of each scanning point is calculated. The distribution of the variance values of all scanning positions on the scanning plane is as shown in Figure 5 shown. Subsequently, the scanning point with the largest variance is selected, as shown in Figure 6 shown, and it is considered that the uncertainty of the radiation field value at this point is high and the radiation field value changes significantly.

[0043] In this embodiment, each time the unscanned point with the largest variance is selected. After multiple iterations, the total number of scanning points obtained by moving the probe with the robotic arm is 510. The magnetic field amplitude values composed of the 510 scanning points are as shown in Figure 7 shown. For the scanning points whose accurate magnetic field amplitude values have not been accurately detected, the RBF linear interpolation function is used to perform interpolation based on the scanning points with detected magnetic field amplitude values to obtain the magnetic field amplitude values of these scanning points, and thus the magnetic field amplitude values of the entire scanned surface are all obtained, as shown in Figure 8 shown. This magnetic field amplitude distribution is very close to the complete magnetic field amplitude distribution of the radiation source in Figure 3 . The fast near-field scanning method proposed by the present invention can reduce the number of scanning points to 1 / 7, and the scanning points are distributed in places where the radiation field value changes significantly, greatly improving the scanning efficiency. Figure 9 shows the distribution diagram of selecting 510 scanning points by the method of randomly selecting points. It can be seen that these 510 scanning points are almost evenly distributed. The complete magnetic field distribution after interpolation by these 510 random scanning points is as shown in Figure 10 shown. It can be seen that there is a large difference between this distribution and the complete magnetic field amplitude distribution of the radiation source in Figure 3 .

[0044] To further quantitatively illustrate the advantages of the method proposed by the present invention, Figure 11 the advantages of the method of the present invention and randomly selecting points in terms of mean square error are compared. This mean square error is calculated from the radiation field of the complete scanned plane after interpolation by the selected scanning points and Figure 3The mean square error between the complete radiation fields in [the specific context] is calculated. It can be seen that as the number of points increases, the mean square error of the present invention is much smaller than that of randomly selected points, which proves the effectiveness of the method proposed by the present invention. It can find the scanning area with more important information with fewer scanning points, greatly improving the near-field scanning efficiency. At the same time, this method has high robustness.

[0045] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A rapid near-field scanning method for radiation sources with active learning, characterized in that, The method includes the following steps: It includes step S1 of detecting the radiation field value at the initial scan points on the scan surface with a probe moved by a robotic arm for a radiation source; It includes step S2 of using the robotic arm to move the probe in real time according to the radiation field values of the initial scan points to perform near-field scan detection on the remaining scan points in an active learning manner; The specific content of step S2 is as follows: S21. In the current round, take the currently undetected scan points as the un-scanned points. Use the radiation field values of the detected scan points to perform interpolation processing through different interpolation functions respectively to obtain different preliminary radiation field values of the un-scanned points. Then calculate the variance of each preliminary radiation field value of the un-scanned points, and select the un-scanned point with the largest variance; S22. Move the probe to the position of the un-scanned point with the largest variance through the robotic arm to perform scan detection to obtain the radiation field value of this un-scanned point, and add this un-scanned point to the detected scan points; S23. Continuously repeat steps S21 to S22 for multiple rounds of processing until the number of scan points detected by moving the probe through the robotic arm reaches the number required by the user. Finally, use the detected scan points and the interpolation function to obtain the scan image of the complete scan plane.

2. The fast near-field scanning method for a radiation source with active learning according to claim 1, wherein: The specific content of step S1 is as follows: For a given radiation source, move the robotic arm to scan and move the probe on the scan surface. There are scan points on the scan surface. Randomly obtain a set of initial scan points. Move the robotic arm to position the probe at each initial scan point for detection to obtain the radiation field values of each initial scan point.

3. The fast near-field scanning method of a radiation source with active learning according to claim 1, characterized in that: The different interpolation functions are different types of interpolation functions, or the same type of interpolation function with different function parameters.

4. A fast near-field scanning method for radiation sources based on active learning according to claim 1, characterized in that: The different interpolation functions are radial basis functions with different kernel functions.

5. A fast near-field scanning method for radiation sources with active learning according to claim 1, characterized in that: In step S21, select the un-scanned point with the largest variance among the un-scanned points as the point to be detected in this round.

Images