Multi-graphics card terminal circuit anomaly detection method and system

By capturing current and voltage signals in the multi-graphics card terminal circuit for Fourier transform analysis, identifying entropy value changes, predicting faults and responding in real time, the problem of reaction lag in the existing technology is solved, and efficient fault detection and prevention is achieved. Reduces system downtime risks and maintenance costs.

CN118914808APending Publication Date: 2024-11-08BEIJING BIGKAT TECH CO LTD
View PDF 0 Cites 4 Cited by

Patent Information

Application Number
CN202410939598.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-15
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The existing fault detection technology does not respond in a timely manner in a high-demand operating environment, and cannot make full use of dynamic data analysis to predict failures, resulting in increased system downtime and maintenance costs, especially in multi-graphics systems, where there is a risk of response lag and equipment damage.

Method used

By capturing the current and voltage signals of the multi-graphics card terminal circuit, Fourier transform analyzes the frequency domain characteristics, extracts entropy information, compares the standard deviation between real-time and historical data, identifys abnormalities and predicts potential faults, adjusts operating parameters and activates faults in real time, and adjusts operating parameters and activates faults in real time. Isolate programs, dynamic monitoring and repair of failures.

Benefits of technology

It significantly improves the speed and accuracy of fault detection, can promptly identify small changes, prevent serious failures, respond quickly and optimize graphics card performance, extend the service life of the equipment, and reduce maintenance costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118914808A_ABST
    Figure CN118914808A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of fault detection, in particular to a multi-display-card terminal circuit anomaly detection method and system, and the method comprises the following steps: capturing current and voltage signals of a multi-display-card terminal circuit through a circuit interface, synchronizing the signals to a data processing center for initial waveform integration, and generating a circuit signal data set; according to the method, the entropy information of the graphics card is extracted by analyzing the frequency domain characteristics through Fourier transform, so that the fault detection speed and accuracy are remarkably improved, the fault detection does not only depend on static or delayed data analysis any more, potential faults can be dynamically analyzed and predicted from continuously updated data, and the fault detection accuracy is improved. By comparing the entropy standard deviation of real-time data and historical data, tiny changes and trends of a circuit can be recognized in time, serious faults are prevented in advance, operation parameters of a display card are adjusted through continuous feedback, the overall performance is optimized, the service life of equipment is prolonged, and the maintenance cost is reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of fault detection, and in particular to a method and system for detecting abnormality of a multi-graphics card terminal circuit. Background Art

[0002] The field of fault detection technology focuses on identifying and preventing potential faults in systems or equipment, including signal processing, data analysis, pattern recognition, and machine learning techniques to extract abnormal indicators from equipment operation data. Fault detection technology is particularly important in electronic circuits, especially in complex systems such as servers or data centers. It can monitor the performance of circuits in real time, detect problems such as short circuits, overheating, or functional failures in a timely manner, and activate early warning systems or automatic repair programs to prevent equipment damage and system downtime. Fault detection not only improves the reliability and safety of equipment, but also helps reduce maintenance costs and extend the service life of equipment.

[0003] Among them, the multi-graphics card terminal circuit anomaly detection method involves monitoring and analyzing circuit behavior in a computing system using multiple graphics cards to identify potential failures or performance degradation. The main purpose is to ensure the efficient operation of the graphics processing unit (GPU), especially in high-performance computing or graphics-intensive applications (such as video games, 3D simulations, video rendering), where the stability of the graphics card is critical to the overall performance. By real-time detection and analysis of graphics card parameters such as voltage, current, and temperature, abnormal situations can be discovered and handled in a timely manner to avoid system overload or damage, ensure stable operation of the system, and optimize performance output.

[0004] Existing fault detection technologies mainly focus on post-analysis and fault response, and only take measures after the fault occurs. In a high-demand operating environment, the response is not timely enough, causing irreversible equipment damage or data loss. For example, in the detection of electronic circuit faults in servers or data centers, traditional technologies rely on regular maintenance inspections or error log analysis, which cannot capture the early stages of fault development in real time, resulting in delayed response to faults, increasing the risk of system downtime and maintenance costs. Existing technologies fail to fully utilize dynamic data analysis to predict faults and lack effective prediction tools to process and analyze large amounts of real-time data, which is particularly critical in complex devices such as multi-graphics card systems, because the stability of the system is directly related to maximizing overall performance and minimizing system failures. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a method and system for detecting abnormality of circuits of a multi-graphics card terminal.

[0006] In order to achieve the above object, the present invention adopts the following technical solution, a method for detecting abnormality of a multi-graphics card terminal circuit, comprising the following steps:

[0007] S1: Capture the current and voltage signals of the multi-graphics terminal circuit through the circuit interface, synchronize the signals to the data processing center for initial waveform integration, and generate a circuit signal data set;

[0008] S2: Performing Fourier transform on the circuit signal data set, analyzing the frequency domain characteristics, extracting entropy information of the graphics card terminal circuit, calculating the entropy variation of each graphics card, and obtaining entropy analysis results;

[0009] S3: According to the entropy value analysis result, compare the entropy value standard deviation recorded in the real-time data and the historical data, identify the change and deviation of the entropy value, detect the abnormality of the multi-graphics terminal circuit, and generate a trend comparison analysis result;

[0010] S4: using the data in the trend comparison and analysis results, identifying and detecting abnormalities in the circuits of the multi-graphics card terminal, classifying potential failure modes of the circuits, and predicting failures in a future time period to generate failure prediction results;

[0011] S5: applying the fault prediction result to the fault response mechanism of the multi-graphics card terminal circuit, adjusting the graphics card operation parameters and activating the fault isolation program, and obtaining a real-time feedback signal by updating the multi-graphics card terminal circuit state in real time;

[0012] S6: Implement a fault response strategy based on the real-time feedback signal, start a maintenance protocol, isolate and repair the detected fault, evaluate the effect of the fault repair, and obtain a fault handling solution.

[0013] As a further solution of the present invention, the circuit signal data set includes current data, voltage data and timestamp, the entropy analysis results include the entropy value of the graphics card, the entropy variation and the statistical distribution of the entropy value, the trend comparison analysis results include the entropy change trend, the entropy standard deviation and the entropy deviation index, the fault prediction results include the predicted fault type, the predicted fault time and the fault probability score, the real-time feedback signal includes the real-time status of the graphics card, the operating parameters and the fault isolation signal, and the fault handling plan includes fault isolation measures, maintenance protocols and fault repair effect evaluation.

[0014] As a further solution of the present invention, the current and voltage signals of the multi-graphics terminal circuit are captured through the circuit interface, and the synchronization signals are sent to the data processing center for initialization waveform integration. The steps of generating the circuit signal data set are specifically as follows:

[0015] S101: collecting current and voltage signals of the multi-graphics card terminal circuit through the circuit interface, adjusting the sampling frequency to match the circuit characteristics, and removing noise to generate an original data set;

[0016] S102: Based on the original data set, the data is transmitted to the data processing center in real time, the consistency of the data transmission is checked through data encryption, the transmission rate and resource consumption are adjusted, and the synchronized current and voltage data are obtained;

[0017] S103: Utilizing the synchronized current and voltage data, processing and integrating the current and voltage data, adjusting the window size and step size to optimize data processing efficiency and avoid information loss, identifying key waveform features, and acquiring a circuit signal data set.

[0018] As a further solution of the present invention, the circuit signal data set is subjected to Fourier transform, the frequency domain characteristics are analyzed, the entropy value information of the graphics card terminal circuit is extracted, the entropy variation of each graphics card is calculated, and the steps of obtaining the entropy value analysis result are specifically as follows:

[0019] S201: According to the circuit signal data set, using a fast Fourier transform algorithm, converting the time domain signal into a frequency domain signal, identifying and recording key frequency components, and generating frequency domain characteristic data;

[0020] S202: Analyzing the distribution density of the differentiated frequency components based on the frequency domain characteristic data, adjusting the interval and step size of entropy value calculation, extracting the entropy value of each graphics card terminal circuit according to the distribution uniformity, and obtaining graphics card entropy value information;

[0021] S203: Using the graphics card entropy information, calculate the variation between the graphics card entropy values, compare the entropy value differences between differentiated graphics cards, and quantify the entropy value variation to obtain an entropy value analysis result.

[0022] As a further solution of the present invention, the formula of the fast Fourier transform algorithm is as follows:

[0023]

[0024] Among them, X k is the kth frequency domain sample, x n is the nth time domain sample, N is the total number of samples, k is the index of the frequency domain sample, w is the window function, α is the modulation coefficient, is the cosine modulation part in the frequency domain.

[0025] As a further solution of the present invention, according to the entropy analysis results, the steps of comparing the entropy standard deviation recorded in the real-time data and the historical data, identifying the change and deviation of the entropy, detecting the abnormality of the multi-graphics terminal circuit, and generating the trend comparison analysis results are specifically as follows:

[0026] S301: Based on the entropy analysis result, compare the entropy standard deviation in the real-time data of the multi-graphics card terminal circuit with the entropy standard deviation of the collected historical data, check the consistency of the time range and the statistical method by setting comparison parameters, calculate the difference between the real-time and historical entropy values, and obtain entropy change data;

[0027] S302: Analyze the change trend and deviation of the entropy value according to the entropy value change data, analyze the abnormal points of the multi-graphics card terminal circuit according to the target threshold, mark the data points that exceed the normal range, and obtain the deviation identification result;

[0028] S303: using the deviation identification result, summarizing the abnormal points identified in the multi-graphics card terminal circuit, and performing comparative analysis with the historical trend, and obtaining the trend comparative analysis result by recording the deviation between the abnormal points and the normal operating state.

[0029] As a further solution of the present invention, the data in the trend comparison and analysis results are used to identify and detect abnormalities in the circuits of multi-graphics card terminals, classify potential circuit failure modes, and predict failures in future time periods. The steps of generating failure prediction results are specifically as follows:

[0030] S401: Based on the trend comparison and analysis results, identify and record the abnormal fluctuations that deviate, match the abnormal fluctuations with the characteristics of the known fault mode, mark the abnormal data points that match the fault mode, and obtain the fault mode recognition result;

[0031] S402: According to the fault mode recognition result, the fault modes identified in the multi-graphics card terminal circuit are classified and sorted, the urgency of the fault mode is evaluated according to the frequency and impact of the abnormality, and risk ranking and priority setting are performed to generate fault classification and evaluation results;

[0032] S403: Using the fault classification and evaluation results, predict the probability and expected time of potential faults within the target time period, identify the risks and risk levels of potential faults, set warning levels and response strategies, and obtain fault prediction results.

[0033] As a further solution of the present invention, the steps of applying the fault prediction result to the fault response mechanism of the multi-graphics card terminal circuit, adjusting the graphics card operating parameters and activating the fault isolation program, and obtaining the real-time feedback signal by updating the multi-graphics card terminal circuit state in real time are specifically as follows:

[0034] S501: Based on the fault prediction result, the fault response parameter is input into the graphics card terminal circuit, the operation parameter of the graphics card is adjusted to avoid the influence of the potential fault, and a preset fault isolation program is started to obtain the adjusted operation parameter;

[0035] S502: using the adjusted operating parameters, cyclically tracking the operating status of the graphics card, extracting key data in the operation of the graphics card, and comparing the early warning indicators to evaluate the adjustment effect of the operating parameters, and generating status monitoring data;

[0036] S503: According to the status monitoring data, update the status display of the multi-graphics terminal circuit, identify and record the real-time operation status and new abnormal alarm of the multi-graphics terminal circuit, and obtain real-time feedback signals.

[0037] As a further solution of the present invention, according to the real-time feedback signal, a fault response strategy is implemented, a maintenance protocol is started, the detected fault is isolated and repaired, and the effect of the fault repair is evaluated. The steps of obtaining a fault handling solution are specifically as follows:

[0038] S601: Based on the real-time feedback signal, a fault response strategy is designed, a pre-designed automatic fault isolation is started using a maintenance protocol, and an emergency notification is sent to perform manual intervention of circuit abnormalities, isolate key components and limit functions, and obtain a fault isolation result;

[0039] S602: performing fault diagnosis of the multi-graphics card terminal circuit based on the fault isolation result, analyzing the fault source, repairing the graphics card and related circuits using a repair tool and replacement parts, and generating a fault repair record;

[0040] S603: Using the fault repair record, evaluate the performance and stability of the circuit after repair, re-run the diagnostic tool to verify the repair effect of the fault, check the normal operation of the circuit parameters and functions, and obtain a fault handling solution.

[0041] A multi-graphics card terminal circuit anomaly detection system, the multi-graphics card terminal circuit anomaly detection system is used to execute the multi-graphics card terminal circuit anomaly detection method, the system comprises:

[0042] The signal capture module is based on the multi-graphics card terminal circuit interface, collects current and voltage data to match different types of graphics cards, identifies the current and voltage signals when the graphics cards are running, screens the signal quality, and generates original circuit signal data;

[0043] The data processing module transmits the synchronization signal to the data processing center according to the original circuit signal data, integrates the waveform data, and sets integration parameters to obtain the integrated waveform data;

[0044] The entropy value calculation module uses the integrated waveform data to convert the waveform data into the frequency domain, analyzes the frequency domain characteristics, calculates the entropy value and variation of the graphics card terminal circuit, and generates an entropy value analysis result;

[0045] The fault analysis module uses the entropy analysis result to compare the entropy standard deviation of real-time data and historical data, identify entropy changes and deviations, and predict and classify fault modes to obtain fault prediction results;

[0046] Based on the fault prediction result, the fault response module adjusts the graphics card operating parameters, activates the fault isolation and maintenance protocol, implements the fault handling strategy, evaluates and updates the operating status of the graphics card terminal circuit, and obtains the fault handling solution.

[0047] Compared with the prior art, the advantages and positive effects of the present invention are:

[0048] In the present invention, by capturing the current and voltage signals of the multi-graphics card terminal circuit of the circuit interface in real time and performing synchronous processing, and using Fourier transform to analyze the frequency domain characteristics to extract the entropy value information of the graphics card, the speed and accuracy of fault detection are significantly improved, so that fault detection no longer relies solely on static or delayed data analysis, and can dynamically analyze and predict potential faults from continuously updated data, which is crucial for maintaining the stable operation of graphics cards for high-performance computing and graphics-intensive applications. By comparing the entropy value standard deviation of real-time data and historical data, it is possible to timely identify minor changes and trends in the circuit and prevent the occurrence of serious faults in advance. By applying the fault prediction results to the fault response mechanism, it is not only possible to react quickly to isolate and repair the problem, but also to adjust the graphics card operating parameters through continuous feedback, optimize the overall performance, extend the service life of the equipment and reduce maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic diagram of the workflow of the present invention;

[0050] Figure 2 This is a detailed flow chart of S1 of the present invention;

[0051] Figure 3 This is a detailed flow chart of S2 of the present invention;

[0052] Figure 4 This is a detailed flow chart of S3 of the present invention;

[0053] Figure 5 This is a detailed flow chart of S4 of the present invention;

[0054] Figure 6 This is a detailed flow chart of S5 of the present invention;

[0055] Figure 7 This is a detailed flow chart of S6 of the present invention;

[0056] Figure 8 It is a system flow chart of the present invention. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with 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 intended to limit the present invention.

[0058] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0059] Embodiment 1

[0060] See also Figure 1 The present invention provides a technical solution, a method for detecting abnormal circuits of a multi-graphics card terminal, comprising the following steps:

[0061] S1: Capture the current and voltage signals of the multi-graphics terminal circuit through the circuit interface, synchronize the signals to the data processing center for initial waveform integration, check the consistency of data transmission through data encryption, and generate a circuit signal data set;

[0062] S2: Perform Fourier transform on the circuit signal data set, analyze the frequency domain characteristics, extract the entropy information of the graphics card terminal circuit, calculate the entropy variation of each graphics card, adjust the interval and step size of the entropy calculation, and obtain the entropy analysis result;

[0063] S3: Based on the entropy analysis results, compare the entropy standard deviation recorded in the real-time data and historical data, identify the changes and deviations in the entropy, detect the abnormalities of the multi-graphics terminal circuit, mark the data points that are beyond the normal range, and generate trend comparison analysis results;

[0064] S4: using the data in the trend comparison analysis results, identifying and detecting the abnormality of the multi-graphics terminal circuit, classifying the potential failure modes of the circuit, and predicting the failure in the future time period to generate the failure prediction results;

[0065] S5: applying the fault prediction result to the fault response mechanism of the multi-graphics card terminal circuit, adjusting the graphics card operation parameters and activating the fault isolation program, and obtaining a real-time feedback signal by updating the multi-graphics card terminal circuit state in real time;

[0066] S6: Implement fault response strategies based on real-time feedback signals, initiate maintenance protocols, isolate and repair detected faults, evaluate the effectiveness of fault repairs, check circuit parameters and normal operation of functions, and obtain fault handling solutions.

[0067] The circuit signal data set includes current data, voltage data and timestamp. The entropy analysis results include the entropy value of the graphics card, entropy variation and statistical distribution of the entropy value. The trend comparison analysis results include the entropy change trend, entropy standard deviation and entropy deviation index. The fault prediction results include the predicted fault type, predicted fault time and fault probability score. The real-time feedback signal includes the real-time status of the graphics card, operating parameters and fault isolation signal. The fault handling plan includes fault isolation measures, maintenance protocol and fault repair effect evaluation.

[0068] See also Figure 2 , the current and voltage signals of the multi-graphics terminal circuit are captured through the circuit interface, and the synchronization signals are sent to the data processing center for initial waveform integration. The specific steps for generating the circuit signal data set are as follows:

[0069] S101: Collect the current and voltage signals of the multi-graphics card terminal circuit through the circuit interface, adjust the sampling frequency to match the circuit characteristics, and remove the noise to generate the original data set. The execution process is as follows;

[0070] The S101 sub-step collects the current and voltage signals of the multi-graphics terminal circuit through the circuit interface. The interface needs to be configured to match the characteristics of the circuit, such as voltage level and current capacity. The sampling frequency is set according to the response speed of the circuit and the signal change rate to ensure that the time resolution of the data is high enough to capture key signal fluctuations. Digital filtering technology is used to remove noise generated by electromagnetic interference or the device itself to ensure that the collected data signal is pure and generate the original data set. The formula used is:

[0071]

[0072] Among them, P data Represents the power signal strength of the processed data set, I i and V i are the current and voltage values ​​of the i-th sampling respectively, and n is the total number of sampling points.

[0073] S102: Based on the original data set, the data is transmitted to the data processing center in real time, the consistency of data transmission is checked through data encryption, the transmission rate and resource consumption are adjusted, and the execution process of obtaining the synchronized current and voltage data is as follows;

[0074] Substep S102 transmits the data to the data processing center in real time based on the original data set. To ensure the security and consistency of the data during transmission, data encryption technology is introduced, such as using AES or RSA algorithm to encrypt the transmitted data to protect the data from being illegally intercepted or tampered with. Monitor the rate and resource consumption during the transmission process, dynamically adjust the transmission rate according to the network status, optimize the use of network resources, ensure the efficiency and economy of data transmission, and obtain the synchronized current and voltage data. The formula used is:

[0075]

[0076] Among them, E sync represents the data synchronization efficiency, D trans is the total amount of data transmitted, T trans is the time required for data transmission, and α is the efficiency adjustment coefficient.

[0077] S103: Using the synchronized current and voltage data, process and integrate the current and voltage data, adjust the window size and step size to optimize data processing efficiency and avoid information loss, and identify key waveform features. The execution process of obtaining the circuit signal data set is as follows;

[0078] Substep S103 uses the synchronized current and voltage data for further processing and integration. By setting the appropriate window size and step size, the efficiency and accuracy of data processing can be optimized to avoid information loss during the analysis process. The choice of window size depends on the frequency characteristics and waveform complexity of the signal, and the step size is adjusted according to the data processing capability and real-time requirements. The signal processing algorithm is used to identify the key waveform features in the current and voltage data, such as peaks, troughs, and periodic fluctuations, and obtain the circuit signal data set. The formula used is:

[0079]

[0080] Among them, F feature represents the feature recognition score, x j is the jth data point in the window, μ x is the average value of the window data, σ x is the standard deviation, β is the normalization coefficient, and m is the total number of data points in the window.

[0081] See also Figure 3 , perform Fourier transform on the circuit signal data set, analyze the frequency domain characteristics, extract the entropy value information of the graphics card terminal circuit, calculate the entropy variation of each graphics card, and obtain the entropy value analysis results in the following steps:

[0082] S201: According to the circuit signal data set, the fast Fourier transform algorithm is used to convert the time domain signal into a frequency domain signal, identify and record the key frequency components, and generate the frequency domain characteristic data. The execution process is as follows;

[0083] Sub-step S201 uses the fast Fourier transform (FFT) algorithm to convert the time domain signal into a frequency domain signal based on the circuit signal data set, and uses the FFT algorithm to process the signal through the collected current and voltage data. The specific operation of FFT includes performing a discrete Fourier transform on the collected time series signal to convert it into a frequency domain signal. Identify the key frequency components in the frequency domain signal, such as the fundamental frequency, harmonics, and noise frequency. The key frequency components are recorded to generate frequency domain characteristic data.

[0084] The formula for the Fast Fourier Transform algorithm is as follows:

[0085]

[0086] Among them, X k is the kth frequency domain sample, x n is the nth time domain sample, N is the total number of samples, k is the index of the frequency domain sample, w is the window function, α is the modulation coefficient, is the cosine modulation part in the frequency domain.

[0087] The execution flow is as follows:

[0088] Apply the window function w n To each time domain sample x n In order to reduce spectrum leakage, the fast Fourier transform algorithm is executed to convert time domain samples into frequency domain samples, and the spectrum resolution is improved by using cosine modulation. The intensity of cosine modulation is adjusted by the α coefficient to make the key frequency components more prominent. The weighted average of each frequency component is calculated to optimize the accuracy of spectrum analysis. The determination of α can be obtained by minimizing the prediction error, including experimenting with a series of preset α values ​​and selecting the α value that minimizes the error.

[0089] S202: Based on the frequency domain characteristic data, the distribution density of the differentiated frequency components is analyzed, the interval and step size of the entropy value calculation are adjusted, and the entropy value of each graphics card terminal circuit is extracted according to the distribution uniformity. The execution process of obtaining the graphics card entropy value information is as follows;

[0090] Sub-step S202 calculates the entropy value based on the frequency domain characteristic data to analyze the frequency component distribution of the graphics card terminal circuit, adjusts the interval and step size of the entropy value calculation, and the parameters affect the accuracy and sensitivity of the entropy value calculation. Entropy value calculation is used to evaluate the uniformity of the distribution of frequency components, that is, the randomness and complexity of frequency domain signals. By performing entropy value analysis on the frequency component distribution of each graphics card terminal circuit, the non-uniformity or abnormality of circuit performance can be identified. The entropy value of each graphics card is calculated and recorded to obtain the graphics card entropy value information, using the formula:

[0091]

[0092] Among them, H represents the entropy value, p k is the proportion of the kth frequency component in the total frequency distribution, and K is the total number of frequency components.

[0093] S203: using the graphics card entropy information, calculating the variation between the graphics card entropy values, comparing the entropy value differences between differentiated graphics cards, and quantifying the entropy value variation, the execution process of obtaining the entropy value analysis result is as follows;

[0094] Sub-step S203 uses the entropy information of the graphics card to calculate the variability of the entropy values ​​between the graphics cards, compare the entropy differences of the terminal circuits of different graphics cards, and identify patterns or anomalies. Through statistical analysis, the variability between entropy values ​​is quantified, which can help identify performance problems or improper configuration of the graphics card. The quantification result of the entropy variability is used to obtain the entropy analysis result, using the formula:

[0095]

[0096] Among them, V represents the variability of entropy value, H j is the entropy value of the jth graphics card, is the average entropy value of all graphics cards, and N is the total number of graphics cards.

[0097] See also Figure 4 ,According to the entropy analysis results, the steps of comparing the entropy standard deviation recorded in the real-time data and the historical data, identifying the changes and deviations of the entropy, detecting the abnormality of the multi-graphics terminal circuit, and generating the trend comparison analysis results are as follows:

[0098] S301: Based on the entropy value analysis result, the entropy value standard deviation in the real-time data of the multi-graphics terminal circuit is compared with the entropy value standard deviation of the collected historical data, and the consistency of the time range and the statistical method is checked by setting comparison parameters, and the difference between the real-time and historical entropy values ​​is calculated to obtain the execution process of the entropy value change data as follows;

[0099] Based on the entropy analysis results, sub-step S301 uses statistical analysis methods to compare the standard deviation of entropy values ​​in real-time data with the standard deviation of entropy values ​​of collected historical data, and sets the parameters of statistical analysis, including time range selection and statistical method determination, to ensure the accuracy and consistency of comparative analysis. By calculating the standard deviation of entropy values ​​of real-time data and historical data, the stability and change trend of graphics card circuit performance can be evaluated. The comparison helps to identify performance degradation or functional abnormalities, providing a basis for further diagnosis, and generating entropy change data using the formula:

[0100]

[0101] Among them, σ is the standard deviation of the weighted sample entropy value, X i is the entropy value of each sample point, μ is the average entropy value of the sample, and w i is the weight of each sample point.

[0102] S302: Analyze the change trend and deviation of the entropy value according to the entropy value change data, analyze the abnormal points of the multi-graphics card terminal circuit according to the target threshold, mark the data points that exceed the normal range, and obtain the deviation identification result;

[0103] Sub-step S302 analyzes the entropy change trend and deviation size based on the entropy change data, and uses statistical methods such as moving average or linear regression to determine the entropy change trend. According to the set target threshold, identify the points in the graphics card terminal circuit where the entropy value deviates abnormally from the normal range. The marked abnormal points are used to indicate potential hardware failures or configuration errors. All identified abnormal points and related data are recorded to obtain the deviation identification results, using the formula:

[0104]

[0105] Among them, V d The variance measure representing the change in entropy value, H real is the actual measured entropy value, H trend is the predicted entropy value and n is the total number of data points.

[0106] S303: using the deviation identification result, summarizing the abnormal points identified in the multi-graphics card terminal circuit, and performing comparative analysis with the historical trend, and obtaining the trend comparative analysis result by recording the deviation between the abnormal points and the normal operation state, the execution process is as follows;

[0107] Sub-step S303 uses the deviation identification results to summarize all abnormal points identified in the multi-graphics terminal circuit and compare and analyze them with historical trends, including evaluating the frequency, severity and time distribution of the abnormal points to determine whether there are any persistent or emerging problem patterns. Comparing the deviations of the abnormal points with the normal operating state helps the circuit maintenance team to develop more effective maintenance strategies and preventive measures to improve the reliability and performance of the graphics card circuit. The trend comparison analysis results are obtained using the formula:

[0108]

[0109] Among them, T c represents the average deviation of trend comparison, H j is the entropy value of each outlier, H norm is the average of the normal entropy values, and m is the total number of abnormal points.

[0110] See also Figure 5 , using the data in the trend comparison analysis results, identify and detect the abnormalities of the multi-graphics terminal circuit, classify the potential failure modes of the circuit, and predict the failures in the future time period. The specific steps for generating the fault prediction results are as follows:

[0111] S401: Based on the trend comparison analysis results, identify and record the abnormal fluctuations that deviate, match the abnormal fluctuations with the characteristics of the known fault mode, mark the abnormal data points that match the fault mode, and obtain the execution process of the fault mode identification result as follows;

[0112] Based on the trend comparison analysis results, sub-step S401 identifies abnormal fluctuations that deviate from the normal operating mode, and identifies data points that significantly deviate from the normal fluctuation range by conducting a detailed inspection of the data points obtained from the trend analysis. Match the abnormal fluctuations with the known fault mode features in the database. Through pattern recognition technology, such as machine learning classification algorithms, the features of the abnormal fluctuations are compared with the features in the fault mode library, and the abnormal data points that match the specific fault mode are marked to obtain the fault mode recognition results.

[0113] S402: According to the fault mode recognition result, the fault modes identified in the multi-graphics terminal circuit are classified and sorted, the urgency of the fault mode is evaluated according to the frequency and impact of the abnormality, and the risk ranking and priority setting are performed. The execution process of generating the fault classification and evaluation results is as follows;

[0114] Sub-step S402 classifies and sorts the fault modes in the multi-graphics terminal circuit according to the fault mode identification results, which involves evaluating the various fault modes identified according to the frequency of occurrence and the degree of impact, so as to determine the urgency of each fault mode, sort the risks and set the priority, and ensure that the most urgent faults are handled first. Through analysis, the fault classification and evaluation results are generated, and the formula used is:

[0115]

[0116] Among them, Z f represents the failure risk score, f j is the occurrence frequency of the jth failure mode, g j is the influence weight of the failure mode, and m is the total number of failure modes.

[0117] S403: Using the fault classification and evaluation results, predict the probability and expected time of potential faults within the target time period, identify the risks and risk levels of potential faults, and set warning levels and response strategies. The execution process of obtaining fault prediction results is as follows;

[0118] Sub-step S403 uses the fault classification and evaluation results to predict the probability and expected time of potential faults within the target time period through statistical analysis and machine learning models, collects historical fault data and current fault modes, and uses regression analysis or time series analysis models to calculate the probability and expected time of each fault mode. Through risk assessment methods, identify the risks and risk levels of each potential fault, including considering the probability of fault occurrence, scope of impact and severity. According to the analysis results, set the warning level and corresponding response strategy, such as adjusting operating parameters or starting fault isolation procedures, and obtain fault prediction results. The formula used is:

[0119]

[0120] Among them, P f represents the probability of occurrence of potential failure, α is a constant term, β and γ are the coefficients of time t and characteristic variable x respectively.

[0121] See also Figure 6 , the fault prediction results are applied to the fault response mechanism of the multi-graphics terminal circuit, the graphics card operating parameters are adjusted and the fault isolation program is activated, and the steps of obtaining the real-time feedback signal by updating the multi-graphics terminal circuit status in real time are as follows:

[0122] S501: Based on the fault prediction result, the fault response parameter is input into the graphics card terminal circuit, the operation parameters of the graphics card are adjusted to avoid the influence of the potential fault, and the preset fault isolation program is started. The execution process of obtaining the adjusted operation parameters is as follows;

[0123] Based on the fault prediction results, sub-step S501 inputs fault response parameters into the graphics card terminal circuit, including adjusting the graphics card operating parameters, such as voltage, current and frequency, to avoid the impact of potential faults. If a high-risk fault is predicted, start the preset fault isolation procedure, such as shutting down certain functions or switching to a backup circuit. It is necessary to monitor the status of the graphics card in real time to ensure that the adjusted parameters can effectively reduce the risk of failure and obtain the adjusted operating parameters. The formula used is:

[0124]

[0125] Among them, R a Represents the adjusted operating parameters, V op and I op are the operating voltage and current, p i is the priority of adjustment, ΔP i is the adjustment amplitude of the ith parameter, and n is the total number of adjusted parameters.

[0126] S502: using the adjusted operating parameters, cyclically tracking the operating status of the graphics card, extracting key data in the operation of the graphics card, and comparing the early warning indicators to evaluate the adjustment effect of the operating parameters, the execution process of generating status monitoring data is as follows;

[0127] Sub-step S502 uses the adjusted operating parameters to cyclically track the operating status of the graphics card, including extracting key data in the graphics card operation, such as temperature, load and frequency, and regularly comparing them with the preset early warning indicators to evaluate the adjustment effect of the operating parameters. Through data acquisition and analysis equipment, the operating status of the graphics card is monitored in real time, any abnormal changes are identified, and the working status and adjustment effect of the graphics card are recorded in detail to ensure that the optimization of operating parameters can effectively improve the performance of the graphics card and avoid the risk of failure. The status monitoring data is generated using the formula:

[0128]

[0129] Among them, S m represents the status monitoring score, f(t) is the actual operating status of the graphics card at time t, W(t) is the warning indicator value, and T is the monitoring period.

[0130] S503: According to the status monitoring data, the status display of the multi-graphics terminal circuit is updated, the real-time operation status and new abnormal alarm of the multi-graphics terminal circuit are identified and recorded, and the execution process of obtaining the real-time feedback signal is as follows;

[0131] Substep S503 updates the status display of the multi-graphics card terminal circuit according to the status monitoring data, collects and analyzes the operation data of the graphics card in real time, compares the new data with the historical data, and identifies and records any new abnormal alarms. Through visualization tools and monitoring panels, the status display of the graphics card is dynamically updated to help the operation and maintenance team quickly respond to and handle abnormal situations, ensure the stable operation of the graphics card system, and obtain real-time feedback signals. The formula used is:

[0132]

[0133] Among them, F r Indicates the accuracy of the real-time feedback signal, R j is the actual monitoring data, A j is the expected normal data and N is the total number of data points.

[0134] See also Figure 7 ,According to the real-time feedback signal, the fault response strategy is implemented, the maintenance protocol is initiated, the detected faults are isolated and repaired, and the effect of the fault repair is evaluated. The specific steps for obtaining the fault handling solution are as follows:

[0135] S601: Based on the real-time feedback signal, a fault response strategy is designed, and the pre-designed automatic fault isolation is started using the maintenance protocol, and an emergency notification is sent to perform manual intervention of circuit abnormalities, isolate key components and limit functions, and the execution process of obtaining the fault isolation result is as follows;

[0136] Substep S601 designs a fault response strategy based on real-time feedback signals, monitors the equipment in real time according to the preset maintenance protocol, starts the automatic fault isolation program, and sends an emergency notification to the operation and maintenance team, requiring manual intervention in circuit anomalies. The fault response strategy includes isolating key components, such as graphics cards or power supply components, and limiting related functions to prevent the spread of faults, and obtaining fault isolation results. The formula used is:

[0137]

[0138] Among them, F i represents the fault isolation effect, C k is the isolation success rate of the kth key component, R k is the corresponding isolation time, D l It is the limiting effect of the first restricted function, E l is the restriction duration, m and n are the number of critical components and restricted functions, respectively.

[0139] S602: Based on the fault isolation result, perform fault diagnosis on the multi-graphics card terminal circuit, analyze the fault source, use the repair tool and replacement parts to repair the graphics card and related circuits, and generate the fault repair record. The execution process is as follows;

[0140] Sub-step S602 performs fault diagnosis of the multi-graphics terminal circuit based on the fault isolation result, analyzes the fault source, and uses diagnostic tools to conduct a comprehensive inspection of the isolated graphics card and related circuits to determine the specific source and impact range of the fault. Repair the found fault points using repair tools and replacement parts, including replacing damaged components and repairing circuit connections, and generate a fault repair record using the formula:

[0141]

[0142] Among them, R f Indicates the fault repair effect, T i is the effect of using the i-th repair tool, S i is the repair time, P j is the performance recovery effect of the jth replacement component, Q j is the replacement time, p and q are the number of repair tools and replacement parts, respectively.

[0143] S603: Using the fault repair record, evaluating the performance and stability of the circuit after repair, re-running the diagnostic tool to verify the repair effect of the fault, checking the normal operation of the circuit parameters and functions, and obtaining the execution process of the fault handling solution is as follows;

[0144] Sub-step S603 uses the fault repair record to evaluate the performance and stability of the circuit after the repair, re-runs the diagnostic tool to verify the repair effect, ensures that all fault points have been repaired, and checks the normal operation of the circuit parameters and functions. Through a series of testing and verification steps, ensure that the graphics card and related circuits are in good working condition after the repair, meet the expected performance standards, and obtain a fault handling plan. The formula used is:

[0145]

[0146] Among them, P v Indicates the performance verification result, V k is the performance parameter of the kth test point after repair, B k is the baseline parameter before repair, and N is the total number of test points.

[0147] See also Figure 8 , a multi-graphics card terminal circuit anomaly detection system, the multi-graphics card terminal circuit anomaly detection system is used to execute the multi-graphics card terminal circuit anomaly detection method, the system includes:

[0148] The signal capture module is based on the multi-graphics card terminal circuit interface, collects current and voltage data to match different types of graphics cards, identifies the current and voltage signals when the graphics cards are running, screens the signal quality, and generates original circuit signal data;

[0149] The data processing module synchronizes the signal to the data processing center according to the original circuit signal data, integrates the waveform data, and sets the integration parameters to obtain the integrated waveform data;

[0150] The entropy value calculation module uses the integrated waveform data to convert the waveform data into the frequency domain, analyzes the frequency domain characteristics, calculates the entropy value and variation of the graphics card terminal circuit, and generates the entropy value analysis result;

[0151] The fault analysis module uses entropy analysis results to compare the entropy standard deviation of real-time data and historical data, identify entropy changes and deviations, and predict and classify fault modes to obtain fault prediction results;

[0152] Based on the fault prediction results, the fault response module adjusts the graphics card operating parameters, activates the fault isolation and maintenance protocol, implements the fault handling strategy, evaluates and updates the operating status of the graphics card terminal circuit, and obtains the fault handling solution.

[0153] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A method for detecting abnormal circuits in a multi-graphics card terminal, characterized in that: The following steps are involved: Capture the current and voltage signals of the multi-graphics terminal circuit through the circuit interface, synchronize the signals to the data processing center for initial waveform integration, and generate a circuit signal data set; Performing Fourier transform on the circuit signal data set, analyzing the frequency domain characteristics, extracting entropy information of the graphics card terminal circuit, calculating the entropy variation of each graphics card, and obtaining entropy analysis results; According to the entropy value analysis results, compare the entropy value standard deviation recorded in the real-time data and the historical data, identify the change and deviation of the entropy value, detect the abnormality of the multi-graphics terminal circuit, and generate trend comparison analysis results; Using the data in the trend comparison and analysis results, identifying and detecting abnormalities in the multi-graphics card terminal circuit, classifying potential circuit failure modes, and predicting failures in a future time period to generate failure prediction results; Applying the fault prediction result to the fault response mechanism of the multi-graphics card terminal circuit, adjusting the graphics card operation parameters and activating the fault isolation program, and obtaining a real-time feedback signal by updating the multi-graphics card terminal circuit state in real time; According to the real-time feedback signal, a fault response strategy is implemented, a maintenance protocol is initiated, the detected fault is isolated and repaired, and the effect of the fault repair is evaluated to obtain a fault handling solution.

2. The method for detecting circuit anomalies of a multi-graphics card terminal according to claim 1, characterized in that: The circuit signal data set includes current data, voltage data and timestamp, the entropy analysis result includes the entropy value of the graphics card, the entropy variation and the statistical distribution of the entropy value, the trend comparison analysis result includes the entropy change trend, the entropy standard deviation and the entropy deviation index, the fault prediction result includes the predicted fault type, the predicted fault time and the fault probability score, the real-time feedback signal includes the real-time status of the graphics card, the operating parameters and the fault isolation signal, and the fault handling plan includes fault isolation measures, maintenance protocols and fault repair effect evaluation.

3. The method for detecting circuit anomalies of a multi-graphics card terminal according to claim 1, characterized in that: The current and voltage signals of the multi-graphics terminal circuit are captured through the circuit interface, and the synchronization signals are sent to the data processing center for initial waveform integration. The specific steps for generating the circuit signal data set are as follows: Through the circuit interface, the current and voltage signals of the multi-graphics terminal circuit are collected, the sampling frequency is adjusted to match the circuit characteristics, and the noise is removed to generate the original data set; Based on the original data set, the data is transmitted to the data processing center in real time, the consistency of data transmission is checked through data encryption, the transmission rate and resource consumption are adjusted, and the synchronized current and voltage data are obtained; The synchronized current and voltage data are used to process and integrate the current and voltage data, adjust the window size and step size to optimize data processing efficiency and avoid information loss, identify key waveform features, and obtain a circuit signal data set.

4. The method for detecting abnormality of a multi-graphics card terminal circuit according to claim 1, characterized in that: The steps of performing Fourier transform on the circuit signal data set, analyzing the frequency domain characteristics, extracting the entropy value information of the graphics card terminal circuit, calculating the entropy variation of each graphics card, and obtaining the entropy value analysis result are specifically as follows: According to the circuit signal data set, a fast Fourier transform algorithm is used to convert the time domain signal into a frequency domain signal, identify and record key frequency components, and generate frequency domain characteristic data; Based on the frequency domain characteristic data, the distribution density of the differentiated frequency components is analyzed, the interval and step size of the entropy value calculation are adjusted, and the entropy value of each graphics card terminal circuit is extracted according to the distribution uniformity to obtain the graphics card entropy value information; The entropy value information of the graphics card is used to calculate the variation between the entropy values ​​of the graphics cards, compare the entropy value differences between differentiated graphics cards, and quantify the entropy value variation to obtain the entropy value analysis result.

5. The method for detecting circuit anomalies of a multi-graphics card terminal according to claim 4, characterized in that: The formula of the Fast Fourier Transform algorithm is as follows: Among them, X k is the kth frequency domain sample, x n is the nth time domain sample, N is the total number of samples, k is the index of the frequency domain sample, w is the window function, α is the modulation coefficient, is the cosine modulation part in the frequency domain.

6. The method for detecting circuit anomalies of a multi-graphics card terminal according to claim 1, characterized in that: According to the entropy value analysis result, the steps of comparing the entropy value standard deviation recorded in the real-time data and the historical data, identifying the change and deviation of the entropy value, detecting the abnormality of the multi-graphics terminal circuit, and generating the trend comparison analysis result are specifically as follows: Based on the entropy analysis results, the entropy standard deviation in the real-time data of the multi-graphics card terminal circuit is compared with the entropy standard deviation of the collected historical data, and the difference between the real-time and historical entropy values ​​is calculated by setting comparison parameters to check the consistency of the time range and the statistical method, so as to obtain the entropy change data; According to the entropy value change data, the change trend and deviation size of the entropy value are analyzed, the abnormal points of the multi-graphics card terminal circuit are analyzed according to the target threshold value, the data points beyond the normal range are marked, and the deviation identification result is obtained; The deviation identification result is used to summarize the abnormal points identified in the multi-graphics card terminal circuit, and compare and analyze them with historical trends. By recording the deviation between the abnormal points and the normal operating state, the trend comparison analysis result is obtained.

7. The method for detecting circuit anomalies of a multi-graphics card terminal according to claim 1, characterized in that: Using the data in the trend comparison and analysis results, the abnormalities of the multi-graphics terminal circuit are identified and detected, the potential failure modes of the circuit are classified, and the failures in the future time period are predicted. The steps for generating the failure prediction results are as follows: Based on the trend comparison and analysis results, identify and record the abnormal fluctuations that deviate, match the abnormal fluctuations with the characteristics of the known fault mode, mark the abnormal data points that match the fault mode, and obtain the fault mode recognition result; According to the fault mode identification results, the fault modes identified in the multi-graphics card terminal circuit are classified and sorted, the urgency of the fault mode is evaluated according to the frequency and impact of the abnormality, and risk sorting and priority setting are performed to generate fault classification and evaluation results; The fault classification and evaluation results are used to predict the probability and expected time of potential faults within the target time period, identify the risks and risk levels of potential faults, set warning levels and response strategies, and obtain fault prediction results.

8. The method for detecting circuit anomalies in a multi-graphics card terminal according to claim 1, characterized in that: The steps of applying the fault prediction result to the fault response mechanism of the multi-graphics card terminal circuit, adjusting the graphics card operation parameters and activating the fault isolation program, and obtaining the real-time feedback signal by updating the multi-graphics card terminal circuit state in real time are specifically as follows: Based on the fault prediction result, the fault response parameter is input into the graphics card terminal circuit, the operation parameter of the graphics card is adjusted to avoid the influence of the potential fault, and a preset fault isolation program is started to obtain the adjusted operation parameter; Using the adjusted operating parameters, cyclically tracking the operating status of the graphics card, extracting key data in the operation of the graphics card, and comparing the early warning indicators to evaluate the adjustment effect of the operating parameters, and generating status monitoring data; According to the status monitoring data, the status display of the multi-graphics terminal circuit is updated, the real-time operation status and new abnormal alarms of the multi-graphics terminal circuit are identified and recorded, and the real-time feedback signal is obtained.

9. The method for detecting circuit anomalies in a multi-graphics card terminal according to claim 1, characterized in that: According to the real-time feedback signal, the fault response strategy is implemented, the maintenance protocol is initiated, the detected fault is isolated and repaired, and the effect of the fault repair is evaluated. The steps of obtaining the fault handling solution are specifically as follows: Based on the real-time feedback signal, a fault response strategy is designed, a pre-designed automatic fault isolation is initiated using a maintenance protocol, and an emergency notification is sent to perform manual intervention of circuit abnormalities, isolate key components and limit functions, and obtain a fault isolation result; Based on the fault isolation results, perform fault diagnosis on the multi-graphics card terminal circuit, analyze the fault source, use repair tools and replacement parts to repair the graphics card and related circuits, and generate a fault repair record; The fault repair record is used to evaluate the performance and stability of the circuit after repair, re-run the diagnostic tool to verify the repair effect of the fault, check the normal operation of the circuit parameters and functions, and obtain the fault handling plan.

10. A multi-graphics card terminal circuit abnormality detection system, characterized in that: According to the method for detecting circuit anomalies in a multi-graphics card terminal according to any one of claims 1 to 9, the system comprises: The signal capture module is based on the multi-graphics card terminal circuit interface, collects current and voltage data to match different types of graphics cards, identifies the current and voltage signals when the graphics cards are running, screens the signal quality, and generates original circuit signal data; The data processing module transmits the synchronization signal to the data processing center according to the original circuit signal data, integrates the waveform data, and sets integration parameters to obtain the integrated waveform data; The entropy value calculation module uses the integrated waveform data to convert the waveform data into the frequency domain, analyzes the frequency domain characteristics, calculates the entropy value and variation of the graphics card terminal circuit, and generates an entropy value analysis result; The fault analysis module uses the entropy analysis result to compare the entropy standard deviation of real-time data and historical data, identify entropy changes and deviations, and predict and classify fault modes to obtain fault prediction results; Based on the fault prediction result, the fault response module adjusts the graphics card operating parameters, activates the fault isolation and maintenance protocol, implements the fault handling strategy, evaluates and updates the operating status of the graphics card terminal circuit, and obtains the fault handling solution.

Citation Information

Cited By

  • Method and system for detecting commercial password capability of FPGA (Field Programmable Gate Array) chip

    CN119135334A

  • Circuit board sensor nondestructive testing method based on big data

    CN119397468A

  • LED display screen HUB board communication detection method and system

    CN119652796A

  • Train control equipment intelligent maintenance method, maintenance system, medium and product

    CN120069845A