AI-driven chip inductance EMC test abnormal signal analysis method

Abstract

The application discloses an AI driving-based chip inductance EMC test abnormal signal analysis method, relates to the chip test technical field, performs EMC test on the chip inductance, compares, analyzes and processes the obtained test data with the preset standard EMC test data, screens and obtains all abnormal sub-data clusters in the test data, allocates corresponding AI computing power resources to all abnormal sub-data clusters based on respective data amounts, groups the abnormal sub-data clusters based on abnormal types, obtains a plurality of abnormal signal sets, selects one set item in each abnormal signal set as an analysis target, obtains corresponding signal characteristics and signal abnormal details, compares the signal characteristics of other set items with the analysis target, selects data multiplexing or data analysis based on the comparison result, and the application guarantees the safety of data in the transmission process, realizes accurate subsequent abnormal signal analysis, and improves the analysis efficiency of abnormal signals.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of chip testing, and in particular to an AI-driven chip inductance EMC test abnormal signal analysis method. BACKGROUND

[0002] As a core passive component in electronic devices, the electromagnetic compatibility of chip inductance directly affects the performance of the whole machine. In electromagnetic compatibility testing, chip inductance is both a potential electromagnetic interference emission source and a sensitive component susceptible to external interference. EMC testing mainly relies on standardized radiation emission and conducted emission testing to determine compliance by collecting and analyzing signals in specific frequency bands.

[0003] The existing chip inductance EMC test analysis method has obvious deficiencies. The test data lacks a security protection mechanism during transmission, affecting the accuracy of subsequent abnormal signal analysis. In addition, there is a lack of reuse mechanism for similar abnormal signals, resulting in repeated analysis of abnormal signals and unnecessary increase in computational overhead, which affects signal analysis efficiency. These deficiencies have become urgent problems to be solved. SUMMARY

[0004] The purpose of the present application is to provide an AI-driven chip inductance EMC test abnormal signal analysis method to solve the problems in the background art.

[0005] To achieve the above purpose, the present application provides the following technical solution: an AI-driven chip inductance EMC test abnormal signal analysis method, comprising the following steps:

[0006] Step S1: Perform EMC testing on the chip inductance, compare and analyze the test data corresponding to the EMC testing with the preset standard EMC test data, and further filter to obtain all abnormal sub-data clusters in the test data;

[0007] Step S2: Assign corresponding AI computing resources to each abnormal sub-data cluster based on its data volume, group all abnormal sub-data clusters based on abnormal types, and obtain several abnormal signal sets;

[0008] Step S3: Select one set item in each abnormal signal set as the analysis target, obtain the corresponding signal features and signal abnormal details, compare the signal features of other set items with the analysis target, and select to perform data reuse or data analysis based on the comparison result.

[0009] In a preferred embodiment, the process of performing EMC testing on the chip inductance, comparing and analyzing the test data corresponding to the EMC testing with the preset standard EMC test data, and further filtering to obtain all abnormal sub-data clusters in the test data includes:

[0010] performing EMC test on the chip inductance based on a preset test procedure, obtaining corresponding test data, processing the test data into a plurality of stage test data, obtaining standard EMC test data and processing the standard EMC test data into a plurality of stage test standard data;

[0011] aligning the plurality of stage test data and the stage test standard data based on time stamp, constructing a plurality of data comparison groups, establishing a liaison path for each data comparison group, the liaison path comprising a plurality of liaison nodes, associating a liaison line with each liaison node;

[0012] uploading the data comparison groups to the virtual machine through the liaison path for comparison analysis, determining whether the data comparison groups exist abnormity, if yes, marking the stage test data in the corresponding data comparison group as an abnormal sub-data cluster, repeating the comparison analysis of the plurality of data comparison groups, and further screening all abnormal sub-data clusters in the test data; if not, no operation is performed.

[0013] In a preferred embodiment, the liaison path, liaison node and liaison line comprise:

[0014] the starting position of the liaison path is used for storing the data comparison groups, and the terminal position is used for connecting the virtual machine, and the data comparison groups transmitted by each liaison path are subjected to path security management;

[0015] the liaison node is arranged at a respective corresponding communication congestion position of each liaison path;

[0016] the liaison line is used for data intercommunication between the liaison nodes on different liaison paths, and a detection point and a notification point are arranged at both ends of the liaison line, the detection point is arranged at the terminal position of the liaison line, and the notification point is arranged at the starting position of the liaison line;

[0017] the detection point and the notification point respectively communicate data with the liaison node.

[0018] In a preferred embodiment, the process of the path security management comprises:

[0019] when the data comparison group is transmitted to the liaison node on the liaison path, the liaison node generates a data empty package which is the same as the data comparison group in data structure, and transfers the data empty package to the notification point on the liaison line, the notification point obtains the data packet address of the data empty package and sends it to the detection point, and the detection point probes the abnormal event on the current liaison path;

[0020] Adapt corresponding path defense measures based on the type of abnormal event, return the abnormal event and the path defense measures accompanying data packet addresses to the notification point, when transmitting data comparison groups under the current contact path, solve the abnormal event on the contact path by executing the path defense measures, and restore the corresponding contact nodes at both ends of the contact line to normal communication;

[0021] The contact nodes on different contact paths generate their own data empty packets and obtain the abnormal events of their respective data empty packets on the contact path. If the data empty packet of the corresponding contact node and the abnormal event of the corresponding contact path are exactly the same as the contact node that has been adapted, then the notification point of the current contact node traverses to the notification point of the exactly same contact node through the notification point of the current contact node, and a communication path is constructed.

[0022] Transmit the data packet address of the data empty packet of the contact node that has been adapted on the communication path, call the corresponding path defense measures based on the data packet address, and restore the contact nodes at both ends of the current contact path to normal communication. Repeat the management of all contact nodes on each contact path until all contact nodes are restored to normal communication.

[0023] In a preferred embodiment, the AI computing resource corresponding to each data amount is allocated to all abnormal sub-data clusters based on their respective data amounts. The process of grouping a plurality of abnormal signal sets based on abnormal types includes:

[0024] Deploy a cloud server to count the amount of computing resources available for scheduling in the cloud server;

[0025] Transmit all abnormal sub-data clusters to the cloud server, obtain the data amount of each abnormal sub-data cluster, allocate the corresponding AI computing resource based on the data amount of each abnormal sub-data cluster, and count the remaining computing resource amount in the cloud server after deducting the allocated AI computing resource;

[0026] Set a critical expenditure threshold corresponding to the normal operation of the cloud server;

[0027] When the remaining computing resource amount is greater than the critical expenditure threshold, do not perform any operation;

[0028] When the remaining computing resource amount is less than or equal to the critical expenditure threshold, perform edge resource release to increase the computing resource in the cloud server until the remaining computing resource is greater than the critical expenditure threshold;

[0029] Process a plurality of abnormal sub-data clusters by a plurality of edge computing devices, aggregate all edge computing devices in the same abnormal type into an edge device set, and group all abnormal sub-data clusters corresponding to an edge device set into an abnormal signal set.

[0030] In a preferred embodiment, one set item in each abnormal signal set is selected as an analysis target, and the process of obtaining corresponding signal features and signal abnormality details includes:

[0031] In a preferred embodiment, several abnormal sub-data clusters in the abnormal signal set are selected as set items of the corresponding set, and the signal features of any set item are selected as an analysis target for data processing to obtain corresponding signal features. Based on the signal features and the pre-set signal feature parameters, the signal abnormality details of each signal feature are obtained.

[0032] In a preferred embodiment, the signal features of other set items are compared with the analysis target, and the process of selecting data multiplexing or data analysis based on the comparison results includes:

[0033] When the signal feature comparison between the selected analysis target and one set item in the abnormal signal set is completely the same, the signal abnormality details of the corresponding analysis target are directly called;

[0034] When the signal feature comparison is not completely the same, the signal abnormality details of the corresponding analysis target on the same signal frequency band are multiplexed, and the signal abnormality details of the corresponding signal frequency band are obtained based on the respective partial signal features and the set signal feature parameters;

[0035] When the signal feature comparison is completely different, the set item corresponding to the current signal feature is selected as a new analysis target, and the signal abnormality details of the new analysis target after data analysis are obtained. The signal features and the signal abnormality details corresponding to the signal features of the new analysis target and the past analysis target are merged to construct a data multiplexing set, and the data analysis and data multiplexing of the corresponding data of the subsequent other set items are performed through the data multiplexing set.

[0036] In a preferred embodiment, the process of performing data analysis and data multiplexing of the corresponding data of the subsequent other set items through the data multiplexing set includes:

[0037] The signal features of the subsequent other set items and the data multiplexing set are compared, and the signal abnormality details corresponding to the signal features that are completely the same between the different analysis targets in the data multiplexing set and the current set item are selected. The signal abnormality details of the current set item are multiplexed, and the signal features that are not matched with the current set item are matched based on the signal features and the signal feature parameters to obtain additional signal abnormality details. The signal abnormality details obtained by screening and the additional signal abnormality details obtained by matching are summarized as the final signal abnormality details.

[0038] In the above technical solutions, the technical effects and advantages provided by the present application are:

[0039] The application establishes a liaison path provided with liaison nodes and liaison lines for each abnormal sub-data cluster, transmits the abnormal sub-data cluster through the liaison path, pre-identifies abnormal events causing communication congestion in a manner of constructing a data empty packet, and adapts a defense measure for solving the abnormal event, thereby ensuring the safety of EMC test related data in the transmission process, providing a reliable data basis for subsequent signal analysis, comparing results based on signal characteristics, directly data multiplexing analysis results for completely same abnormal signals, and performing partial data multiplexing and data analysis for partially same abnormal signals, thereby reducing redundant calculation and improving the analysis efficiency of abnormal signals. BRIEF DESCRIPTION OF DRAWINGS

[0040] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art based on these drawings.

[0041] Figure 1 The method flowchart of the present application.

[0042] Figure 2 The structural schematic diagram among the liaison path, the liaison node and the liaison line in the present application. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical scheme and advantages of the embodiments of the present application more clear, the technical scheme in the embodiments of the present application will be described clearly and completely below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0044] Please refer to Figure 1 The AI-driven chip inductance EMC test abnormal signal analysis method based on the chip inductance EMC test abnormal signal analysis method shown in the figure, including the following steps:

[0045] Step S1: performing EMC test on the chip inductance, comparing and analyzing the test data corresponding to the EMC test with the preset standard EMC test data, and then screening all abnormal sub-data clusters in the test data;

[0046] Step S2: allocating corresponding AI computing resource to each abnormal sub-data cluster based on the data amount of each abnormal sub-data cluster, grouping all abnormal sub-data clusters based on abnormal types, and obtaining a plurality of abnormal signal sets;

[0047] Step S3: selecting one set item in each abnormal signal set as the analysis target, obtaining the corresponding signal characteristics and signal abnormal details, comparing the signal characteristics of other set items with the analysis target, and selecting to perform data multiplexing or data analysis based on the comparison result.

[0048] It needs to be further explained that, in the specific implementation process, the EMC test is performed on the chip inductor, the EMC test corresponding test data is compared and analyzed with the preset standard EMC test data, and then the process of screening all abnormal sub-data clusters in the test data includes:

[0049] Based on the preset test procedure, the EMC test is performed on the chip inductor, and the test data of the chip inductor after the EMC test is completed is obtained. The test data is processed into several stage test data based on the time stamp in the test process and the test procedure of each stage.

[0050] It needs to be explained that the test scene for EMC test is built, the test scene is a semi-electric wave darkroom or a shielding room, the environmental background noise is set to be at least 6 dB lower than the standard limit value, all related instruments and equipment for EMC test are calibrated, including spectrum analyzer, signal receiver and power amplifier, etc., the transmission path of the signal is calibrated using a standard signal source, according to the preset test procedure, the frequency range, resolution bandwidth, type of detector and scanning time length that meet the current signal transmission requirements are set on the signal receiver;

[0051] The chip inductor or its evaluation board is fixed on a non-conductive test table and powered to purify the power supply and provide standard measurement impedance. The chip inductor is set in the working mode specified by the preset test procedure, and the working state of the chip inductor is set. The specific setting content includes full load condition, light load condition, dynamic load condition and static switch state of the chip inductor.

[0052] The full load condition is used to test the EMC performance of the chip inductor under the maximum allowed power consumption.

[0053] The light load condition is used to evaluate the efficiency of the chip inductor in the low power consumption mode and the intermittent and burst noise that may be generated thereby.

[0054] The dynamic load condition is used to simulate the transient response of the chip inductor when responding to rapidly changing loads, test its power integrity and capture the high frequency noise generated thereby.

[0055] The static switch state is used to test the difference in electromagnetic behavior of the chip inductor in two extreme states of complete conduction (working) and complete shutdown (standby), and to locate standby power consumption and power-on or power-off pulse interference.

[0056] The EMC test includes two categories of emission test and immunity test, and all relevant data corresponding to the emission test and immunity test after execution are taken as test data, taking the emission test as an example, the process of the emission test is that a signal receiver is connected to a signal corresponding measurement port, frequency scanning is performed in a range of 150 kHz-30 MHz, peak and quasi-peak type detectors are used for measurement respectively, if the peak measurement result is lower than the limit value of the quasi-peak, it is passed, otherwise, quasi-peak continues to measure and confirm, and all frequency points and amplitudes close to or exceeding the standard limit value are recorded; similarly, the immunity test includes conducted immunity test and radiated immunity test, and one method of the immunity test is selected and executed according to the actual test requirement of the EMC test;

[0057] The standard EMC test data corresponding to the normal working state of the chip inductance is obtained, and the standard EMC test data is processed into a plurality of stage test standard data based on different test stages corresponding to the test procedure of the EMC test;

[0058] The plurality of stage test data and stage test standard data are aligned based on timestamps, a plurality of data comparison groups are constructed, the plurality of data comparison groups are labeled, the label is marked as i, i.e., i=1, 2, 3, …, n, wherein n is a natural number greater than 0, a liaison path is established for each data comparison group, the liaison path includes a plurality of liaison nodes, and a liaison line is associated with each liaison node;

[0059] The connection relationship among the liaison path, the liaison node and the liaison line, please refer to Figure 2 shown;

[0060] The starting position of the liaison path is used for storing the data comparison group, and the terminal position is used for connecting the virtual machine, so as to upload any one data comparison group to the virtual machine through the liaison path for comparison and analysis, and determine whether each data comparison group is abnormal;

[0061] Wherein, whether the data comparison group is normal is determined by whether the fields of the stage test data and the stage test standard data are consistent, if yes, it is determined to be normal, otherwise, it is determined to be abnormal;

[0062] If yes, the stage test data in the corresponding data comparison group is marked as an abnormal sub-data cluster, the comparison and analysis of the plurality of data comparison groups are repeated, and all abnormal sub-data clusters in the test data are screened; if no, no operation is performed;

[0063] The liaison node is arranged at a communication congestion position corresponding to each liaison path;

[0064] The contact line is used for data intercommunication between contact nodes on different contact paths, and a detection point and a notification point are arranged at both ends of the contact line, wherein the detection point is arranged at the termination position of the contact line, and the notification point is arranged at the starting position of the contact line;

[0065] The detection point and the notification point each perform data communication with the contact nodes;

[0066] The data transmitted on each contact path is managed by path security, and the path security management includes: when the data comparison group is transmitted to the contact node on the contact path, the contact node generates a data empty packet, the data structure of the data empty packet is the same as that of the corresponding transmitted data comparison group, the data empty packet corresponding to the contact node is handed over to the notification point on the contact line, the data packet address corresponding to the data empty packet is obtained by the notification point, and the data packet address is sent to the detection point through the contact line, and the detection point probes the type of abnormal event that affects the safety of the data corresponding to the current data empty packet on the current contact line;

[0067] Based on the type of abnormal event, the corresponding path defense measure is adapted from the preset measure library, and the abnormal event and the path defense measure are returned to the notification point together with the data packet address, and when the corresponding data comparison group on the current contact path is transmitted, the corresponding abnormal event on the contact path is solved by executing the path defense measure, and then the corresponding contact nodes at both ends of the contact line are restored to normal communication;

[0068] Each contact node on different contact paths generates a corresponding data empty packet, obtains the abnormal event of the respective data empty packet on the contact line, and if the data empty packet of the corresponding contact node and the abnormal event of the corresponding contact line are exactly the same as the contact node that has been adapted, then the notification point corresponding to the current contact node is traversed to the notification point corresponding to the exactly same contact node, and a communication path is constructed;

[0069] Wherein, the respective data empty packets of the two contact nodes are the same in data structure, and the types of the abnormal events existing on the respective contact lines are also the same, which means that the two contact nodes are exactly the same;

[0070] The data packet address of the data empty packet corresponding to the contact node that has been adapted is transmitted on the communication path, and the corresponding path defense measure is called based on the data packet address, so as to restore the contact nodes at both ends of the current contact line to normal communication;

[0071] The above operation is repeated to manage all contact nodes on each contact path, until all contact nodes on all contact paths are restored to normal communication, so as to ensure that each data comparison group can be safely transmitted.

[0072] It should be noted that each contact path is used for corresponding transmission of one data comparison group, and the contact nodes on the contact path are communication blockages, which affect the normal progress of data transmission. By setting a contact line between the contact nodes on different contact paths, and setting a notification point at one end of the contact line and a detection point at the other end, a data empty packet identical in data structure to the data comparison group to be transmitted is generated. The detection point acquires abnormal events of the data empty packet in the transmission process, and adapts the path defense measures to solve the abnormal events. On the one hand, a data empty packet identical in data structure to the data comparison group is generated, and the abnormal events and the corresponding path defense measures are obtained through the data empty packet. When the data comparison group that really needs to be transmitted on the contact path is transmitted, the path defense measures are used to protect the current path data, which protects the original data comparison group from data damage, and also accurately adapts the method measures for solving the communication blockage, effectively ensuring data security. On the other hand, based on the data packet address corresponding to the data empty packet, the corresponding communication blockage problems of different contact nodes with the same data structure and the same type of abnormal events encountered in the communication process are solved, avoiding the low efficiency problem caused by repeated analysis of the same problem.

[0073] It should be further explained that in the specific implementation process, the AI computing resource corresponding to each data amount is allocated to all abnormal sub-data clusters, and the process of grouping all abnormal sub-data clusters based on the abnormal type to obtain a plurality of abnormal signal sets includes:

[0074] Deploy a cloud server to real-time count the amount of computing resources available for scheduling in the cloud server;

[0075] Transfer all abnormal sub-data clusters to the cloud server to obtain the data amount corresponding to each abnormal sub-data cluster, allocate the corresponding AI computing resource based on the data amount of each abnormal sub-data cluster in the cloud server, and count the remaining computing resource amount in the cloud server after deducting the allocated AI computing resource;

[0076] Set a critical threshold value corresponding to the normal operation of the cloud server, and set a shutdown threshold value corresponding to the collapse of the cloud server, wherein the critical threshold value represents a buffer value when the cloud server is normal, and when the critical threshold value is lower, a warning is issued. The shutdown threshold value is a critical value corresponding to the normal operation of the cloud server that cannot be maintained, and when the remaining computing resources are between the critical threshold value and the shutdown threshold value, the cloud server can still maintain normal operation for a certain period of time. At this time, the computing resources corresponding to the cloud server need to be urgently increased;

[0077] When the amount of remaining computing resources is greater than the critical threshold value, no operation is performed;

[0078] When the remaining computing resource amount is less than or equal to the critical overhead threshold, edge resource release is performed, and the corresponding computing resource in the cloud server is increased through the edge resource release until the remaining computing resource is greater than the critical overhead threshold, and the increase is stopped;

[0079] The plurality of abnormal sub-data clusters are processed by a plurality of edge computing devices, each edge computing device obtains an abnormal type of the processed abnormal sub-data cluster, all edge computing devices in the same abnormal type are aggregated into an edge device set, and all abnormal sub-data clusters corresponding to the edge device set are grouped into an abnormal signal set.

[0080] The process of the edge resource release includes: setting a plurality of processing nodes on each edge computing device, each processing node obtains an AI computing resource for maintaining a subsequent edge computing related resource amount of itself, and intercepts a part of data in the abnormal sub-data cluster for processing, after processing the intercepted part of data, the AI computing resource obtained by the corresponding processing node is released, and the released AI computing resource is returned to the cloud server through the edge computing device.

[0081] It should be further explained that, in the specific implementation process, the process of selecting one set item in each abnormal signal set as an analysis target and obtaining corresponding signal features and signal abnormal details includes:

[0082] A plurality of abnormal sub-data clusters in the abnormal signal set are taken as set items of the corresponding set, any one set item is selected as an analysis target, data processing is performed on the analysis target, signal features corresponding to the analysis target are obtained, data analysis is performed based on the signal features and pre-set signal feature parameters, and signal abnormal details corresponding to each signal feature are obtained;

[0083] It should be noted that in the EMC test, an abnormal signal set is composed of a plurality of abnormal points gathered in the frequency domain, time domain or spatial domain, and these gathered abnormal points constitute an abnormal sub-data cluster. For example, in the radiation emission test corresponding to the EMC test, the fundamental wave with a clock frequency of 1 GHz is taken as the emission center source, and the third (3 GHz), fifth (5 GHz) and seventh (7 GHz) harmonics respectively form corresponding over-standard frequency bands, which constitute three abnormal sub-data clusters corresponding to different over-standard frequency bands. The over-standard frequency band is different from the standard frequency band, and the corresponding data is also abnormal. For each abnormal sub-data cluster, the corresponding data segment is extracted from the corresponding complete EMC test data, the data is sliced using EMC test software (such as MATLAB), a data set including only the current analysis target is created, and interference from other frequency band signals to the abnormal sub-data cluster is avoided. The data set is pre-processed, specifically including subtracting the environmental background noise from the signal amplitude of the data segment corresponding to the data set to obtain the corresponding net disturbance intensity. When there is data jitter in the data segment and the jitter amplitude exceeds the preset amplitude standard, a filter is used to smooth the fitting curve corresponding to the data segment, highlight the signal change trend corresponding to the data segment, and keep the peak value information of the data segment from being lost. The purpose is to improve the signal-to-noise ratio and ensure that the subsequent analysis is of the real abnormal signal generated by the chip inductor itself.

[0084] It should be noted that after the data preprocessing of the abnormal sub-data cluster is completed, the extraction of the signal characteristics corresponding to the abnormal sub-data cluster is performed, and based on the obtained signal characteristics and signal characteristic parameters, the subsequent data analysis of the corresponding abnormal sub-data cluster is completed. The signal characteristics include frequency domain characteristics and time domain characteristics, the point with the most concentrated energy of the abnormal sub-data cluster in the frequency spectrum diagram is determined, that is, the frequency corresponding to the highest peak in the frequency spectrum diagram, the amplitude value at the center frequency is taken as the peak amplitude, when the center frequency is a wide frequency cluster, the maximum amplitude value in the wide frequency cluster is taken as the peak amplitude, whether there are regular sidebands on both sides of the peak value is observed, for example, a pair of sidebands appear on both sides with equal intervals (±100 kHz) with the center frequency as the symmetry axis, which indicates that the signal is modulated by periodic noise, and the modulation characteristics of the corresponding signal are obtained. When there are multiple abnormal sub-data clusters, whether the center frequencies of the abnormal sub-data clusters have an integer multiple relationship (such as Fc, 2Fc, 3Fc…) is checked, and the harmonic relationship of the corresponding signal is obtained. The center frequency, the peak amplitude, the modulation characteristics, and the harmonic relationship are collectively used as the frequency domain characteristics. In addition, for each abnormal sub-data cluster, the Fourier transform of the time domain waveform of the signal represented by the abnormal sub-data cluster is performed, the fundamental frequency appearing in the frequency spectrum is obtained as the pulse repetition frequency, the time required for the pulse to rise from 10% amplitude to 90% amplitude on the time domain waveform is measured as the rise time, and the time required for the pulse to fall from 90% amplitude to 10% amplitude is measured as the fall time, and the duration of the pulse at 50% amplitude is measured as the corresponding pulse width. The pulse repetition frequency, the rise time, the fall time, and the pulse width are all time domain characteristics.

[0085] The extracted signal characteristics are physically associated with the hardware and working state of the device in the EMC test scene, and the abstract signal characteristics are converted into signal abnormal details with actual engineering significance. The signal abnormal details are used to represent the interference of the EMC test on the chip inductance, or the interference of the chip inductance itself on the hardware and working state of the device in the test scene. For example:

[0086] Example one: large current injection test causes the power inductance of the chip inductance to appear saturation (EMC test causes interference to the chip inductance), and the signal characteristics obtained from the test data are as follows:

[0087] Center frequency: 120 MHz;

[0088] Peak amplitude: 15 dB higher than the normal state;

[0089] Modulation characteristics: none;

[0090] Harmonic relationship: 2Fc;

[0091] Based on the above signal characteristics, the corresponding signal abnormality details are determined as: the radio frequency immunity test (such as large current injection) in the EMC test or the noise of other power supply networks, which affects the switching power supply circuit (including chip inductance) for powering the USB controller through conduction or induction;

[0092] Example two: the chip inductance itself becomes an electromagnetic disturbance source due to its physical characteristics (such as resonance, saturation) or faults (such as damage, deterioration), which destroys the normal work of other parts in the system. For example, the self-resonance of the decoupling inductor causes clock signal jitter. The signal characteristics obtained from the test data are as follows:

[0093] Center frequency: 800 MHz (also including harmonics of 1.6 GHz and 2.4 GHz);

[0094] Peak amplitude: severely exceeds the standard at 800 MHz;

[0095] Modulation characteristics: in the frequency spectrum, it shows a phase noise base lift and discrete sideband centered on the clock fundamental frequency and its harmonics;

[0096] Based on the above signal characteristics, the corresponding signal abnormality details are determined as: the high impedance and antenna effect of the decoupling inductor at its self-resonance frequency (800 MHz) cause the decoupling inductor to become an interference source, which causes interference to the hardware and working state of the device in the test scene.

[0097] It should be further explained that, in the specific implementation process, the signal characteristics of other collection items are compared with the analysis target, and based on the comparison result, the process of selecting to perform data multiplexing or data analysis includes:

[0098] When the signal characteristics comparison between a collection item in the abnormal signal set and the selected analysis target is completely the same, the signal abnormality details of the corresponding analysis target are directly called as the final signal abnormality details of the current collection item;

[0099] When the signal characteristics comparison is not completely the same, the signal abnormality details of the corresponding analysis target on the same signal frequency band are data multiplexed, and the signal abnormality details of different signal frequency bands are analyzed based on their respective partial signal characteristics and the set signal characteristic parameters. The signal abnormality details of the obtained part are obtained, and the signal abnormality details obtained by data multiplexing are summarized as the final signal abnormality details of the current collection item;

[0100] When the signal feature comparison is completely different, the current signal feature corresponding set item is selected as a new analysis target, and the signal abnormality details after the new analysis target completes data analysis are obtained. The signal features of the new analysis target and the past analysis target are merged, and the signal abnormality details corresponding to the signal features are constructed as a data reuse set. The data reuse set is used for subsequent data analysis and data reuse of other set items.

[0101] The subsequent other set items and the data reuse set are compared with each other in terms of signal features, and the signal abnormality details corresponding to the signal features that are completely the same between each of the different analysis targets in the data reuse set and the current set item are screened out. The signal abnormality details of the current set item are reused, and the signal features and signal feature parameters are compared and matched based on the signal features that are not matched with the current set item. Additional signal abnormality details are obtained. The signal abnormality details obtained through screening and the additional signal abnormality details obtained through matching are summarized and taken as the final signal abnormality details.

[0102] By comparing the signal features of all set items in each abnormal signal set, the signal abnormality details corresponding to each set item in the abnormal signal set are obtained. The signal abnormality details are used to represent specific abnormal information content in the chip inductance EMC test process, that is, the abnormal influence of the chip inductance itself on the signal and the abnormal influence of other devices in the test scene on the signal at the chip inductance.

[0103] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. An AI-driven method for analyzing abnormal signals in chip inductor EMC testing, characterized in that, Includes the following steps: Step S1: Perform EMC testing on the chip inductor, compare and analyze the EMC test data with the preset standard EMC test data, and then filter out all abnormal sub-data clusters in the test data; Step S2: Allocate corresponding AI computing resources to all abnormal sub-data clusters based on their respective data volume, and group all abnormal sub-data clusters based on abnormality type to obtain several abnormal signal sets; Step S3: Select one set item from each abnormal signal set as the analysis target, obtain the corresponding signal characteristics and signal abnormality details, compare the signal characteristics of other set items with the analysis target, and select to perform data reuse or data analysis based on the comparison results.

2. The AI-driven chip inductor EMC test abnormal signal analysis method according to claim 1, characterized in that, The process of performing EMC testing on chip inductors, comparing and analyzing the EMC test data with preset standard EMC test data, and then filtering out all abnormal sub-data clusters in the test data includes: Based on the preset test procedures, EMC tests are performed on the chip inductor to obtain the corresponding test data. The test data is then processed into several stages of test data to obtain standard EMC test data, which is then processed into several stages of standard test data. The test data and standard test data of several phases are aligned based on timestamps to construct several data comparison groups. A communication path is established for each data comparison group. The communication path includes several communication nodes, and a communication line is associated with each communication node. The data comparison group is uploaded to the virtual machine via the communication path for comparison and analysis to determine whether there are any anomalies in the data comparison group. If so, the stage test data in the corresponding data comparison group is marked as an abnormal sub-data cluster. The comparison and analysis of several data comparison groups is repeated to filter all abnormal sub-data clusters in the test data. If not, no operation is performed.

3. The AI-driven chip inductor EMC test abnormal signal analysis method according to claim 2, characterized in that, The communication path, communication node, and communication line include: The starting point of the communication path is used to store data comparison groups, and the ending point is used to connect virtual machines. Path security management is performed on the data comparison groups transmitted on each communication path. The contact nodes are set at the corresponding communication congestion locations for each contact path; The communication line is used to enable data communication between communication nodes on different communication paths. At both ends of the communication line, there are detection points and notification points. The detection points are set at the end of the communication line, and the notification points are set at the beginning of the communication line. The detection point and the notification point each communicate with the contact node.

4. The AI-driven chip inductor EMC test abnormal signal analysis method according to claim 3, characterized in that, The path security management process includes: When the data comparison group is transmitted to the communication node on the communication path, the communication node generates a data empty packet with the same data structure as the data comparison group and hands the data empty packet over to the notification point on the communication line. The notification point obtains the data packet address of the data empty packet and sends it to the probe point, which then investigates abnormal events on the current communication path. Based on the type of abnormal event, the corresponding path defense measures are adapted, and the abnormal event and path defense measures are returned to the notification point along with the data packet address. When transmitting the data comparison group under the current communication path, the abnormal event on the communication path is resolved by executing the path defense measures, and the corresponding communication nodes at both ends of the communication line are restored to normal communication. The communication nodes on different communication paths generate their own empty data packets and obtain the abnormal events of their respective empty data packets on the communication path. If the empty data packets of the corresponding communication node and the abnormal events of the corresponding communication path are exactly the same as those of the communication nodes that have been adapted, then a communication path is constructed by traversing through the notification points of the current communication node to the corresponding notification points of the completely identical communication nodes. Transmit the data packet address of the data empty packet corresponding to the adapted contact node on the communication path, invoke the corresponding path defense measures based on the data packet address, restore the contact nodes at both ends of the current contact path to normal communication, repeat the management of all contact nodes on each contact path until all contact nodes are restored to normal communication.

5. The AI-driven chip inductor EMC test abnormal signal analysis method according to claim 4, characterized in that, The process of allocating corresponding AI computing resources to all anomalous sub-data clusters based on their respective data volume, and grouping all anomalous sub-data clusters based on anomalous type to obtain several sets of anomalous signals includes: Deploy cloud servers and calculate the amount of computing resources available for scheduling on the cloud servers; All abnormal sub-data clusters are transmitted to the cloud server to obtain the data volume of each abnormal sub-data cluster. Based on the data volume of each abnormal sub-data cluster, corresponding AI computing power resources are allocated, and the remaining computing power resources after deducting the allocated AI computing power resources are calculated in the cloud server. Set the critical overhead threshold for maintaining normal operation of the cloud server; When the remaining computing power resources exceed the critical overhead threshold, no operation is performed. When the remaining computing power resources are less than or equal to the critical overhead threshold, edge resource release is performed to increase the computing power resources in the cloud server until the remaining computing power resources are greater than the critical overhead threshold and then the increase stops. Several edge computing devices process several abnormal sub-data clusters, all edge computing devices of the same abnormal type are aggregated into an edge device set, and all abnormal sub-data clusters corresponding to an edge device set are grouped into an abnormal signal set.

6. The AI-driven chip inductor EMC test abnormal signal analysis method according to claim 5, characterized in that, The process of selecting one set item from each abnormal signal set as the analysis target and obtaining the corresponding signal characteristics and signal anomaly details includes: Several abnormal sub-data clusters in the abnormal signal set are taken as set items of the corresponding set. Any set item is selected as the analysis target for data processing to obtain the corresponding signal features. Based on the signal features and the pre-set signal feature parameters, data analysis is performed to obtain the signal abnormality details of each signal feature.

7. The AI-driven chip inductor EMC test abnormal signal analysis method according to claim 6, characterized in that, The process of comparing the signal characteristics of other set items with the analysis target, and selecting whether to perform data reuse or data analysis based on the comparison results, includes: When a set item in the abnormal signal set is exactly the same as the signal feature of the selected analysis target, the signal abnormality details of the corresponding analysis target are directly called. When the signal features are not completely identical, the signal anomaly details of the corresponding analysis targets in the same signal frequency band are reused, and data analysis is performed on different signal frequency bands based on their respective partial signal features and set signal feature parameters to obtain the signal anomaly details of the corresponding signal frequency band. When the signal features are completely different, the set item corresponding to the current signal feature is selected as the new analysis target, and the signal anomaly details after the new analysis target completes the data analysis are obtained. The signal features and signal anomaly details corresponding to the signal features of the new analysis target and the previous analysis targets are merged to construct a data reuse set. The data reuse set is used for subsequent data analysis and data reuse corresponding to other set items.

8. The AI-driven chip inductor EMC test abnormal signal analysis method according to claim 7, characterized in that, The process of using a data reuse set for subsequent data analysis and data reuse of other set items includes: The signal characteristics of subsequent other set items are compared with those of the data reuse set. The signal anomaly details corresponding to different analysis targets in the data reuse set that have exactly the same signal characteristics as the current set item are selected and the signal anomaly details are reused for the current set item. For signals that are not matched in the current set item, the signal matching is performed based on the signal characteristics and signal characteristic parameters to obtain additional signal anomaly details. The selected signal anomaly details and the additional signal anomaly details obtained by matching are summarized as the final signal anomaly details.

Images