Main board fault detection method and computer device

Abstract

An embodiment of the present invention discloses a mainboard fault detection method and a computer device, belonging to the technical field of data analysis. By introducing innovative technical means such as time-frequency domain data extraction, sensing state trend vector pairing, and fault detection execution feature generation, the embodiment of the present invention realizes efficient and accurate fault detection of multi-mode mainboard sensing signals. In this way, it can solve the deficiencies of traditional fault detection methods in terms of accuracy, efficiency, and dealing with complex fault modes, and provides a new and more effective solution for the fault detection of the mainboard.

Description

Technical Field

[0001] The present invention belongs to the technical field of data analysis, and particularly relates to a motherboard fault detection method and a computer device. Background Art

[0002] The motherboard is a core component in electronic devices, and its quality and reliability are crucial for the performance of the entire device. Motherboard faults may include soldering problems, component shortages, circuit board damage, etc. These defects may lead to device failures, performance degradation, or even inability to operate normally. Therefore, accurately and reliably detecting motherboard faults is of great significance for ensuring the normal operation of the device.

[0003] In the field of motherboard fault detection, traditional detection methods often rely on a single sensing signal or simple signal feature extraction, resulting in limitations in the accuracy and efficiency of fault detection. Especially when facing complex and variable fault modes, traditional methods often fail to provide accurate and reliable detection results. Summary of the Invention

[0004] The present invention provides a motherboard fault detection method and a computer device, which can solve or partially solve the technical problems involved in the above background art.

[0005] An embodiment of the present invention provides a motherboard fault detection method, which is applied to a computer device. The method includes: obtaining multi-mode motherboard sensing signals to be subjected to fault detection, extracting time-frequency domain data from the multi-mode motherboard sensing signals to obtain time-frequency domain sensing signals associated with the multi-mode motherboard sensing signals; determining a target sensing state trend vector of the multi-mode motherboard sensing signals according to the time-frequency domain sensing signals; performing sensing state trend vector pairing between the target sensing state trend vector and the past sensing state trend vectors of past multi-mode motherboard sensing signals that have completed fault detection to obtain a trend vector pairing view; the past sensing state trend vectors are determined according to the past time-frequency domain sensing signals associated with the past multi-mode motherboard sensing signals; determining a multi-mode motherboard sensing reference signal of the multi-mode motherboard sensing signals from the past multi-mode motherboard sensing signals according to the trend vector pairing view; generating execution features for fault detection of the multi-mode motherboard sensing signals based on the fault detection conclusion view of the multi-mode motherboard sensing reference signal to obtain fault detection execution features of the multi-mode motherboard sensing signals; the fault detection execution features are used to indicate fault detection analysis for the multi-mode motherboard sensing signals.

[0006] An embodiment of the present invention provides a computer device, including at least one processor and a memory; the memory stores computer execution instructions; the at least one processor executes the computer execution instructions stored in the memory, so that the at least one processor executes the above method.

[0007] An embodiment of the present invention provides a readable storage medium, on which a program or instructions are stored, and when the program or instructions are executed by a processor, the steps of the above method are implemented.

[0008] In the embodiment of the present invention, through innovative technical means such as time-frequency domain data extraction, sensing state trend vector pairing, and fault detection execution feature generation, efficient and accurate fault detection of multimode motherboard sensing signals is achieved. In this way, the deficiencies of traditional fault detection methods in terms of accuracy, efficiency, and dealing with complex fault modes can be solved, providing a new and more effective solution for the fault detection of the motherboard. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 It is a flowchart of a motherboard fault detection method provided by an embodiment of the present invention.

[0010] Figure 2 It is a schematic structural diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0011] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0012] The terms "first", "second", etc. in the present invention are used to distinguish similar objects, rather than to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances, so that the embodiments of the present invention can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. generally belong to the same category, and the number of objects is not limited. For example, the first object can be one or multiple. In addition, "and / or" in the present invention means at least one of the connected objects, and the character " / " generally represents an "or" relationship between the associated objects before and after.

[0013] Figure 1 A motherboard fault detection method is shown, which is applied to a computer device, and the method includes the following steps 110-step 150.

[0014] In modern computer systems, the motherboard serves as a bridge connecting various hardware components, and its stability and reliability are crucial for the operation of the entire system. However, due to the integration of a large number of electronic components and complex circuits on the motherboard, it has also become a common source of system failures. To effectively detect and prevent motherboard failures, steps 110 - 150 introduce a multi-mode motherboard sensing signal fault detection technical solution based on time-frequency domain analysis and sensing state trend vector pairing. The following will elaborate on the implementation steps of this solution through specific application scenario examples.

[0015] The computer device first obtains the multi-mode motherboard sensing signals to be fault-detected through the sensor interface on the motherboard. These sensing signals can include data in various modes such as temperature, voltage, and current, which together reflect the real-time status of the motherboard under different working conditions. After obtaining the sensing signals, the computer device extracts time-frequency domain data from these signals. Time-frequency domain analysis is a method of converting a signal from the time domain to the frequency domain, and using time-frequency domain analysis can more clearly observe the frequency components in the signal and their changes over time. Specifically, the computer device uses algorithms such as the Fast Fourier Transform (FFT) or wavelet transform to convert the multi-mode motherboard sensing signals from the time domain to the frequency domain, obtaining time-frequency domain sensing signals.

[0016] After obtaining the time-frequency domain sensing signals, the computer device needs to further analyze these signals to determine the target sensing state trend vector of the multi-mode motherboard sensing signals. The sensing state trend vector is a feature vector that can reflect the change trend of the motherboard sensing signal state, and it contains key information such as the main frequency components, amplitude changes, and phase relationships of the signal. To determine the target sensing state trend vector, the computer device first extracts features from the time-frequency domain sensing signals, extracting the main frequency components and their corresponding amplitudes and phases of the signal. Then, based on these feature information, the computer device uses machine learning algorithms (such as support vector machines, neural networks, etc.) to classify and identify the state of the sensing signal, and finally determines the target sensing state trend vector.

[0017] After determining the target sensing state trend vector, the computer device then needs to pair it with the past sensing state trend vectors of the past multi-mode main board sensing signals for which fault detection has been completed. The purpose of this step is to find past sensing state trend vectors similar to the target sensing state trend vector, so as to use the fault detection experience of these past vectors to guide the fault detection of the current signal. To achieve trend vector pairing, the computer device first establishes a database containing a large number of past multi-mode main board sensing signals and their corresponding past sensing state trend vectors. Then, similarity calculation algorithms (such as cosine similarity, Euclidean distance, etc.) are used to evaluate the similarity between the target sensing state trend vector and the past sensing state trend vectors in the database. Finally, the computer device selects several past sensing state trend vectors with the highest similarity to the target sensing state trend vector as the pairing result.

[0018] After completing the pairing of the sensing state trend vectors, the computer device then needs to determine the sensing reference signals similar to the current multi-mode main board sensing signal from the past multi-mode main board sensing signals. These sensing reference signals are signals whose states or fault types have been confirmed in past fault detections, and they can provide valuable reference information for the fault detection of the current signal. To determine the sensing reference signals, the computer device first retrieves the past sensing signals similar to the target sensing state trend vector from the database according to the trend vector pairing result. Then, these past sensing signals are further analyzed and screened, and the signals that are most similar to the current multi-mode main board sensing signal in terms of time-frequency domain characteristics, working mode, environmental conditions, etc. are selected as the sensing reference signals.

[0019] Finally, based on the conclusion of the fault detection of the multi-mode motherboard sensing reference signal, the computer device generates execution features for the current multi-mode motherboard sensing signal. The purpose of this step is to convert the fault detection experience of the sensing reference signal into the fault detection execution features for the current signal, so as to perform subsequent fault detection analysis. During the generation of the execution features, the computer device first analyzes the conclusion of the fault detection of the sensing reference signal and extracts the information related to the fault type, fault degree, fault occurrence time, etc. Then, based on this information and the specific characteristics of the current multi-mode motherboard sensing signal, the computer device uses machine learning or deep learning algorithms to generate the fault detection execution features for the current signal. These execution features are a high-dimensional vector or matrix, which contain the key information such as the possible fault type, fault degree, and the time window of the fault occurrence of the current signal. After generating the fault detection execution features, the computer device can use these features to perform fault detection analysis. Specifically, the computer device can input the fault detection execution features into a trained fault detection model, and the model outputs the fault detection conclusion of the current multi-mode motherboard sensing signal. This conclusion can include information such as whether the signal has a fault, the type and degree of the fault, and the recommended repair measures. In this way, the computer device can achieve accurate and efficient fault detection of the multi-mode motherboard sensing signal.

[0020] In summary, through steps such as obtaining the multi-mode motherboard sensing signal, extracting time-frequency domain data, determining the target sensing state trend vector, pairing the sensing state trend vectors, determining the multi-mode motherboard sensing reference signal, and generating the fault detection execution features, accurate and efficient fault detection of the motherboard sensing signal is achieved. In this way, not only can the stability and reliability of the computer system be improved, but also the maintenance cost can be reduced and the service life of the equipment can be extended.

[0021] Next, combined with the above content, steps 110 - 150 will be introduced in detail respectively.

[0022] Step 110: Obtain the multi-mode motherboard sensing signal to be subjected to fault detection, and extract time-frequency domain data from the multi-mode motherboard sensing signal to obtain the time-frequency domain sensing signal associated with the multi-mode motherboard sensing signal.

[0023] In step 110, the multi-mode motherboard sensing signal to be subjected to fault detection is to obtain a series of sensing signals through the sensor interface on the motherboard, and these signals are the multi-mode motherboard sensing signals to be subjected to fault detection. These signals may include data in multiple modes such as temperature, voltage, and current, and they jointly reflect the real-time status of the motherboard under different working conditions. By performing fault detection analysis on these multi-mode motherboard sensing signals, abnormal conditions on the motherboard can be detected in a timely manner, and corresponding repair measures can be taken to ensure the stable operation of the system.

[0024] The time-frequency domain sensing signal is the result obtained after extracting time-frequency domain data from the original sensing signal. Time-frequency domain analysis is a method of converting a signal from the time domain to the frequency domain, which can help to more clearly observe the frequency components in the signal and their changes over time. In the application scenario of motherboard fault detection, the computer device extracts time-frequency domain data from the multi-mode motherboard sensing signal to obtain the time-frequency domain sensing signal. These time-frequency domain sensing signals contain information about the original sensing signal in the frequency domain, such as the main frequency components, amplitude changes, phase relationships, etc., which are crucial for subsequent motherboard fault detection and analysis.

[0025] Specifically, in the application scenario of motherboard fault detection, the computer device first needs to obtain the multi-mode motherboard sensing signal to be detected for faults. These signals are collected in real time through the sensor interface on the motherboard and may include data in multiple modes such as temperature, voltage, current, etc. After obtaining these sensing signals, the computer device extracts time-frequency domain data from them.

[0026] Time-frequency domain data extraction is a process of converting a signal from the time domain to the frequency domain. In this step, the computer device uses algorithms such as the Fast Fourier Transform (FFT) or wavelet transform to process the multi-mode motherboard sensing signal. These algorithms can convert the time information in the signal into frequency information, thereby obtaining the time-frequency domain sensing signal. The time-frequency domain sensing signal contains detailed information about the original sensing signal in the frequency domain, such as the main frequency components, amplitude changes, phase relationships, etc. These information are very important for subsequent motherboard fault detection and analysis because they can reflect the state change trend and potential fault characteristics of the motherboard sensing signal.

[0027] Through time-frequency domain data extraction, the computer device can obtain the time-frequency domain sensing signal associated with the multi-mode motherboard sensing signal. These time-frequency domain sensing signals provide an important data basis for subsequent fault detection and analysis, enabling the computer device to more accurately identify information such as the type of fault, the degree of fault, and the time window of fault occurrence on the motherboard. This is of great significance for improving the stability and reliability of the computer system, reducing maintenance costs, and extending the service life of the device.

[0028] Step 120: Determine the target sensing state trend vector of the multi-mode motherboard sensing signal based on the time-frequency domain sensing signal.

[0029] In step 120, the target sensing state trend vector is a feature vector that can comprehensively reflect the state change trend of the multi-mode motherboard sensing signal and provides key data support for subsequent fault detection. This vector is not simply composed of several numerical values, but covers multiple key information such as the main frequency components, amplitude changes, and phase relationships of the signal.

[0030] For example, there is a multi-mode motherboard sensing signal. Its time-frequency domain sensing signal shows that within a certain specific time period, the main frequency components of the signal are 50 Hz and 100 Hz, with corresponding amplitudes of 2 V and 1 V respectively, and the phase of the 50 Hz component shows a linear growth trend over time. Then, the target sensing state trend vector of this sensing signal can be represented as a vector containing these key information: [50 Hz, 2 V, linear growth; 100 Hz, 1 V,...]. Here, "..." indicates that there may be other minor frequency components or phase relationships and other information.

[0031] In this way, the target sensing state trend vector presents the state change trend of the multi-mode motherboard sensing signal in a mathematical form, providing great convenience for subsequent fault detection and analysis.

[0032] Specifically, the implementation of step 120 mainly depends on the in-depth analysis and processing of the time-frequency domain sensing signal. First, the computer device obtains the time-frequency domain sensing signal of the multi-mode motherboard sensing signal. These signals have been converted from the original time-domain signal to the frequency-domain signal through time-frequency domain data extraction technology, and contain multiple key information such as the main frequency components, amplitude changes, and phase relationships of the signal. Then, the computer device further analyzes and processes these time-frequency domain sensing signals. It extracts the main frequency components in the signal, as well as the amplitude and phase information corresponding to these frequency components. These information are important elements that make up the target sensing state trend vector. Then, the computer device constructs a target sensing state trend vector that can comprehensively reflect the state change trend of the multi-mode motherboard sensing signal according to the extracted information. This vector not only contains the main frequency components and amplitude information of the signal, but also reflects the change trend of these frequency components over time, as well as the phase relationship between them. In this way, the computer device can determine the target sensing state trend vector of the multi-mode motherboard sensing signal. This vector provides key data support for subsequent fault detection and analysis, enabling the computer device to more accurately identify potential faults on the motherboard and take corresponding maintenance measures in a timely manner.

[0033] Step 130: Pair the target sensing state trend vector with the past sensing state trend vectors of past multi-mode motherboard sensing signals that have completed fault detection to obtain a trend vector pairing view; the past sensing state trend vectors are determined based on the past time-frequency domain sensing signals associated with the past multi-mode motherboard sensing signals.

[0034] In step 130, the trend vector pairing view refers to the view or conclusion about the similarity or difference degree between the target sensing state trend vector and the past sensing state trend vectors of the past multi-mode main board sensing signals for which fault detection has been completed, obtained after pairing and comparing them. This view or conclusion is of great significance for judging whether there are potential faults in the current main board.

[0035] For example, there is a target sensing state trend vector, which reflects the state change trend of the current multi-mode main board sensing signal. At the same time, there is a past sensing state trend vector library, which stores a large number of trend vectors of the past multi-mode main board sensing signals for which fault detection has been completed. By pairing and comparing the target trend vector with the vectors in the past trend vector library, a trend vector pairing view about the similarity between the two can be obtained. If the target trend vector is highly similar to a certain past fault trend vector, then it can be preliminarily judged that the current main board may have a similar fault.

[0036] Specifically, the core task of step 130 is to pair and compare the target sensing state trend vector with the past sensing state trend vectors of the past multi-mode main board sensing signals for which fault detection has been completed, so as to obtain the trend vector pairing view. First, the computer device obtains the target sensing state trend vector of the current multi-mode main board sensing signal. This vector is obtained by deeply analyzing and processing the current sensing signal, and it comprehensively reflects the state change trend of the signal. Then, the computer device accesses the past sensing state trend vector library. This library stores a large number of trend vectors of the past multi-mode main board sensing signals for which fault detection has been completed. These past trend vectors are determined based on the past time-frequency domain sensing signals, and they represent the sensing signal trends in various known fault states. Then, the computer device pairs and compares the target sensing state trend vector with the vectors in the past sensing state trend vector library. This comparison process will consider each dimension of the vector, including frequency components, amplitude changes, phase relationships, etc., to comprehensively evaluate the similarity or difference degree between the two. Finally, the computer device generates a trend vector pairing view according to the result of the pairing comparison. This view will clearly point out the similarity or difference degree between the target trend vector and the past trend vector, thus providing an important reference basis for subsequent fault detection and analysis. If the target trend vector is highly similar to a certain past fault trend vector, then the computer device can preliminarily judge that the current main board may have a similar fault and take corresponding maintenance measures.

[0037] Step 140: Determine the multi-mode main board sensing reference signal of the multi-mode main board sensing signal from the past multi-mode main board sensing signals according to the trend vector pairing view.

[0038] In step 140, the multi-mode motherboard sensing reference signal is a crucial concept. It refers to the signal selected from past multi-mode motherboard sensing signals that has a high degree of similarity or referenceability in terms of the state change trend with the current multi-mode motherboard sensing signal to be detected. This reference signal provides an important benchmark and basis for subsequent fault detection and diagnosis.

[0039] For example, currently, the sensing signal of a multi-mode motherboard is being detected to determine whether there is a fault. During the detection process, it is found that the sensing signal of this motherboard is similar to the sensing signal of a known faulty motherboard in some aspects. Then, the sensing signal of this known faulty motherboard can be used as the "multi-mode motherboard sensing reference signal" for the current motherboard to be detected. By comparing the current signal with the reference signal, it is possible to more accurately determine whether there is a fault in the current motherboard, as well as the type and degree of the fault.

[0040] During the process of motherboard fault detection, the core task of step 140 is to determine the sensing reference signal similar to the current multi-mode motherboard sensing signal from past multi-mode motherboard sensing signals according to the trend vector pairing view. First, the computer device obtains the trend vector pairing view generated in the previous step. This view reflects the similarity or difference degree between the current multi-mode motherboard sensing signal and the past sensing state trend vectors. Then, based on this trend vector pairing view, the computer device starts to search the past multi-mode motherboard sensing signal library for signals similar to the current signal. This search process takes into account various dimensions of the signal, including frequency components, amplitude changes, phase relationships, etc., to ensure that the found signal has a high degree of similarity in the state change trend with the current signal. Then, the computer device selects the past multi-mode motherboard sensing signal that is most similar to the current signal as the multi-mode motherboard sensing reference signal for the current signal. This reference signal will serve as an important benchmark for subsequent fault detection and diagnosis. Finally, the computer device uses this multi-mode motherboard sensing reference signal to conduct a more in-depth analysis and diagnosis of the current multi-mode motherboard sensing signal. By comparing the differences between the current signal and the reference signal, the computer device can more accurately determine whether there is a fault in the current motherboard, as well as the type and degree of the fault. Thus, it provides strong support for subsequent repair and replacement work.

[0041] Step 150: Based on the fault detection conclusion view of the multi-mode motherboard sensing reference signal, perform feature generation on the multi-mode motherboard sensing signal to obtain the fault detection execution feature of the multi-mode motherboard sensing signal; the fault detection execution feature is used to indicate the fault detection analysis for the multi-mode motherboard sensing signal.

[0042] In step 150, the fault detection conclusion view is a judgment or conclusion about whether there is a fault in the motherboard and the type of the fault, which is obtained based on the analysis and comparison of the sensing signals of the multi-mode motherboard. This view is the direct output of the fault detection process and provides clear guidance for subsequent repair and replacement work.

[0043] For example, after a series of analysis and comparison, if the computer device finds that the current sensing signals of the multi-mode motherboard are highly similar to a sensing reference signal indicating a memory fault, then the fault detection conclusion view may be that "there is a memory fault in the current motherboard".

[0044] The fault detection execution features are the feature information extracted from the sensing signals of the multi-mode motherboard according to specific detection methods and algorithms when performing motherboard fault detection, which is used to indicate the process and execution status of the fault detection analysis. These feature information may include the frequency components, amplitude changes, phase relationships, etc. of the signals, and they together constitute the basis of the fault detection analysis.

[0045] For example, when performing fault detection, the computer device can extract specific frequency components and amplitude changes from the sensing signals of the multi-mode motherboard as the fault detection execution features, and these features will be used for subsequent fault analysis and judgment.

[0046] In the process of motherboard fault detection, step 150 generates execution features for the current sensing signals of the multi-mode motherboard based on the fault detection conclusion view of the previously determined sensing reference signals of the multi-mode motherboard, in order to obtain the fault detection execution features for indicating the fault detection analysis. First, the computer device reviews the fault detection conclusion view of the sensing reference signals of the multi-mode motherboard obtained in the previous steps. This view provides clear guidance for the current step and indicates which aspects of the sensing signal features need to be concerned. Second, based on this fault detection conclusion view, the computer device begins to extract execution features from the current sensing signals of the multi-mode motherboard. This extraction process will consider multiple dimensions of the signals, including frequency components, amplitude changes, phase relationships, etc., to ensure that the extracted features can comprehensively reflect the state and change trend of the signals. Then, the computer device organizes and combines these extracted execution features to form the fault detection execution features for indicating the fault detection analysis. This execution feature is a comprehensive information set that contains all the important information of the current sensing signals of the multi-mode motherboard in the fault detection process. Finally, the computer device uses this fault detection execution feature for subsequent fault analysis and judgment. By comparing the differences between the execution feature and the known fault features, the computer device can more accurately judge whether there is a fault in the current motherboard and the type and degree of the fault. Thus, it provides strong support for subsequent repair and replacement work.

[0047] It can be understood that the embodiments of the present invention significantly improve the fault detection efficiency and accuracy of multi-mode motherboard sensing signals through a series of innovative technical means. Specifically, the embodiments of the present invention first obtain the multi-mode motherboard sensing signals to be detected and extract the time-frequency domain data thereof, so as to obtain the time-frequency domain sensing signals closely related to the multi-mode motherboard sensing signals. This step provides rich and comprehensive signal feature information for subsequent fault detection.

[0048] Next, the embodiments of the present invention determine the target sensing state trend vector of the multi-mode motherboard sensing signals according to the extracted time-frequency domain sensing signals. The introduction of this vector enables the embodiments of the present invention to more accurately grasp the state change trend of the sensing signals and provides a strong basis for fault detection.

[0049] On this basis, the embodiments of the present invention pair the target sensing state trend vector with the past sensing state trend vectors of the past multi-mode motherboard sensing signals that have completed fault detection, so as to obtain the trend vector pairing view. The realization of this step enables the embodiments of the present invention to make full use of the past fault detection experience and provide an accurate fault detection reference for the current multi-mode motherboard sensing signals.

[0050] Furthermore, the embodiments of the present invention determine the sensing reference signals similar to the current multi-mode motherboard sensing signals from the past multi-mode motherboard sensing signals according to the trend vector pairing view. The completion of this step provides a direct basis and reference for subsequent fault detection analysis.

[0051] Finally, the embodiments of the present invention generate execution features for the current multi-mode motherboard sensing signals based on the fault detection conclusion view of the multi-mode motherboard sensing reference signals, so as to obtain the fault detection execution features for indicating fault detection analysis. The realization of this step enables the embodiments of the present invention to more accurately judge whether there are faults in the current multi-mode motherboard sensing signals and the type and degree of the faults, and provides strong support for subsequent maintenance and replacement work.

[0052] In summary, the embodiments of the present invention realize the efficient and accurate fault detection of multi-mode motherboard sensing signals through innovative technical means, significantly improve the efficiency and accuracy of fault detection, and have broad application prospects and important practical values.

[0053] In an alternative embodiment, determining the target sensing state trend vector of the multi-mode main board sensing signal based on the time-frequency domain sensing signal includes: pairing the time-frequency domain sensing signal with each sensing signal index respectively to obtain a signal index pairing result; the sensing signal index is obtained by index-based clustering of the past time-frequency domain sensing signals associated with the past multi-mode main board sensing signals; determining the sensing signal index in the signal index pairing result representing the completion of pairing as the target sensing signal index to which the time-frequency domain sensing signal belongs; and obtaining the target sensing state trend vector of the multi-mode main board sensing signal based on the sensing state deduction characteristics of the target sensing signal index.

[0054] In this embodiment, when a computer device performs a fault detection on a multi-mode main board sensing signal, it needs to determine the target sensing state trend vector of the multi-mode main board sensing signal based on the time-frequency domain sensing signal. The realization of this process not only involves in-depth analysis of the time-frequency domain sensing signal, but also relies on the experience and knowledge of past multi-mode main board sensing signals.

[0055] First, the computer device acquires the multi-mode main board sensing signals to be subjected to fault detection and extracts time-frequency domain data from these signals. The purpose of this step is to convert the multi-mode main board sensing signals from the original time domain or frequency domain to the time-frequency domain, so as to more comprehensively capture the characteristics and changes of the signals. After the time-frequency domain data extraction is completed, the computer device obtains the time-frequency domain sensing signals closely associated with the multi-mode main board sensing signals.

[0056] Next, the computer device needs to pair these time-frequency domain sensing signals with each sensing signal index. The sensing signal index is obtained by index-based clustering of the past time-frequency domain sensing signals associated with the past multi-mode main board sensing signals. This means that each sensing signal index represents a group of past time-frequency domain sensing signals with similar characteristics or change trends. By pairing the current time-frequency domain sensing signal with these sensing signal indexes, the computer device can find the group of past signals most similar to the current signal.

[0057] After the pairing is completed, the computer device obtains a signal index pairing result. This result contains all the sensing signal indexes paired with the current time-frequency domain sensing signal. The computer device further analyzes this result and determines the sensing signal index in the signal index pairing result representing the completion of pairing as the target sensing signal index to which the current time-frequency domain sensing signal belongs. This target sensing signal index represents the group of signals most similar to the current time-frequency domain sensing signal in past experience, and thus also contains the most likely fault mode and change trend of the current signal.

[0058] Finally, based on the sensing state deduction features indexed by the target sensing signal of the computer device, the target sensing state trend vector of the multi-mode main board sensing signal is obtained. The sensing state deduction features are extracted based on the past fault detection experience and knowledge of the multi-mode main board sensing signal, which represent the typical change trends and features of the signal group during the fault development process. By applying these deduction features to the current multi-mode main board sensing signal, the computer device can obtain a target sensing state trend vector that can accurately reflect the change trend of the signal state.

[0059] The generation of the target sensing state trend vector indicates that the computer device has completed the preliminary fault detection and analysis of the multi-mode main board sensing signal. Next, the computer device will use this target sensing state trend vector for further fault detection and diagnosis.

[0060] Specifically, the computer device performs a sensing state trend vector pairing of the target sensing state trend vector with the past sensing state trend vectors of the past multi-mode main board sensing signals for which fault detection has been completed. The purpose of this pairing process is to find the past sensing state trend vector that is most similar to the current target sensing state trend vector, so as to use the past fault detection experience and knowledge for more accurate fault detection and analysis of the current multi-mode main board sensing signal.

[0061] After the pairing is completed, the computer device obtains a trend vector pairing view. This view reflects the similarity and difference between the current multi-mode main board sensing signal and the past sensing state trend vectors, providing an important reference basis for subsequent fault detection and analysis.

[0062] Then, based on the trend vector pairing view, the computer device determines the multi-mode main board sensing reference signal of the current multi-mode main board sensing signal from the past multi-mode main board sensing signals. This reference signal is one or a group of signals in the past multi-mode main board sensing signals that are most similar to the current signal, and it contains the most likely fault modes and features of the current signal. By using this reference signal, the computer device can perform more in-depth fault detection and analysis on the current multi-mode main board sensing signal.

[0063] Finally, based on the fault detection conclusion view of the multi-mode main board sensing reference signal, the computer device generates execution features for the current multi-mode main board sensing signal. This execution feature is a comprehensive information set that contains all the important information and features of the current multi-mode main board sensing signal during the fault detection process. By analyzing and comparing the differences between this execution feature and the known fault features, the computer device can more accurately determine whether there is a fault in the current multi-mode main board sensing signal and the type and degree of the fault.

[0064] Thus, through a series of innovative technical means and methods such as time-frequency domain data extraction, sensing signal index pairing, target sensing state trend vector generation, trend vector pairing view determination, and fault detection execution feature generation, the efficient and accurate fault detection and analysis of multi-mode mainboard sensing signals are realized. This not only improves the accuracy and efficiency of fault detection, but also provides strong support and guidance for subsequent repair and replacement work.

[0065] In another alternative embodiment, the pairing of the time-frequency domain sensing signal with each sensing signal index respectively to obtain a signal index pairing result includes: respectively determining the difference discriminant variable between the time-frequency domain sensing signal and each sensing signal index; obtaining the signal index pairing result based on the difference discriminant variable; and determining the sensing signal index in the signal index pairing result representing the completion of pairing as the target sensing signal index to which the time-frequency domain sensing signal belongs, including: determining the sensing signal index in the signal index pairing result whose included difference discriminant variable is not greater than the discriminant variable threshold as the target sensing signal index to which the time-frequency domain sensing signal belongs.

[0066] In the actual implementation process, when a computer device executes the fault detection task of multi-mode mainboard sensing signals, it faces a core challenge: how to accurately find the signal that best matches the current time-frequency domain sensing signal from a large amount of past sensing signals in order to effectively identify and judge the fault mode. For this reason, this embodiment proposes an innovative signal pairing and index determination strategy, and the following is a detailed elaboration of this strategy.

[0067] First, the computer device acquires the multi-mode mainboard sensing signal to be detected and conducts in-depth time-frequency domain data extraction on it. This step is crucial because it can convert the original sensing signal into data containing rich time-frequency features, providing a solid foundation for subsequent signal pairing. After completing the data extraction, the computer device obtains a time-frequency domain sensing signal closely related to the multi-mode mainboard sensing signal.

[0068] Next, the computer device needs to perform a fine pairing of this time-frequency domain sensing signal with each sensing signal index. The sensing signal index is obtained through an index-based clique technique based on past multi-mode mainboard sensing signals and their associated time-frequency domain sensing signals. Each sensing signal index represents a group of past sensing signals with similar characteristics or change patterns. In order to find the signal index that best matches the current time-frequency domain sensing signal, the computer device calculates the difference discriminant variable between the current signal and each sensing signal index respectively. This difference discriminant variable is a quantitative index used to measure the similarity or difference degree between the current signal and each sensing signal index.

[0069] In the process of calculating the difference discrimination variable, the computer device comprehensively considers various characteristics of the time-frequency domain sensing signal, including frequency components, amplitude changes, phase relationships, etc. By comparing and analyzing these characteristics with the corresponding characteristics in the sensing signal index, the computer device can obtain a discrimination variable that can accurately reflect the degree of difference between the current signal and each sensing signal index.

[0070] After obtaining the difference discrimination variables, the computer device generates signal index pairing results based on these variables. This process is actually a process of screening and sorting. The computer device first screens out those signal index pairing results whose difference discrimination variables are not greater than a preset discrimination variable threshold. This threshold is an empirical value, which represents the maximum acceptable degree of difference between the current signal and the past sensing signals. Only those signal index pairing results whose degree of difference is less than or equal to this threshold will be retained.

[0071] Then, the computer device further sorts and processes these screened signal index pairing results. The basis for sorting is the magnitude of the difference discrimination variable, that is, the similarity between the current signal and the sensing signal index. The signal index pairing results with higher similarity will be ranked in a more forward position. In this way, the computer device can obtain a list of signal index pairing results sorted from high to low according to similarity.

[0072] Next, the computer device needs to determine the sensing signal index in the signal index pairing result list as the target sensing signal index to which the current time-frequency domain sensing signal belongs. This process is actually a process of selection. The computer device selects those sensing signal indices that are ranked at the top and have the smallest difference discrimination variable as the target sensing signal indices. These target sensing signal indices represent the past sensing signal groups that are most similar to the current time-frequency domain sensing signal and are most likely to contain the current signal fault mode.

[0073] Finally, the computer device generates a target sensing state trend vector for the current multi-mode main board sensing signal based on the sensing state deduction characteristics of these target sensing signal indices. The sensing state deduction characteristics are extracted from the past multi-mode main board sensing signals, and they represent the typical change trends and characteristics of the signal group during the fault development process. By applying these deduction characteristics to the current multi-mode main board sensing signal, the computer device can obtain a target sensing state trend vector that can accurately reflect the change trend of the signal state. The generation of this vector marks that the computer device has completed the preliminary fault detection and analysis of the multi-mode main board sensing signal.

[0074] Thus, through a series of innovative technical means and methods such as introducing differential discriminant variables, signal index pairing results, and determination of target sensing signal indices, efficient and accurate fault detection and analysis of multimode motherboard sensing signals are achieved. In this way, not only the accuracy and efficiency of fault detection are improved, but also strong support and guidance are provided for subsequent repair and replacement work. Through this method, the computer device can quickly find the group of past sensing signals that best match the current time-frequency domain sensing signal, and conduct in-depth fault detection and analysis of the current signal based on the fault experience and knowledge of these signal groups. As a result, not only is the time for fault detection greatly shortened, but also the accuracy and reliability of fault detection are improved.

[0075] In some other possible embodiments, the pairing of the target sensing state trend vector with the past sensing state trend vectors of past multimode motherboard sensing signals that have completed fault detection to obtain a trend vector pairing view includes: obtaining the past sensing state trend vectors of past multimode motherboard sensing signals that have completed fault detection; determining the Minkowski similarity between the target sensing state trend vector and the past sensing state trend vectors; and obtaining a trend vector pairing view based on the Minkowski similarity.

[0076] In the actual application process, when the computer device executes the fault detection task of multimode motherboard sensing signals, it faces an important challenge: how to effectively match and compare the current multimode motherboard sensing signals with the past multimode motherboard sensing signals that have completed fault detection in order to accurately judge the fault type and degree of the current signal. For this reason, this embodiment proposes an innovative sensing state trend vector pairing strategy.

[0077] First, the computer device needs to obtain the past sensing state trend vectors of past multimode motherboard sensing signals that have completed fault detection. These past sensing state trend vectors are extracted from historical data, and they represent the typical change trends and characteristics of past multimode motherboard sensing signals during the fault development process. Each past sensing state trend vector contains a series of eigenvalue, which can comprehensively describe the state and change trend of the past signal.

[0078] After obtaining the past sensing state trend vectors, the computer device needs to determine the similarity between the target sensing state trend vector of the current multimode motherboard sensing signal and these past sensing state trend vectors. To accurately measure the similarity, this embodiment uses the Minkowski similarity as the metric standard. The Minkowski similarity is a commonly used vector similarity measurement method, which can comprehensively consider each dimension of the vector and give a quantitative similarity value.

[0079] In the process of calculating the Minkowski similarity, the computer device first calculates the absolute value of the difference between the corresponding elements of the target sensing state trend vector and each past sensing state trend vector. Then, these absolute values are raised to the power of p, and the results of all dimensions are added together. Finally, the p-th root of the added result is taken to obtain the Minkowski similarity. Here, p is a real number greater than 1, which represents the accuracy and sensitivity of the similarity calculation. The larger the value of p, the more sensitive the similarity calculation is to the differences between vectors.

[0080] By calculating the Minkowski similarity, the computer device can obtain a quantified value that can accurately reflect the similarity degree between the target sensing state trend vector and each past sensing state trend vector. The higher the similarity, the closer the current multi-mode mainboard sensing signal is to the past sensing signals in terms of fault mode and change trend.

[0081] Based on the calculated Minkowski similarity, the computer device can further obtain the trend vector pairing view. The trend vector pairing view is a comprehensive judgment result that represents the matching degree and similarity between the current multi-mode mainboard sensing signal and the past sensing signals. Specifically, the computer device will take the several past sensing state trend vectors with the highest similarity as the signal group most matching the current signal, and use the fault types and degrees of these signal groups as the fault detection reference for the current signal.

[0082] In practical applications, the computer device can set a similarity threshold to judge the matching degree between the current signal and the past signals. If the highest similarity is greater than this threshold, it means that the current signal is very well matched with the past signals, and the computer device can use the fault types and degrees of the past signals as the fault detection conclusion for the current signal. If the highest similarity is less than this threshold, it means that the current signal is not well matched with the past signals, and the computer device needs to further analyze the characteristics and change trends of the current signal to more accurately judge its fault type and degree.

[0083] In addition to the similarity threshold, the computer device can also consider other factors to obtain the trend vector pairing view. For example, the computer device can analyze factors such as the temporal proximity between the current signal and the past signals, the frequency similarity, and the amplitude change degree between them to further judge the fault type and degree of the current signal.

[0084] Therefore, by introducing a series of innovative technical means and methods such as the Minkowski similarity and the trend vector pairing view, the effective matching and comparison between the current multi-mode mainboard sensing signal and the past multi-mode mainboard sensing signals that have completed fault detection are realized. This not only improves the accuracy and efficiency of fault detection but also provides strong support and guidance for subsequent repair and replacement work.

[0085] In practical applications, computer devices can improve the accuracy and reliability of fault detection by continuously adjusting and optimizing similarity calculation parameters, similarity thresholds, and other related factors. For example, a computer device can select an appropriate p value to calculate the Minkowski similarity according to the actual application scenario and requirements; it can also set different similarity thresholds according to different fault types and degrees to determine the matching degree between the current signal and the past signal. In addition, it can be combined with other fault detection techniques and methods to further improve the accuracy and efficiency of fault detection. For example, a computer device can first use traditional fault detection techniques and methods to conduct preliminary detection and analysis of the current signal; then use the method proposed in this embodiment to match and compare the current signal with the past signal; finally, comprehensively obtain the final fault detection conclusion based on the detection results of the two methods.

[0086] Thus, by introducing an innovative sensing state trend vector pairing strategy and the Minkowski similarity metric method, an effective matching and comparison between the current multi-mode motherboard sensing signal and the past multi-mode motherboard sensing signals that have completed fault detection are achieved. This not only provides a more accurate and reliable basis and support for fault detection; but also brings greater convenience and benefits to subsequent maintenance and replacement work.

[0087] In some alternative embodiments, the trend vector pairing view includes the Minkowski similarity; according to the trend vector pairing view, determining the multi-mode motherboard sensing reference signal of the multi-mode motherboard sensing signal from the past multi-mode motherboard sensing signals includes: obtaining a preset similarity threshold; determining, from the trend vector pairing view, the target trend vector pairing view whose included Minkowski similarity is higher than the preset similarity threshold; and determining the past multi-mode motherboard sensing signal associated with the target trend vector pairing view as the multi-mode motherboard sensing reference signal of the multi-mode motherboard sensing signal.

[0088] In the embodiment of the present invention, processing and analyzing multi-mode motherboard sensing signals is a complex but crucial task. This process aims to find the reference signal most similar to the current signal from a large amount of past sensing signal data for further analysis, prediction, or decision support. This embodiment achieves this goal by introducing the trend vector pairing view, especially the Minkowski similarity as the core index.

[0089] First, the computer device needs to obtain and analyze the current multi-mode motherboard sensing signals. The multi-mode motherboard sensing signals refer to the data collected from multiple sensors on the computer motherboard, which may include various types of physical quantities such as temperature, voltage, current, etc. These data together reflect the working state of the motherboard. The purpose of analyzing these signals is to identify patterns, trends, or anomalies therein, so as to provide a basis for subsequent processing.

[0090] Next, the computer device turns to process the historical multi-mode motherboard sensing signal data. These data were collected from the same or similar motherboards before, so they contain a large amount of historical information and potential patterns. The first step in processing these data is to extract feature vectors, that is, to convert the original sensing signal data into a vector form that can represent its key characteristics. The selection of feature vectors is crucial for subsequent analysis because they will directly affect the accuracy and efficiency of similarity calculation.

[0091] After the feature vectors are extracted, the computer device will use a method called the trend vector pairing view to further process these data. The trend vector pairing view is an analysis technique that identifies similarities or differences in data by associating feature vectors with specific trends or patterns. In this embodiment, special attention is paid to the Minkowski similarity metric.

[0092] The Minkowski similarity is a measure method to measure the similarity degree between two vectors. It considers the differences between the elements in the vectors and calculates an overall similarity score based on these differences. The higher this score, the more similar the two vectors are; conversely, the less similar they are. When calculating the Minkowski similarity, different parameters can be selected to adjust the calculation method of similarity to adapt to different application scenarios and data characteristics.

[0093] With the trend vector pairing view and the Minkowski similarity as the basis, the computer device will then perform a key step: determining the reference signal that is most similar to the current signal from the historical multi-mode motherboard sensing signals. This process first involves setting a preset similarity threshold, which is a threshold used to screen out those historical signals that are similar enough to the current signal.

[0094] The selection of the preset similarity threshold is an important decision point because it directly affects the quantity and quality of the finally determined reference signals. If the threshold is set too high, there may not be enough historical signals to meet the conditions, resulting in not being able to find enough references; while if the threshold is set too low, it may introduce too many irrelevant or noisy signals, reducing the effectiveness of the reference signals.

[0095] Once the preset similarity threshold is determined, the computer device will start traversing all the trend vector paired viewpoints and calculate their Minkowski similarity with the current signal feature vector. This calculation process needs to take into account the weights and distribution characteristics of each element in the feature vector to ensure the accurate calculation of similarity.

[0096] After the calculation is completed, those trend vector paired viewpoints with Minkowski similarity higher than the preset similarity threshold will be screened out as target trend vector paired viewpoints. The past multi-mode motherboard sensing signals associated with these target viewpoints are the reference signals that the computer device is looking for.

[0097] Finally, the computer device will use these reference signals for further analysis, prediction, or decision support. For example, these reference signals can be used to train a machine learning model to better understand and predict the future working state of the motherboard; or they can be compared with the current signals to detect any possible anomalies or faults.

[0098] In this way, by introducing the trend vector paired viewpoint and Minkowski similarity as the core metrics, an effective method for the computer device to process and analyze multi-mode motherboard sensing signals is provided. This method can not only accurately find the reference signals most similar to the current signal from a large amount of past data, but also provide a strong data basis for subsequent analysis, prediction, or decision support. In practical applications, it can significantly improve the computer device's monitoring and management capabilities of the motherboard working state, thereby providing a more stable and reliable computing environment for users.

[0099] Furthermore, to specifically illustrate the implementation details and effects of this technical solution, some specific numerical values and scenarios can be given. For example, the computer device has collected 1000 pieces of past multi-mode motherboard sensing signal data, and each piece of data contains 10 feature vector elements, representing different physical quantities. After extracting the feature vectors, the computer device uses the trend vector paired viewpoint to convert these data into a series of paired viewpoints, and each viewpoint contains a specific trend description and a related feature vector.

[0100] Next, when the computer device receives a new multi-mode motherboard sensing signal, it will first extract the feature vector of this signal and calculate the Minkowski similarity between this vector and all past signal feature vectors. The preset similarity threshold is set to 0.85, which means that only those past signals with similarity higher than 0.85 to the current signal will be selected as reference signals.

[0101] In the process of calculating similarity, the computer device may find that the similarity of 50 past signals is higher than 0.85, and these signals are thus determined as reference signals. These reference signals may contain the sensing data of the motherboard under different working conditions before. Their high similarity with the current signal indicates that the current motherboard may be in a similar working state or facing similar challenges.

[0102] With these reference signals, the computer device can further analyze them to extract more useful information. For example, it can compare the differences between these reference signals and the current signal in each element of the feature vector to identify any possible anomalies or change trends. Or, it can also use these reference signals to train a machine learning model, which can be used to predict the future working state or failure risk of the motherboard.

[0103] Through such a processing and analysis process, the computer device can more accurately understand and respond to the sensing signal data of the motherboard, thereby providing users with more efficient and reliable computing services. Whether in performance monitoring, fault prediction or status management, this technical solution can bring significant improvements and advantages to the computer device.

[0104] In some alternative embodiments, based on the fault detection conclusion view of the multi-mode motherboard sensing reference signal, performing feature generation on the multi-mode motherboard sensing signal to obtain the fault detection execution feature of the multi-mode motherboard sensing signal, including: obtaining the fault detection conclusion view of the multi-mode motherboard sensing reference signal; performing global analysis on the fault detection conclusion view to obtain the fault detection conclusion heat distribution; and generating an execution feature for the multi-mode motherboard sensing signal based on the fault detection conclusion heat distribution to obtain the fault detection execution feature of the multi-mode motherboard sensing signal.

[0105] In the multi-mode motherboard sensing signal fault detection task executed by the computer device, one of the core technical links is to generate execution features for these sensing signals based on the fault detection conclusion view of the multi-mode motherboard sensing reference signal, so as to obtain the fault detection execution features that can reflect the motherboard state and potential fault information. In this way, it not only depends on the accurate capture and processing of the sensing signal itself, but also needs to combine advanced fault detection algorithms and heat distribution analysis techniques to ensure that the finally obtained execution features are both comprehensive and accurate.

[0106] First, the computer device collects sensing reference signals from the multi-mode motherboard. These signals cover multiple dimensions such as temperature, voltage, and current, and are important data reflecting the operating state of the motherboard. During the collection process, the device uses high-precision sensors to ensure the accuracy and reliability of the data, and at the same time adopts real-time sampling technology to ensure the timeliness and continuity of the signals. The collected sensing reference signals will be sent to the next fault detection process.

[0107] Next, the computer device performs fault detection on these sensing reference signals to generate conclusion views. This process involves complex algorithm models, such as machine learning models, deep learning models, or rule-based expert systems. These models will analyze whether the current sensing signals are normal or whether there are abnormal conditions deviating from the expected range based on the pre-trained knowledge base and historical data. The output of the fault detection algorithm is a series of health status evaluations of various parts of the motherboard, that is, the fault detection conclusion views, which constitute a preliminary judgment of the overall state of the motherboard.

[0108] After obtaining the fault detection conclusion views, the computer device conducts a global analysis with the goal of obtaining an intuitive fault detection conclusion heat map. This step is the key bridge connecting the original sensing signals and the final execution features. Through comprehensive consideration of all fault detection conclusion views, the global analysis identifies the fault risk levels of different regions or components on the motherboard. The heat map represents the level of risk with the depth of color. For example, red represents high-risk areas, and green represents low-risk or risk-free areas. This visual expression method enables technicians to quickly locate potential fault points and improves the efficiency of fault troubleshooting.

[0109] Based on the fault detection conclusion heat map, the computer device further generates execution features for the multi-mode motherboard sensing signals. This process involves feature extraction and selection techniques, aiming to extract the key information from the original sensing signals that can best represent the fault characteristics and heat distribution patterns. The execution features not only need to include the type and degree of the fault, but also should reflect the evolution trend of the fault over time and the mutual influence between faults. Therefore, methods such as time series analysis and correlation analysis may be used in the feature generation stage to ensure that the extracted features have rich information content and good discrimination.

[0110] Specifically, the generation of execution features may include the following steps: First, decompose the sensing signal by time series to extract the trend term, seasonal term, and random term, which represent the long-term trend, periodic change, and random fluctuation of the signal respectively; Second, calculate the statistical features of the signal, such as mean, variance, skewness, kurtosis, etc., which can reflect the distribution characteristics and stability of the signal; Finally, use the information in the heat map to encode the fault risk level in the spatial position as part of the feature to form a multi-dimensional feature set of space-time-attribute.

[0111] Through the above process, the computer device finally obtains the fault detection execution features of the multi-mode motherboard sensing signal. These features not only contain rich fault information but also visually display the health status of the motherboard in the form of thermal distribution, providing strong data support for subsequent fault diagnosis and predictive maintenance. In practical applications, these execution features can be used to train a more accurate fault prediction model or to achieve real-time fault warning in an online monitoring system, significantly improving the intelligent level and operation and maintenance efficiency of motherboard maintenance.

[0112] In addition, in some alternative embodiments, to further enhance the effectiveness and robustness of the fault detection execution features, the integration of external data sources can also be considered. For example, by combining environmental factors (such as temperature and humidity), workload information, and historical maintenance records, the execution features can be enriched and refined. Such multi-dimensional data fusion helps to more comprehensively understand the context environment in which motherboard faults occur and improves the accuracy of fault detection and early warning capabilities.

[0113] In this way, the technical solution of generating execution features for the sensing signal based on the fault detection conclusion view of the multi-mode motherboard sensing reference signal is a complex process that integrates multiple fields such as signal processing, fault detection, thermal distribution analysis, and feature engineering. It not only requires the computer device to have powerful data processing capabilities but also relies on advanced algorithm models and in-depth industry knowledge to jointly improve the intelligent level and operation and maintenance efficiency of motherboard fault detection.

[0114] In an exemplary design idea, the extraction of time-frequency domain data from the multi-mode motherboard sensing signal to obtain the time-frequency domain sensing signal associated with the multi-mode motherboard sensing signal includes: extracting time-frequency domain data from the multi-mode motherboard sensing signal to determine the motherboard sensing detection information contained in the multi-mode motherboard sensing signal; determining the target time-frequency domain feature from a preset time-frequency domain feature set according to the motherboard sensing detection information contained in the multi-mode motherboard sensing signal; and performing feature decoding based on the target time-frequency domain feature to obtain the time-frequency domain sensing signal associated with the multi-mode motherboard sensing signal.

[0115] One of the key technical links in the embodiments of the present invention is to extract time-frequency domain data from these sensing signals to obtain the associated time-frequency domain sensing signals. This process not only involves in-depth analysis of the sensing signals but also requires selecting the most appropriate features from a preset time-frequency domain feature set according to the signal characteristics and performing feature decoding to finally generate time-frequency domain sensing signals that can accurately reflect the motherboard status. The following is a detailed introduction to this technical solution.

[0116] First, the computer device extracts time-frequency domain data from the collected multi-mode motherboard sensing signals. The purpose of this step is to determine the motherboard sensing detection information contained in the sensing signals. Time-frequency domain analysis is a powerful signal processing technique that can simultaneously display the characteristics of signals in both the time and frequency dimensions, and is very suitable for analyzing motherboard sensing signals with complex time-varying characteristics. When extracting time-frequency domain data, the computer device uses advanced algorithms such as the Short-Time Fourier Transform (STFT), wavelet transform, or Hilbert-Huang Transform (HHT) to convert the sensing signals from the original time domain to the time-frequency domain, thereby revealing the energy distribution and change patterns of the signals in different time periods and frequency bands.

[0117] Specifically, the time-frequency domain data extraction process may involve segmenting the sensing signals, with each segment corresponding to a specific time window. For the signals within each time window, the computer device calculates their amplitudes or power spectra at different frequencies to form a two-dimensional time-frequency distribution graph. This distribution graph can clearly display the energy intensity and changes of the signals in different time periods and frequency bands, providing rich information for subsequent feature selection and decoding.

[0118] Secondly, based on the motherboard sensing detection information contained in the multi-mode motherboard sensing signals, the computer device determines the target time-frequency domain features from a preset time-frequency domain feature set. The time-frequency domain feature set is a collection containing various possible features that can reflect the specific attributes of signals in different time-frequency domains. For example, some features may focus on the energy intensity of the signal in a specific frequency band, while others may focus on the transfer or change of signal energy between different time periods and frequency bands.

[0119] When determining the target time-frequency domain features, the computer device makes selections according to the specific content and requirements of the sensing detection information. For example, if the sensing signal reflects the temperature change of the motherboard, then time-frequency domain features related to temperature may be selected, such as the energy intensity or energy change rate in a specific frequency band. If voltage or current sensing signals are selected, then the stability and fluctuations of the signals in different time periods and frequency bands may be of concern.

[0120] To illustrate this process more specifically, the computer device is analyzing a sensing signal that reflects the temperature change of the motherboard. After extracting the time-frequency domain data, a two-dimensional time-frequency distribution graph is obtained, where the horizontal axis represents time, the vertical axis represents frequency, and the value at each point represents the energy intensity in the corresponding time period and frequency band. Based on this distribution graph, the computer device may select the following target time-frequency domain features:

[0121] The average energy intensity in the 0 - 10 Hz frequency band, which can reflect the low-frequency change situation of the motherboard temperature;

[0122] The energy change rate in the frequency band of 10 - 100 Hz, which can capture the mid - frequency fluctuations in the temperature signal;

[0123] The maximum energy value in the entire time - frequency distribution diagram and its corresponding time and frequency positions, which can indicate the sudden high - energy events in the temperature signal.

[0124] Finally, based on the selected target time - frequency domain features, the computer device will perform feature decoding to obtain the time - frequency domain sensing signal associated with the multi - mode mainboard sensing signal. Feature decoding is a process of converting time - frequency domain features into sensing signals that can be used for subsequent analysis or decision - making. In this step, the computer device will construct a new sensing signal according to the specific values and characteristics of the target time - frequency domain features, and this signal can more intuitively reflect the state and potential fault information of the mainboard.

[0125] Taking the target time - frequency domain features of the previously selected temperature sensing signal as an example, the computer device may perform the following decoding operations:

[0126] For the average energy intensity feature in the frequency band of 0 - 10 Hz, the computer device will calculate the average value of all energy intensities within this frequency band and use this average value as a component of the decoded sensing signal;

[0127] For the energy change rate feature in the frequency band of 10 - 100 Hz, the computer will calculate the change rate of the energy intensity within this frequency band, that is, the difference in energy intensity between adjacent time periods, and use this change rate as another component of the decoded sensing signal;

[0128] For the maximum energy value in the entire time - frequency distribution diagram and its corresponding time and frequency position features, the computer will use this maximum value and its corresponding time and frequency positions as additional information of the decoded sensing signal to indicate the sudden high - energy events in the temperature signal.

[0129] Through feature decoding, the computer device finally obtains a time - frequency domain sensing signal associated with the multi - mode mainboard sensing signal. This signal not only contains the main information of the original sensing signal but also presents the state and potential fault characteristics of the mainboard in a more intuitive and easier - to - analyze form. In practical applications, this time - frequency domain sensing signal can be used for subsequent fault diagnosis, performance evaluation, or predictive maintenance tasks, providing strong support for the intelligent management and operation and maintenance of computer devices.

[0130] It can be seen that extracting time-frequency domain data from the multi-mode motherboard sensing signal and obtaining the associated time-frequency domain sensing signal is a complex process involving multiple steps such as signal processing, feature selection, and feature decoding. This process requires the computer device to have powerful data processing capabilities and advanced algorithm support to ensure that the finally obtained time-frequency domain sensing signal can accurately and comprehensively reflect the status and potential fault information of the motherboard.

[0131] In the following example, determining the target time-frequency domain feature from the preset time-frequency domain feature set according to the motherboard sensing detection information included in the multi-mode motherboard sensing signal includes: determining the sensing signal group to which the multi-mode motherboard sensing signal belongs; obtaining the signal index relationship topology paired with the sensing signal group; the signal index relationship topology includes various index units for characterizing the multi-mode motherboard sensing signals belonging to the sensing signal group; in the signal index relationship topology, index unit pairing is performed according to the motherboard sensing detection information included in the multi-mode motherboard sensing signal to obtain the target time-frequency domain feature.

[0132] Another core technical link in the embodiments of the present invention is to determine the target time-frequency domain feature from the preset time-frequency domain feature set according to the motherboard sensing detection information included in the sensing signal. This process not only involves in-depth analysis of the sensing signal, but also requires selecting the most appropriate feature from the complex feature set according to the signal characteristics. To achieve this goal, the computer device will take a series of delicate operations, including determining the sensing signal group, obtaining the signal index relationship topology, and performing index unit pairing according to the sensing detection information.

[0133] First of all, the computer device needs to determine the sensing signal group to which the multi-mode motherboard sensing signal belongs. The sensing signal group is a classification of the sensing signal according to factors such as the signal source, type, and acquisition frequency. For the multi-mode motherboard sensing signal, different signals may reflect different aspects of the motherboard, such as temperature, voltage, current, etc. Therefore, grouping these signals according to their characteristics helps to better manage and analyze them.

[0134] When determining the sensing signal group, the computer device will identify and classify the sensing signal according to the preset rules and algorithms. These rules and algorithms may be based on various factors such as the physical characteristics of the signal, the acquisition method, and historical data. For example, for the temperature sensing signal, the computer device may divide it into groups reflecting different aspects such as CPU temperature, memory temperature, and graphics card temperature.

[0135] Secondly, the computer device needs to obtain the signal index relationship topology paired with the sensing signal group. The signal index relationship topology is a data structure used to characterize the relationship between sensing signals. It contains various index units of the multi-mode mainboard sensing signals belonging to the same sensing signal group, and these index units are connected through specific relationships to form a complex network.

[0136] In this signal index relationship topology, each index unit represents a specific aspect or feature of the sensing signal. For example, for the temperature sensing signal group, the index units may include temperature characteristics reflecting different temperature regions, different time periods, and different workloads. These index units are connected through relationships to form a comprehensive description of the temperature sensing signal.

[0137] To illustrate the construction and application of the signal index relationship topology more specifically, an example can be given. The computer device is analyzing a multi-mode sensing signal group reflecting the mainboard temperature. In this group, there are sensing signals respectively reflecting the CPU temperature, memory temperature, and graphics card temperature. To comprehensively describe the relationship and characteristics between these signals, the computer device will construct a signal index relationship topology.

[0138] In this topology, each index unit represents a specific temperature characteristic, such as the idle temperature of the CPU, the high-load temperature of the memory, the gaming-load temperature of the graphics card, etc. These index units are connected through relationships to form a comprehensive description of the mainboard temperature. For example, there may be a certain correlation between the idle temperature of the CPU and the high-load temperature of the memory. Because when the CPU is idle, the memory may take on more work tasks, resulting in a temperature increase. This correlation is clearly reflected in the signal index relationship topology.

[0139] Then, in the signal index relationship topology, the computer device will pair the index units based on the mainboard sensing detection information contained in the multi-mode mainboard sensing signals. This process is a key step in determining the target time-frequency domain characteristics. The computer device will search for the matching index units in the signal index relationship topology according to the specific content and requirements of the sensing detection information and perform pairing.

[0140] The pairing process may involve various algorithms and strategies. For example, the computer device may adopt a similarity-based pairing algorithm and pair according to the similarity degree between the sensing detection information and the index units. The similarity may be calculated based on various factors such as the waveform, frequency, and amplitude of the signal. In addition, the computer device may also adopt a rule-based pairing strategy and pair the index units according to the preset rules and conditions.

[0141] During the pairing process, the computer device comprehensively considers multiple aspects and characteristics of the sensing detection information to ensure the accuracy and effectiveness of the pairing. For example, for the temperature sensing signal, the computer device may consider multiple factors such as the amplitude, frequency, and change trend of the signal for pairing. If the sensing detection information indicates an abnormal upward trend in the motherboard temperature, then the computer device may search for index units in the signal index relationship topology that reflect a similar trend for pairing.

[0142] Finally, through the process of index unit pairing, the computer device can obtain the target time-frequency domain characteristics. These characteristics are obtained through careful selection and pairing in the signal index relationship topology based on the motherboard sensing detection information contained in the multi-mode motherboard sensing signal. They can accurately reflect the state and potential fault information of the motherboard, providing strong support for subsequent tasks such as fault diagnosis, performance evaluation, or predictive maintenance.

[0143] Thus, determining the target time-frequency domain characteristics from the preset time-frequency domain characteristic set based on the motherboard sensing detection information contained in the multi-mode motherboard sensing signal is a complex process involving multiple steps such as signal grouping, obtaining the signal index relationship topology, and index unit pairing. This process requires the computer device to have powerful data processing capabilities and advanced algorithm support to ensure that the finally obtained target time-frequency domain characteristics can accurately and comprehensively reflect the state and potential fault information of the motherboard. Through the implementation of this technical solution, the computer device can better manage and analyze the multi-mode motherboard sensing signal, providing strong support for intelligent management and operation and maintenance.

[0144] Under some other design ideas, the method further includes: obtaining the past time-frequency domain sensing signals associated with the past multi-mode motherboard sensing signals; the past time-frequency domain sensing signals are obtained by performing time-frequency domain data extraction on the past multi-mode motherboard sensing signals; performing index-based clustering on each of the past time-frequency domain sensing signals to obtain at least one sensing signal index formed by the clustering; and obtaining the past sensing state trend vector of the past multi-mode motherboard sensing signals included in each of the sensing signal indexes based on the sensing state trend vectors of the sensing signal indexes.

[0145] In the task of analyzing the multi-mode motherboard sensing signal performed by the computer device, in addition to processing the current sensing signal in real time, it is often necessary to consider the past sensing signal data in order to more comprehensively understand the state and trend of the motherboard. Therefore, under some design ideas, the method further includes the processing and analysis of the past multi-mode motherboard sensing signals.

[0146] First, the computer device needs to obtain the past time-frequency domain sensing signals associated with the past multi-mode motherboard sensing signals. These past time-frequency domain sensing signals are obtained by performing time-frequency domain data extraction on the past multi-mode motherboard sensing signals. Time-frequency domain data extraction is a powerful signal processing technique that can transform the sensing signal from the original time domain to the time-frequency domain, thereby revealing the energy distribution and change patterns of the signal in different time periods and frequency bands. In this way, the computer device can obtain the past time-frequency domain sensing signals containing rich information, providing a basis for subsequent analysis and processing.

[0147] Secondly, the computer device performs index-based clustering on each past time-frequency domain sensing signal to obtain at least one sensing signal index formed by the clustering. The purpose of this step is to organize and classify the past sensing signal data for better management and analysis. The sensing signal index is a data structure used to characterize the relationship between sensing signals. It can group sensing signals with similar characteristics or attributes into one category and form an index unit. Through the index-based clustering operation, the computer device can classify the past time-frequency domain sensing signals according to their characteristics and attributes and form several sensing signal indexes.

[0148] When performing index-based clustering, the computer device may adopt various algorithms and strategies. For example, it may use a clustering algorithm to group sensing signals with similar time-frequency domain characteristics into the same category. The clustering algorithm can calculate based on the characteristics such as the amplitude, frequency, and phase of the sensing signal and divide signals with similar characteristics into the same cluster. In this way, each cluster corresponds to a sensing signal index, which contains the past multi-mode motherboard sensing signals with similar time-frequency domain characteristics.

[0149] Then, the computer device obtains the past sensing state trend vector of the past multi-mode motherboard sensing signals included in each sensing signal index based on the sensing state trend vector of each sensing signal index. The sensing state trend vector is a vector used to describe the change trend of the sensing signal state, which can reflect the change situation and trend of the signal in different time periods. By calculating the sensing state trend vector of each sensing signal index, the computer device can obtain a comprehensive description of the change trend of the past multi-mode motherboard sensing signals within the index.

[0150] To illustrate this process more specifically, an example can be given. The computer device is analyzing a multi-mode sensing signal reflecting the motherboard temperature. It first obtains the past multi-mode motherboard temperature sensing signals and performs time-frequency domain data extraction to obtain the past time-frequency domain temperature sensing signals. Then, it uses a clustering algorithm to perform index-based clustering on these past time-frequency domain temperature sensing signals to form several temperature sensing signal indexes. Each index contains the past temperature sensing signals with similar time-frequency domain characteristics.

[0151] Next, the computer device calculated the sensing state trend vectors for each temperature sensing signal index. For each index, it calculated a vector that describes the change trend of the temperature sensing signals within that index over different time periods. For example, for a certain index, its sensing state trend vector may indicate that the temperature of the motherboard has shown a gradually increasing trend over a past period of time. Such information is very important for predicting the future state of the motherboard, detecting potential faults, and formulating maintenance strategies.

[0152] Through this embodiment, the computer device can more comprehensively understand the past state and trends of the motherboard, providing strong support for subsequent tasks such as fault diagnosis, performance evaluation, or predictive maintenance. At the same time, by processing and analyzing the past multi-mode motherboard sensing signals, the computer device can also discover some potential patterns and regularities, providing useful references for optimizing the motherboard design and improving system performance. In addition, this embodiment also has good scalability and flexibility. In practical applications, the computer device can select different time-frequency domain data extraction algorithms, clustering algorithms, and calculation methods for the sensing state trend vectors according to specific requirements and scenarios. At the same time, it can also flexibly adjust the grouping of sensing signals and the construction method of indexes according to the actual situation and monitoring requirements of the motherboard to achieve more accurate and effective monitoring and analysis of the motherboard state.

[0153] Thus, obtaining the past time-frequency domain sensing signals associated with the past multi-mode motherboard sensing signals and performing index-based clustering and calculation of the sensing state trend vectors is a very effective method for monitoring and analyzing the motherboard state. Through this embodiment, the computer device can more comprehensively understand the past state and trends of the motherboard, providing strong support for subsequent tasks such as fault diagnosis, performance evaluation, or predictive maintenance.

[0154] In some other possible embodiments, the method further includes: when the fault detection and analysis of the multi-mode motherboard sensing signals are completed, obtaining a fault detection and analysis report of the multi-mode motherboard sensing signals; the fault detection and analysis report includes a fault location area, a fault location category, and a fault element analysis text; based on the fault location area, the fault location category, and the target sensing state trend vector, obtaining a fault detection relationship vector of the multi-mode motherboard sensing signals; based on the fault detection relationship vector and the fault element analysis text, constructing a motherboard defect prediction vector of the multi-mode motherboard sensing signals; and pairing the motherboard defect prediction vector with the past motherboard defect prediction vector of the past multi-mode motherboard sensing signals to obtain a defect prediction vector pairing view for the multi-mode motherboard sensing signals.

[0155] It is worth mentioning that in the multi-mode motherboard sensing signal analysis task performed by a computer device, in addition to processing the current sensing signal in real time and performing fault detection and analysis, it is often necessary to consider the past sensing signal data in order to more comprehensively understand the state, trend, and potential defects of the motherboard. Therefore, under some design concepts, the method not only includes the processing of the current multi-mode motherboard sensing signal, but also covers the analysis and utilization of the past sensing signals.

[0156] First, the computer device acquires the past time-frequency domain sensing signals associated with the past multi-mode motherboard sensing signals. These signals are obtained by extracting time-frequency domain data from the past multi-mode motherboard sensing signals, and they contain the energy distribution and change patterns of the sensing signals in different time periods and frequency bands. For example, for the temperature sensing signal, the past time-frequency domain sensing signals may show the change trend of the motherboard temperature in one year, and the temperature records of a specific temperature sensing signal in the past year (in months) are as follows: [26.5, 27.1, 28.2, 29.5, 30.7, 31.8, 32.4, 32.1, 31.2, 30.4, 29.1, 28.0].

[0157] Second, the computer device performs index-based clustering on each of the past time-frequency domain sensing signals to form at least one sensing signal index. Each sensing signal index contains the past multi-mode motherboard sensing signals with similar time-frequency domain characteristics. For example, a temperature sensing signal index may contain all the past time-frequency domain sensing signals related to the motherboard temperature, and the temperature trend vector under its index can be an average vector representing the average temperature of each month: [28.25, 28.35, 29.05, 29.85, 30.5, 31.55, 32.2, 32.05, 31.15, 30.35, 29.05, 27.95].

[0158] Then, based on the sensing state trend vectors of each sensing signal index, the computer device obtains the past sensing state trend vectors of the past multi-mode motherboard sensing signals included in each sensing signal index. This vector describes the change situation and trend of the sensing signal in different time periods. Continuing with the example of the temperature sensing signal index, its sensing state trend vector may be a difference vector representing the change amount of the temperature of each month: [0.1, 0.7, 0.8, 0.65, 0.95, 0.7, -0.15, -0.95, -0.9, -0.8, -1.3, -1.1].

[0159] In a further embodiment, after the multi-mode motherboard sensing signal completes the fault detection and analysis, the computer device will obtain its fault detection and analysis report. This report includes important information such as the fault location area, fault location category, and fault element analysis text. For example, the report may indicate that the fault location area is in the power module of the motherboard, the fault location category is voltage anomaly, and provides a detailed fault element analysis text.

[0160] Based on this fault detection and analysis report, the computer device will further perform a series of processing and analysis. It will combine the fault location area and fault location category in the report with the previously calculated target sensing state trend vector (i.e., the sensing state trend vector of the current multi-mode motherboard sensing signal, which is also a differential vector: [0.2, 0.5, 0.7, 0.4, 0.3, -0.2, -0.5, -0.8, -0.6, -0.4, -0.3, 0.1]) to obtain the fault detection relationship vector of the multi-mode motherboard sensing signal. This vector can reflect the correlation between the fault and the sensing state trend. Taking the voltage anomaly in the power module as an example, the fault detection relationship vector may be a correlation vector representing the correlation between the monthly temperature difference and the voltage anomaly: [0.8, 0.75, 0.9, 0.6, 0.55, -0.7, -0.65, -0.85, -0.7, -0.5, -0.45, 0.2].

[0161] Next, the computer device will construct the motherboard defect prediction vector of the multi-mode motherboard sensing signal based on the fault detection relationship vector and the fault element analysis text. This vector is a quantitative representation of the potential defects of the current multi-mode motherboard sensing signal. Taking the voltage anomaly in the power module as an example, the motherboard defect prediction vector may be a vector combining the correlation and the fault element weight: [0.8×0.5, 0.75×0.5, 0.9×0.5, 0.6×0.5, 0.55×0.5, -0.7×0.3, -0.65×0.3, -0.85×0.3, -0.7×0.3, -0.5×0.3, -0.45×0.3, 0.2×0.1], that is, [0.4, 0.375, 0.45, 0.3, 0.275, -0.21, -0.195, -0.255, -0.21, -0.15, -0.135, 0.02].

[0162] Finally, the computer device will pair the motherboard defect prediction vector with the past motherboard defect prediction vectors of past multi-mode motherboard sensing signals to obtain a defect prediction vector pairing view for the multi-mode motherboard sensing signals. This pairing view is a comprehensive evaluation of the similarities and differences in defects between the current multi-mode motherboard sensing signals and past sensing signals. Taking the abnormal voltage of the power module as an example, if the past motherboard defect prediction vectors also contain similar temperature and voltage change trends, then the pairing view may indicate that the power module of the current motherboard has similar defect situations as in the past and needs to be repaired and maintained as soon as possible.

[0163] In other alternative embodiments, the method further includes: in response to a motherboard production line evaluation task, determining an evaluation task keyword and production line finished product status output data; retrieving target multi-mode motherboard sensing signals based on the evaluation task keyword and production line finished product status output data; generating a target fault detection report for evaluation based on the evaluation task keyword according to the target multi-mode motherboard sensing signals, and performing a production line anomaly evaluation on the target fault detection report based on the production line finished product status output data.

[0164] In another alternative embodiment, the method further includes: when the fault detection execution feature characterization meets the fault detection index, obtaining a fault detection reference report of the multi-mode motherboard sensing reference signal; determining a key fault detection reference event of the multi-mode motherboard sensing reference signal from the fault detection reference report; generating a fault detection attention feature for the multi-mode motherboard sensing signals according to the key fault detection reference event; the fault detection attention feature includes a local feature recognition decision vector for fault detection of the multi-mode motherboard sensing signals.

[0165] It can be understood that in addition to real-time processing of current sensing signals and performing fault detection analysis, it is often necessary to consider past sensing signal data and production line evaluation tasks in order to more comprehensively understand the status, trends, and potential defects of the motherboard, and further perform anomaly evaluation on the production line finished products. In addition, during the fault detection process, the fault detection report of the reference signal is also used to generate a fault detection attention feature to improve the accuracy of fault detection.

[0166] First, in response to a motherboard production line evaluation task, the computer device needs to determine an evaluation task keyword and production line finished product status output data. The evaluation task keyword is a word or phrase used to describe the specific requirements or goals of the production line evaluation task, such as "temperature stability", "voltage fluctuation", etc. The production line finished product status output data is the actual status data of the motherboard finished products on the production line, including real-time readings of various sensing signals, finished product test results, etc.

[0167] Secondly, based on the evaluation task keywords and the production line finished product status output data, the computer device retrieves the target multi-mode motherboard sensing signals. This step is achieved by searching for signals related to the evaluation task keywords in the stored historical sensing signal data. For example, if the evaluation task keyword is "temperature stability", then the computer device will retrieve all temperature-related sensing signals and filter out those signals that exhibit stable or unstable characteristics over time as the target multi-mode motherboard sensing signals.

[0168] After retrieving the target multi-mode motherboard sensing signals, the computer device generates a target fault detection report for evaluation based on the evaluation task keywords. This report will analyze in detail the characteristics of the target multi-mode motherboard sensing signals and determine whether there are faults or anomalies related to the evaluation task keywords. For example, if the evaluation task keyword is "voltage fluctuation", then the fault detection report will analyze whether there are abnormal fluctuations in the voltage sensing signals and give specific parameters such as the fluctuation amplitude and frequency.

[0169] Meanwhile, the computer device also conducts a production line anomaly assessment for the target fault detection report based on the production line finished product status output data. This step is to compare the analysis in the fault detection report with the production line finished product status output data to determine whether there are anomalies or defects in the motherboard finished products on the production line that are consistent with the fault detection report. For example, if the fault detection report indicates an abnormal voltage fluctuation, and the production line finished product status output data also shows that there are significant differences in the voltage readings of the motherboard finished products in the same batch, then the computer device will determine that there are voltage-related anomalies or defects on the production line.

[0170] In an alternative optional embodiment, when the fault detection execution feature characterization meets the fault detection criteria, the computer device obtains a fault detection reference report for the multi-mode motherboard sensing reference signals. This report is the result of fault detection on the sensing reference signals based on historical data or preset standards, and it contains information such as whether there are faults in the sensing reference signals, the types and degrees of the faults.

[0171] From the fault detection reference report, the computer device determines the key fault detection reference events for the multi-mode motherboard sensing reference signals. These events are the most relevant or representative events related to fault detection in the report, such as the time points of fault occurrence, the transitions of fault types, etc. By identifying these key events, the computer device can better understand the behavioral characteristics of the sensing reference signals in the fault state.

[0172] Finally, based on the key fault detection reference events, the computer device generates fault detection attention features for the multi-mode motherboard sensing signals. This feature includes local feature recognition decision vectors that should be focused on during fault detection. For example, if the key fault detection reference event indicates that there has been a significant change in the frequency components of the sensing signal within a specific time period, then the fault detection attention feature will include the frequency feature recognition vector for this time period and assign higher weights or attentions during subsequent fault detection processes.

[0173] The following is a specific numerical example to further illustrate the technical solution of the above embodiment: The evaluation task keyword is "temperature stability", and the output data of the production line finished product status includes the real-time temperature readings of 100 motherboard finished products. The computer device first retrieves the target multi-mode motherboard sensing signals related to temperature and generates a target fault detection report for evaluation based on temperature stability. The report indicates that within the past month, the temperature sensing signals of 5 motherboards have shown abnormal fluctuations, and the fluctuation amplitude exceeds the preset threshold (e.g., ±2°C).

[0174] Next, the computer device further evaluates the abnormality of these 5 motherboards based on the output data of the production line finished product status. By comparing the temperature readings of other motherboards in the same batch, it is found that there are indeed obvious differences in the temperatures of these 5 motherboards, and this difference is consistent with the fluctuation situation indicated in the fault detection report. Therefore, the computer device determines that there are abnormalities or defects related to temperature stability on the production line.

[0175] In another embodiment, when performing fault detection on a new motherboard, the computer device first obtains the fault detection execution feature representation of its multi-mode motherboard sensing signals and determines whether it meets the fault detection criteria. If it meets, the computer device obtains the fault detection reference report of the motherboard sensing reference signal. The report indicates that within a past period of time, the temperature sensing signal of this motherboard has experienced a significant change in frequency components, which is a key fault detection reference event.

[0176] Based on this key event, the computer device generates fault detection attention features for the motherboard sensing signals. This feature includes local feature recognition decision vectors that should be focused on during fault detection, namely the frequency feature recognition vector of the temperature sensing signal. During subsequent fault detection processes, the computer device will assign higher weights or attentions to this vector to more accurately identify potential faults or abnormalities.

[0177] Thus, through the technical solution that combines the production line evaluation task and the fault detection attention feature, the computer device can more comprehensively and accurately evaluate the state and potential defects of the motherboard, and provide strong support for subsequent tasks such as fault diagnosis, performance evaluation, or predictive maintenance. This not only improves the accuracy of motherboard state monitoring and analysis, but also helps to detect and handle abnormalities or defects on the production line in a timely manner, thereby ensuring product quality and production efficiency.

[0178] In some independent embodiments, after generating the fault detection attention feature for the multimode motherboard sensing signal according to the key fault detection reference event, the method further includes: using the local feature recognition decision vector included in the fault detection attention feature to perform local fault detection on the multimode motherboard sensing signal to obtain the local fault detection view of the multimode motherboard sensing signal.

[0179] Specifically, the using the local feature recognition decision vector included in the fault detection attention feature to perform local fault detection on the multimode motherboard sensing signal to obtain the local fault detection view of the multimode motherboard sensing signal includes: using the local feature recognition decision vector to extract the sensing signal change data sequence of the multimode motherboard sensing signal, where the sensing signal change data sequence includes consecutive X sensing signal change data, and X is an integer greater than or equal to 1; obtaining the abnormal signal change data sequence according to the sensing signal change data sequence, where the abnormal signal change data sequence includes consecutive X abnormal signal change data; based on the sensing signal change data sequence, obtaining the sensing signal dense hidden danger vector relationship network sequence through the first dense hidden danger feature recognition branch included in the local fault detection algorithm, where the sensing signal dense hidden danger vector relationship network sequence includes X sensing signal dense hidden danger vector relationship networks; based on the abnormal signal change data sequence, obtaining the abnormal signal dense hidden danger vector relationship network sequence through the second dense hidden danger feature recognition branch included in the local fault detection algorithm, where the abnormal signal dense hidden danger vector relationship network sequence includes X abnormal signal dense hidden danger vector relationship networks; based on the sensing signal dense hidden danger vector relationship network sequence and the abnormal signal dense hidden danger vector relationship network sequence, obtaining the fault voting weight corresponding to the sensing signal change data sequence through the detection and discrimination branch included in the local fault detection algorithm; determining the local fault detection view of the sensing signal change data sequence according to the fault voting weight.

[0180] In some other independent embodiments, based on the sequence of the dense hidden danger vector relation networks of the sensing signals and the sequence of the dense hidden danger vector relation networks of the abnormal signals, the fault voting weights corresponding to the sensing signal change data sequence are obtained through the detection and discrimination branch included in the local fault detection algorithm, including: based on the sequence of the dense hidden danger vector relation networks of the sensing signals, X first one-hot encodings are obtained through the first long short-term memory model included in the local fault detection algorithm, where each first one-hot encoding corresponds to a dense hidden danger vector relation network of a sensing signal; based on the sequence of the dense hidden danger vector relation networks of the abnormal signals, X second one-hot encodings are obtained through the second long short-term memory model included in the local fault detection algorithm, where each second one-hot encoding corresponds to a dense hidden danger vector relation network of an abnormal signal; the X first one-hot encodings and the X second one-hot encodings are concatenated to obtain X target one-hot encodings, where each target one-hot encoding includes a first one-hot encoding and a second one-hot encoding; based on the X target one-hot encodings, the fault voting weights corresponding to the sensing signal change data sequence are obtained through the detection and discrimination branch included in the local fault detection algorithm.

[0181] In some other independent embodiments, the obtaining of X first one-hot encodings through the first long short-term memory model included in the local fault detection algorithm based on the sequence of the dense hidden danger vector relation networks of the sensing signals includes: for each dense hidden danger vector relation network of the sensing signals in the sequence of the dense hidden danger vector relation networks of the sensing signals, a first local downsampling vector relation network is obtained through the local downsampling layer included in the first long short-term memory model, where the first long short-term memory model belongs to the local fault detection algorithm; for each dense hidden danger vector relation network of the sensing signals in the sequence of the dense hidden danger vector relation networks of the sensing signals, a first global downsampling vector relation network is obtained through the global downsampling layer included in the first long short-term memory model; for each dense hidden danger vector relation network of the sensing signals in the sequence of the dense hidden danger vector relation networks of the sensing signals, based on the first local downsampling vector relation network and the first global downsampling vector relation network, a first dense hidden danger residual connection vector is obtained through the residual connection layer included in the first long short-term memory model; for each dense hidden danger vector relation network of the sensing signals in the sequence of the dense hidden danger vector relation networks of the sensing signals, based on the first dense hidden danger residual connection vector and the dense hidden danger vector relation network of the sensing signal, a first one-hot encoding is obtained through the first global downsampling layer included in the first long short-term memory model.

[0182] In the multi - mode motherboard sensing signal analysis task performed by a computer device, in addition to real - time processing of current sensing signals and performing fault detection and analysis, fault detection attention features can also be utilized to improve the accuracy of fault detection. Subsequently, details will be provided on how to use fault detection attention features for local fault detection and obtain the local fault detection viewpoints of multi - mode motherboard sensing signals.

[0183] First, after generating fault detection attention features for multi - mode motherboard sensing signals based on key fault detection reference events, the computer device uses the local feature recognition decision vectors contained in these features to perform local fault detection on the multi - mode motherboard sensing signals. This step is achieved by extracting the sensing signal change data sequence, where the sensing signal change data sequence includes consecutive X sensing signal change data, and X is an integer greater than or equal to 1. These data sequences can reflect the change trend and characteristics of the sensing signal over time.

[0184] Next, the computer device obtains the abnormal signal change data sequence based on the sensing signal change data sequence. The abnormal signal change data sequence also includes consecutive X abnormal signal change data, which are the parts of the sensing signal change data sequence that do not conform to normal behavior or expected behavior and may indicate potential faults or abnormalities.

[0185] Then, based on the sensing signal change data sequence, the computer device obtains the sensing signal dense hidden - danger vector relationship network sequence through the first dense hidden - danger feature recognition branch of the local fault detection algorithm. This sequence includes X sensing signal dense hidden - danger vector relationship networks, and each network reflects the dense hidden - danger features of the sensing signal within a certain time period, that is, the features closely related to faults or abnormalities.

[0186] Meanwhile, based on the abnormal signal change data sequence, the computer device obtains the abnormal signal dense hidden - danger vector relationship network sequence through the second dense hidden - danger feature recognition branch of the local fault detection algorithm. This sequence also includes X abnormal signal dense hidden - danger vector relationship networks, and each network focuses on the dense hidden - danger features of the abnormal signal within a certain time period.

[0187] Next, based on the sensing signal dense hidden - danger vector relationship network sequence and the abnormal signal dense hidden - danger vector relationship network sequence, the computer device obtains the fault voting weight corresponding to the sensing signal change data sequence through the detection and discrimination branch of the local fault detection algorithm. This weight is an important indicator for measuring the degree of correlation between the sensing signal change data sequence and faults or abnormalities.

[0188] Finally, the computer device determines the local fault detection view of the sensed signal change data sequence according to the fault voting weight. If the fault voting weight is high, it indicates that the sensed signal change data sequence is closely related to faults or anomalies. Therefore, the local fault detection view will tend to consider that there are faults or anomalies. On the contrary, if the fault voting weight is low, it indicates that the sensed signal change data sequence conforms to normal behavior or expected behavior. Therefore, the local fault detection view will tend to consider that there are no faults or anomalies.

[0189] To illustrate this process more specifically, let X be 5, that is, both the sensed signal change data sequence and the abnormal signal change data sequence include 5 consecutive data points. The computer device will first extract these 5 data points as the sensed signal change data sequence and obtain the abnormal signal change data sequence based on these data points. Then, through the two dense hidden hazard feature recognition branches of the local fault detection algorithm, the sensed signal dense hidden hazard vector relation network sequence and the abnormal signal dense hidden hazard vector relation network sequence are respectively obtained, and each sequence includes 5 vector relation networks. Next, the computer device obtains the fault voting weight based on these two sequences through the detection and discrimination branch and determines the local fault detection view according to the weight.

[0190] In some other independently implementable embodiments, when the computer device obtains the fault voting weight corresponding to the sensed signal change data sequence based on the sensed signal dense hidden hazard vector relation network sequence and the abnormal signal dense hidden hazard vector relation network sequence, it will use a long short-term memory model (LSTM) for further processing. Specifically, the computer device will obtain 5 first one-hot encodings based on the sensed signal dense hidden hazard vector relation network sequence through the first long short-term memory model, and each one-hot encoding corresponds to a sensed signal dense hidden hazard vector relation network. At the same time, 5 second one-hot encodings are obtained based on the abnormal signal dense hidden hazard vector relation network sequence through the second long short-term memory model, and each one-hot encoding corresponds to an abnormal signal dense hidden hazard vector relation network. Then, the computer device will splice these 5 first one-hot encodings and 5 second one-hot encodings to obtain 5 target one-hot encodings, and each target one-hot encoding includes a first one-hot encoding and a second one-hot encoding. Finally, based on these 5 target one-hot encodings, the fault voting weight corresponding to the sensed signal change data sequence is obtained through the detection and discrimination branch.

[0191] In some other independent embodiments, when the computer device obtains the first one-hot encoding through the first long short-term memory model based on the sequence of dense hidden danger vector relation networks of sensing signals, a series of downsampling and residual connection operations will be performed. Specifically, for each dense hidden danger vector relation network of sensing signals in the sequence of dense hidden danger vector relation networks of sensing signals, the computer device will obtain the first local downsampling vector relation network through the local downsampling layer included in the first long short-term memory model; at the same time, obtain the first global downsampling vector relation network through the global downsampling layer. Then, based on the first local downsampling vector relation network and the first global downsampling vector relation network, obtain the first dense hidden danger residual connection vector through the residual connection layer. Finally, based on the first dense hidden danger residual connection vector and the dense hidden danger vector relation network of sensing signals, obtain the first one-hot encoding through the first global downsampling layer. This process can help the computer device extract the dense hidden danger features in the sensing signals more accurately and further improve the accuracy of fault detection.

[0192] Furthermore, Figure 2 FIG. 5 is a schematic structural diagram of a computer device 200 provided by an embodiment of the present invention. As Figure 2 shown, the computer device 200 includes a processor 210. The processor 210 can call and run a computer program from a memory to implement the method in the embodiment of the present invention. Optionally, as Figure 2 shown, the computer device 200 may further include a memory 230. Among them, the processor 210 can call and run a computer program from the memory 230 to implement the method in the embodiment of the present invention. Among them, the memory 230 may be a separate device independent of the processor 210 or integrated in the processor 210. Optionally, as Figure 2 shown, the computer device 200 may further include a transceiver 220. The processor 210 can control the transceiver 220 to interact with other devices. Specifically, it can send information or data to other devices or receive information or data sent by other devices. Optionally, the computer device 200 can implement the corresponding processes of the storage engine or components in the storage engine (such as a processing module) or the device deployed with the storage engine in each method of the embodiment of the present invention. For the sake of brevity, it will not be elaborated here. It should be understood that the processor in the embodiment of the present invention may be an integrated circuit chip with signal processing capabilities. It can be understood that the memory in the embodiment of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. It should be noted that the memory of the system and method described in this article is intended to include but not limited to suitable types of memories.

[0193] Based on the above, a readable storage medium is provided. A program or instruction is stored on the readable storage medium, and when the program or instruction is executed by a processor, the steps of the above method are implemented.

[0194] The embodiments of the present invention have been described above in conjunction with the accompanying drawings. However, the present invention is not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make many forms without departing from the spirit and scope of the present invention, and all of them fall within the protection scope of the present invention.

Claims

1. A motherboard fault detection method, characterized in that Applied to a computer device, the method includes: Obtain multi-mode motherboard sensing signals to be subject to fault detection, extract time-frequency domain data from the multi-mode motherboard sensing signals, and obtain time-frequency domain sensing signals associated with the multi-mode motherboard sensing signals; Determine the target sensing state trend vector of the multi-mode motherboard sensing signals according to the time-frequency domain sensing signals; Perform sensing state trend vector pairing between the target sensing state trend vector and the past sensing state trend vectors of past multi-mode motherboard sensing signals that have completed fault detection to obtain a trend vector pairing view; the past sensing state trend vectors are determined according to the past time-frequency domain sensing signals associated with the past multi-mode motherboard sensing signals; Determine the multi-mode motherboard sensing reference signals of the multi-mode motherboard sensing signals from the past multi-mode motherboard sensing signals according to the trend vector pairing view; Based on the fault detection conclusion view of the multi-mode motherboard sensing reference signals, generate execution features for the multi-mode motherboard sensing signals to obtain the fault detection execution features of the multi-mode motherboard sensing signals; the fault detection execution features are used to indicate the fault detection analysis for the multi-mode motherboard sensing signals; The method further includes: in response to a motherboard production line evaluation task, determine evaluation task keywords and production line finished product status output data; based on the evaluation task keywords and the production line finished product status output data, retrieve target multi-mode motherboard sensing signals; generate a target fault detection report for evaluation based on the evaluation task keywords according to the target multi-mode motherboard sensing signals, and perform production line anomaly evaluation on the target fault detection report based on the production line finished product status output data; The method further includes: when the fault detection execution features characterize compliance with fault detection indicators, obtain the fault detection reference report of the multi-mode motherboard sensing reference signals; from the fault detection reference report, determine the key fault detection reference events of the multi-mode motherboard sensing reference signals; according to the key fault detection reference events, generate fault detection attention features for the multi-mode motherboard sensing signals; the fault detection attention features include local feature recognition decision vectors for performing fault detection on the multi-mode motherboard sensing signals; use the local feature recognition decision vectors included in the fault detection attention features to perform local fault detection on the multi-mode motherboard sensing signals to obtain the local fault detection view of the multi-mode motherboard sensing signals.

2. The method according to claim 1, wherein The determining the target sensing state trend vector of the multi-mode motherboard sensing signals according to the time-frequency domain sensing signals includes: Pair the time-frequency domain sensing signals with each sensing signal index respectively to obtain a signal index pairing result; the sensing signal index is obtained by index-based clustering of the past time-frequency domain sensing signals associated with the past multi-mode motherboard sensing signals; Determine the sensing signal index in the signal index pairing result indicating completion of pairing as the target sensing signal index to which the time-frequency domain sensing signals belong; Based on the sensing state deduction features of the target sensing signal index, obtain the target sensing state trend vector of the multi-mode motherboard sensing signals; Pairing the time-frequency domain sensing signal with each sensing signal index respectively to obtain a signal index pairing result includes: respectively determining a difference discrimination variable between the time-frequency domain sensing signal and each sensing signal index; Obtaining a signal index pairing result based on the difference discrimination variable; Determining the sensing signal index in the signal index pairing result representing the completion of pairing as the target sensing signal index to which the time-frequency domain sensing signal belongs includes: determining the sensing signal index in the signal index pairing result whose included difference discrimination variable is not greater than the discrimination variable threshold as the target sensing signal index to which the time-frequency domain sensing signal belongs.

3. The method according to claim 1, wherein Pairing the target sensing state trend vector with the past sensing state trend vectors of the past multi-mode main board sensing signals for which fault detection has been completed to obtain a trend vector pairing view includes: Obtaining the past sensing state trend vectors of the past multi-mode main board sensing signals for which fault detection has been completed; Determining the Minkowski similarity between the target sensing state trend vector and the past sensing state trend vectors; Obtaining a trend vector pairing view based on the Minkowski similarity.

4. The method according to claim 1, wherein The trend vector pairing view includes the Minkowski similarity; according to the trend vector pairing view, determining the multi-mode main board sensing reference signal of the multi-mode main board sensing signal from the past multi-mode main board sensing signals includes: Obtaining a preset similarity threshold; Determining, from the trend vector pairing view, the target trend vector pairing view whose included Minkowski similarity is higher than the preset similarity threshold; Determining the past multi-mode main board sensing signal associated with the target trend vector pairing view as the multi-mode main board sensing reference signal of the multi-mode main board sensing signal.

5. The method according to claim 1, wherein Performing execution feature generation on the multi-mode main board sensing signal based on the fault detection conclusion view of the multi-mode main board sensing reference signal to obtain the fault detection execution feature of the multi-mode main board sensing signal includes: Obtaining the fault detection conclusion view of the multi-mode main board sensing reference signal; Performing global analysis on the fault detection conclusion view to obtain a fault detection conclusion heat distribution; Performing execution feature generation on the multi-mode main board sensing signal according to the fault detection conclusion heat distribution to obtain the fault detection execution feature of the multi-mode main board sensing signal.

6. The method according to claim 1, wherein Performing time-frequency domain data extraction on the multi-mode main board sensing signal to obtain the time-frequency domain sensing signal associated with the multi-mode main board sensing signal includes: Performing time-frequency domain data extraction on the multi-mode main board sensing signal to determine the main board sensing detection information included in the multi-mode main board sensing signal; Determining a target time-frequency domain feature from a preset time-frequency domain feature set according to the main board sensing detection information included in the multi-mode main board sensing signal; Performing feature decoding based on the target time-frequency domain feature to obtain the time-frequency domain sensing signal associated with the multi-mode main board sensing signal; Determining a target time-frequency domain feature from a preset time-frequency domain feature set according to the main board sensing detection information included in the multi-mode main board sensing signal includes: Determine the sensing signal group to which the multi-mode main board sensing signal belongs; Obtain the signal index relationship topology paired with the sensing signal group; various index units for characterizing the multi-mode main board sensing signals belonging to the sensing signal group are included in the signal index relationship topology; In the signal index relationship topology, perform index unit pairing according to the main board sensing detection information included in the multi-mode main board sensing signal to obtain the target time-frequency domain feature.

7. The method according to any one of claims 1 to 6, characterized in that The method further includes: Obtain the past time-frequency domain sensing signals associated with the past multi-mode main board sensing signals; the past time-frequency domain sensing signals are obtained by performing time-frequency domain data extraction on the past multi-mode main board sensing signals; Perform index-based clustering on each of the past time-frequency domain sensing signals to obtain at least one sensing signal index formed by the clustering; Based on the sensing state trend vectors of each of the sensing signal indexes, obtain the past sensing state trend vectors of the past multi-mode main board sensing signals included in each of the sensing signal indexes.

8. The method according to claim 1, characterized in that The method further includes: In the case where the fault detection and analysis of the multi-mode main board sensing signal are completed, obtain the fault detection and analysis report of the multi-mode main board sensing signal; the fault detection and analysis report includes the fault location area, the fault location category, and the fault element analysis text; Based on the fault location area, the fault location category, and the target sensing state trend vector, obtain the fault detection relationship vector of the multi-mode main board sensing signal; Based on the fault detection relationship vector and the fault element analysis text, construct the main board defect prediction vector of the multi-mode main board sensing signal; Perform defect prediction vector pairing between the main board defect prediction vector and the past main board defect prediction vector of the past multi-mode main board sensing signal to obtain the defect prediction vector pairing view for the multi-mode main board sensing signal.

9. A computer device, characterized in that, Comprising at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, such that the at least one processor executes the method according to any one of claims 1-8.

Images