Solid state disk fault detection method and device, computer equipment and storage medium

Abstract

The invention relates to a solid state disk fault detection method and device, computer equipment and a storage medium, and the method comprises the steps: collecting operation data of a solid state disk, preprocessing the operation data, and generating a standardized data set; performing preliminary dimension reduction on the standardized data set through a multi-stage dimension reduction algorithm, compressing the standardized data set, and extracting corresponding fault related features; performing secondary dimension reduction on the fault related features through a nonlinear dimension reduction algorithm, extracting corresponding fault deep features, and further obtaining fault feature data; and inputting the fault feature data into a preset fault detection model, and generating a fault early warning message when it is determined that the fault feature data is hard disk abnormality. The method has the effect of improving the detection accuracy of the solid state disk.

Description

Technical Field

[0001] The present application relates to the technical field of hard disk detection, and in particular to a solid state hard disk fault detection method, device, computer equipment and storage medium. Background Art

[0002] At present, solid-state drives are widely used in the field of data storage due to their high-speed data reading and writing and high reliability. However, since solid-state drives may be affected by multiple factors such as the number of read and write times, temperature changes, and operating environment during long-term use, resulting in hard disk performance degradation or even failure, existing solid-state drive fault detection methods are mainly based on SMART attributes, namely self-monitoring, analysis, and reporting technology. The SMART detection method mainly evaluates the status by collecting basic data such as the hard disk's read and write error rate and temperature. This method, which simply relies on changes in raw data values, has the problem of insufficient accuracy in practical applications.

[0003] The above-mentioned existing technical solutions have the following defects: the existing fault detection method cannot perform effective multi-level dimensionality reduction processing on hard disk fault characteristics, resulting in insufficient detection accuracy, so there is room for improvement. Summary of the invention

[0004] In order to improve the accuracy of hard disk detection, the present application provides a solid state hard disk fault detection method, device, computer equipment and storage medium.

[0005] The above-mentioned invention objective of the present application is achieved through the following technical solutions: A solid state hard disk fault detection method, the solid state hard disk fault detection method comprising: Collecting operation data of the solid state drive, and preprocessing the operation data to generate a standardized data set; Performing preliminary dimensionality reduction on the standardized data set through a multi-level dimensionality reduction algorithm, compressing the standardized data set, and extracting corresponding fault-related features; Performing secondary dimensionality reduction on the fault-related features through a nonlinear dimensionality reduction algorithm to extract corresponding deep fault features, thereby obtaining fault feature data; The fault characteristic data is input into a preset fault detection model, and when it is confirmed that the fault characteristic data indicates that the hard disk is abnormal, a fault warning message is generated.

[0006] By adopting the above technical solution, by collecting the operation data of the solid-state hard disk and pre-processing the operation data to generate a standardized data set, it is possible to remove noise and outliers in the data, ensure the consistency and reliability of the data, and thus improve the accuracy of subsequent dimensionality reduction and fault detection processing; by performing preliminary dimensionality reduction on the standardized data set through a multi-level dimensionality reduction algorithm, compressing the standardized data set, and extracting the corresponding fault-related features, it is possible to retain the key features of the fault information while reducing the data dimension, thereby reducing the model calculation amount and improving the accuracy of fault detection; by performing secondary dimensionality reduction on the fault-related features through a nonlinear dimensionality reduction algorithm, extracting the corresponding deep fault features, and then generating fault feature data, it is possible to further extract hidden deep-level fault information, thereby enhancing the model's ability to identify complex faults; by inputting the fault feature data into a preset fault detection model, a fault warning message is generated when it is confirmed that the fault feature data is a hard disk abnormality, and an alarm can be issued in time when a potential hard disk fault occurs, thereby reducing the risk of system downtime and ensuring the security of data storage.

[0007] In one example, the present application may be further configured as follows: performing preliminary dimensionality reduction on the standardized data set by a multi-level dimensionality reduction algorithm, compressing the standardized data set, and extracting corresponding fault-related features, specifically including: Performing principal component decomposition on all features in the standardized data set using a principal component analysis algorithm, and calculating the eigenvalue of each principal component; The principal component whose characteristic value is higher than a preset characteristic threshold is selected as the fault-related characteristic.

[0008] By adopting the above technical solution, by using the principal component analysis algorithm to decompose all the features in the standardized data set into principal components and calculating the eigenvalue of each principal component, it is possible to identify the main components in the data that are most relevant to the fault and remove redundant features, thereby reducing the data dimension while retaining the main information and improving the applicability of the data after dimensionality reduction; by selecting the principal components whose eigenvalues ​​are higher than the preset feature threshold as the fault-related features, it is possible to screen out features that have a high contribution to the fault information, thereby improving the fault detection model's ability to recognize key features and the accuracy of fault detection.

[0009] In one example, the present application may be further configured as follows: the selecting of the principal component having a feature value higher than a preset feature threshold as the fault-related feature further includes: Acquire the operation scenario of the solid state drive, and dynamically adjust the fault-related features according to the operation scenario to eliminate noise features that appear in a specific scenario; By utilizing the online feature screening mechanism, the dimension reduction threshold is updated according to the real-time operating status of the solid-state drive to ensure the applicability of the fault-related features under different working conditions.

[0010] By adopting the above technical solution, by obtaining the operating scenario of the solid-state drive and dynamically adjusting the fault-related features according to the operating scenario, the noise features that appear in specific scenarios are eliminated, and the feature set can be optimized in real time in different operating environments, reducing the interference of noise features on the detection results, thereby improving the adaptability and accuracy of fault detection; by utilizing the online feature screening mechanism, the dimensionality reduction threshold is updated according to the real-time operating status of the solid-state drive, which can ensure the representativeness of the fault features under different working conditions, thereby enhancing the detection accuracy and stability of the model under complex working conditions.

[0011] In one example, the present application may be further configured as follows: performing secondary dimensionality reduction on the fault-related features by a nonlinear dimensionality reduction algorithm to extract deep fault features, and then generating fault feature data, specifically including: Using the t-SNE algorithm to further reduce the dimension of the fault-related features after the initial dimension reduction, and mapping the multi-dimensional fault-related features to a low-dimensional space; By setting the number of neighbors and the learning rate of the t-SNE algorithm, the distribution of data in the low-dimensional space is optimized to ensure the distinguishability of fault features.

[0012] By adopting the above technical solution, by using the t-SNE algorithm to further reduce the dimension of the fault-related features after the initial dimensionality reduction, the multi-dimensional fault-related features are mapped to the low-dimensional space. The adjacent relationship of the fault features can be retained in the dimensionality reduction process, and the deep correlation between the features can be reflected, thereby improving the effectiveness of the data after dimensionality reduction in fault detection; by setting the number of neighbors and the learning rate of the t-SNE algorithm, the distribution of data in the low-dimensional space is optimized to ensure the distinguishability of the fault features, which can enhance the separation effect between different fault features, thereby improving the accuracy of the fault detection model.

[0013] In one example, the present application may be further configured as follows: the t-SNE algorithm is used to further reduce the dimension of the fault-related features after the preliminary dimension reduction, and the multi-dimensional fault-related features are mapped to a low-dimensional space, specifically including: Obtaining the distribution of the fault-related features in the multidimensional space, calculating the average neighbor distance of each fault feature data point in the high-dimensional space, and then determining the neighborhood density mean, and setting the neighborhood density mean as the neighborhood retention threshold; The number of iterations of the t-SNE algorithm is dynamically adjusted based on the average neighbor distance, so that the distribution of data points in the low-dimensional space is more consistent with the neighborhood relationship of the initial space.

[0014] By adopting the above technical solution, by obtaining the distribution of fault-related features in multi-dimensional space, calculating the average neighbor distance of each fault feature data point in high-dimensional space, and then determining the mean neighborhood density, the mean neighborhood density is set as the neighborhood retention threshold, which can maintain the data neighborhood relationship in the original space and provide a basis for the spatial distribution in the dimensionality reduction process, thereby ensuring the consistency of the spatial structure of the data after dimensionality reduction and enhancing the model's ability to recognize fault features; by dynamically adjusting the number of iterations of the t-SNE algorithm based on the average neighbor distance, the distribution of data points in the low-dimensional space is more consistent with the neighborhood relationship of the initial space, and the spatial consistency of the fault features can be maintained during the dimensionality reduction process, thereby improving the fault detection model's recognition effect on deep-level features.

[0015] In one example, the present application may be further configured as follows: the construction of the fault detection model specifically includes: Obtaining historical fault types and fault feature data corresponding to the historical fault types, and constructing a feature label mapping relationship, associating each of the fault feature data with the corresponding fault category, and obtaining a model training data set; The preset machine learning model is iteratively trained through the model training data set, and the fault detection model is generated by learning the corresponding relationship between the fault feature data and the fault category.

[0016] By adopting the above technical solution, by obtaining historical fault types and fault feature data corresponding to the historical fault types, and constructing a feature label mapping relationship, each fault feature data is associated with the corresponding fault category to obtain a model training data set, which can provide clearly labeled fault data for model training, ensure that the model learns the characteristic patterns of different fault categories, thereby improving the accuracy and reliability of the fault detection model; through the model training data set, the preset machine learning model is iteratively trained, and the fault detection model is generated by learning the corresponding relationship between the fault feature data and the fault category, which can continuously optimize the feature recognition ability of the model, improve the judgment accuracy of the fault category, thereby enhancing the performance of the fault detection model, so that it has higher accuracy and stability in practical applications.

[0017] The second object of the invention is achieved by the following technical solutions: A solid state hard disk fault detection device, the solid state hard disk fault detection device comprising: A data acquisition module, used to collect the operating data of the solid state drive, and pre-process the operating data to generate a standardized data set; A preliminary dimensionality reduction module, used to perform preliminary dimensionality reduction on the standardized data set through a multi-level dimensionality reduction algorithm, compress the standardized data set, and extract corresponding fault-related features; A deep feature extraction module is used to perform secondary dimensionality reduction on the fault-related features through a nonlinear dimensionality reduction algorithm, extract the corresponding deep features of the fault, and then obtain fault feature data; The fault detection module is used to input the fault characteristic data into a preset fault detection model, and generate a fault warning message when it is confirmed that the fault characteristic data is a hard disk abnormality.

[0018] By adopting the above technical solution, by collecting the operation data of the solid-state hard disk and pre-processing the operation data to generate a standardized data set, it is possible to remove noise and outliers in the data, ensure the consistency and reliability of the data, and thus improve the accuracy of subsequent dimensionality reduction and fault detection processing; by performing preliminary dimensionality reduction on the standardized data set through a multi-level dimensionality reduction algorithm, compressing the standardized data set, and extracting the corresponding fault-related features, it is possible to retain the key features of the fault information while reducing the data dimension, thereby reducing the model calculation amount and improving the accuracy of fault detection; by performing secondary dimensionality reduction on the fault-related features through a nonlinear dimensionality reduction algorithm, extracting the corresponding deep fault features, and then generating fault feature data, it is possible to further extract hidden deep-level fault information, thereby enhancing the model's ability to identify complex faults; by inputting the fault feature data into a preset fault detection model, a fault warning message is generated when it is confirmed that the fault feature data is a hard disk abnormality, and an alarm can be issued in time when a potential hard disk fault occurs, thereby reducing the risk of system downtime and ensuring the security of data storage.

[0019] The third objective of the present application is achieved through the following technical solutions: A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned solid state drive fault detection method when executing the computer program.

[0020] The fourth objective of the present application is achieved through the following technical solutions: A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the solid-state hard disk fault detection method are implemented.

[0021] In summary, this application includes the following beneficial technical effects: 1. By collecting the operation data of the solid-state hard disk and preprocessing the operation data to generate a standardized data set, it is possible to remove noise and outliers in the data, ensure the consistency and reliability of the data, and thus improve the accuracy of subsequent dimensionality reduction and fault detection processing; by using a multi-level dimensionality reduction algorithm to perform preliminary dimensionality reduction on the standardized data set, compress the standardized data set, and extract the corresponding fault-related features, it is possible to retain the key features of the fault information while reducing the data dimension, thereby reducing the amount of model calculation and improving the accuracy of fault detection; by using a nonlinear dimensionality reduction algorithm to perform secondary dimensionality reduction on the fault-related features, extract the corresponding deep fault features, and then generate fault feature data, it is possible to further extract hidden deep-level fault information, thereby enhancing the model's ability to identify complex faults; by inputting the fault feature data into the preset fault detection model, a fault warning message is generated when it is confirmed that the fault feature data is a hard disk abnormality, and an alarm can be issued in time when a potential hard disk fault occurs, thereby reducing the risk of system downtime and ensuring the security of data storage; 2. By using the principal component analysis algorithm to decompose all the features in the standardized data set into principal components and calculating the eigenvalue of each principal component, the main components in the data that are most relevant to the fault can be identified and redundant features can be removed, thereby reducing the data dimension while retaining the main information and improving the applicability of the data after dimensionality reduction; by selecting the principal components with eigenvalues ​​higher than the preset feature threshold as the fault-related features, the features that have a high contribution to the fault information can be screened out, thereby improving the fault detection model's ability to identify key features and the accuracy of fault detection; 3. By obtaining the operating scenario of the SSD and dynamically adjusting the fault-related features according to the operating scenario, the noise features that appear in specific scenarios can be eliminated. This can optimize the feature set in real time in different operating environments and reduce the interference of noise features on the detection results, thereby improving the adaptability and accuracy of fault detection. By utilizing the online feature screening mechanism and updating the dimensionality reduction threshold according to the real-time operating status of the SSD, the representativeness of the fault features under different working conditions can be ensured, thereby enhancing the detection accuracy and stability of the model under complex working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 is a flow chart of a solid state drive fault detection method in one embodiment of the present application; Figure 2 is a flowchart for implementing step S20 in the solid state drive fault detection method in one embodiment of the present application; Figure 3 is a flowchart for implementing step S22 in the solid state drive fault detection method in one embodiment of the present application; Figure 4 is a flowchart for implementing step S30 in a solid state drive fault detection method in an embodiment of the present application; Figure 5 is a flowchart for implementing step S31 in a solid state drive fault detection method in an embodiment of the present application; Figure 6 is a flowchart for implementing the construction of a fault detection model in a solid state drive fault detection method in one embodiment of the present application; Figure 7 is a principle block diagram of a solid state hard disk fault detection device in one embodiment of the present application; Figure 8 It is a schematic diagram of a device in an embodiment of the present application. DETAILED DESCRIPTION

[0023] The present application is further described in detail below in conjunction with the accompanying drawings.

[0024] In one embodiment, if Figure 1 As shown, the present application discloses a solid state drive fault detection method, which specifically includes the following steps: S10: Collecting the operation data of the solid state drive, and preprocessing the operation data to generate a standardized data set.

[0025] Specifically, the collection of operating data includes key attributes of the hard disk, such as read and write error rate, temperature, and power-on time. Each eigenvalue is continuously collected through internal sensors to form complete time series data, and the operating data is preprocessed by removing abnormal data, filling missing values, and performing data standardization. By setting upper and lower thresholds, outliers that are too high or too low are eliminated, and interpolation or mean filling methods are used to ensure data continuity. Finally, the preprocessed data set is standardized to make the range of its eigenvalues ​​relatively consistent, so as to reduce the scale differences between features and obtain a standardized data set.

[0026] S20: Perform preliminary dimensionality reduction on the standardized data set through a multi-level dimensionality reduction algorithm, compress the standardized data set, and extract corresponding fault-related features.

[0027] Specifically, a multi-level dimensionality reduction process is performed on the standardized data set. The high-dimensional features are first converted into multiple principal components. The principal components are sorted and the principal components with higher contributions are selected. Redundant features are removed to reduce the complexity of the data set and the interference caused by irrelevant data. This ensures that the retained features can fully represent the fault feature information. At the same time, the dimension of the data set is compressed for subsequent analysis.

[0028] S30: Perform secondary dimensionality reduction on the fault-related features through a nonlinear dimensionality reduction algorithm, extract the corresponding deep fault features, and then obtain fault feature data.

[0029] Specifically, a nonlinear dimensionality reduction algorithm is used to further reduce the fault-related features after the initial dimensionality reduction, so as to retain important deep fault features in the low-dimensional space and compress the fault-related features into a representative data set. The algorithm maps the multidimensional fault-related features to the low-dimensional space based on the similarity and distance relationship between the feature points, and extracts the potential relationship between the features to form a deep feature data set for fault detection.

[0030] S40: Inputting the fault characteristic data into a preset fault detection model, and generating a fault warning message when confirming that the fault characteristic data indicates that the hard disk is abnormal.

[0031] Specifically, the generated fault feature data is input into the preset fault detection model. The model performs a comprehensive analysis based on the distribution, similarity and change trend of the feature data to determine whether there is an abnormal state. When the model detects an abnormality, it generates a fault warning message and notifies the user or system to take preventive measures. The warning message can be sent to the monitoring interface or further disseminated through the notification system.

[0032] In one embodiment, if Figure 2 As shown, in step S20, the standardized data set is preliminarily reduced in dimension by a multi-level dimensionality reduction algorithm, the standardized data set is compressed, and the corresponding fault-related features are extracted, specifically including: S21: Use the principal component analysis algorithm to perform principal component decomposition on all features in the standardized data set and calculate the eigenvalue of each principal component.

[0033] Specifically, principal component decomposition is performed on the standardized data set, and each feature is converted into multiple principal components through linear combination. The eigenvalue and eigenvector of each principal component are calculated to sort the features. The principal components that have a greater contribution to the data are screened out according to the size of the eigenvalues, and redundant principal components with low eigenvalues ​​are eliminated to simplify the data structure and extract the main features.

[0034] S22: Selecting a principal component whose characteristic value is higher than a preset characteristic threshold as a fault-related characteristic.

[0035] Specifically, fault-related features are filtered according to the size of eigenvalues, and the eigenvalue threshold is set to filter out principal components with higher contributions. Principal components with eigenvalues ​​higher than the threshold are retained in the fault-related feature set, and principal components with low contributions are removed to compress the dimension of the data set. At the same time, redundant data is reduced to ensure that the fault-related features have a high amount of information and provide effective input for subsequent detection models.

[0036] In one embodiment, if Figure 3 As shown, in step S22, the principal component whose characteristic value is higher than the preset characteristic threshold is selected as the fault-related characteristic, and further includes: S221: Obtain an operation scenario of the solid state drive, and dynamically adjust fault-related features according to the operation scenario to eliminate noise features that appear in a specific scenario.

[0037] Specifically, the fault-related features are adjusted according to different SSD operating scenarios. The noise features in specific scenarios are identified by analyzing the operating parameters such as the hard disk's usage environment, load conditions, and temperature fluctuations. The fault feature screening criteria are dynamically adjusted based on the operating scenario information to eliminate irrelevant features in the scenario. By setting adjustment parameters, low-correlation noise features are eliminated in real time in different operating scenarios to ensure that the data set has high adaptability under different working conditions.

[0038] S222: Using an online feature screening mechanism, the dimension reduction threshold is updated according to the real-time operating status of the solid state drive to ensure the applicability of the fault-related features under different working conditions.

[0039] Specifically, the operating status of the solid-state drive is monitored in real time, and the dimensionality reduction threshold is adjusted through the online feature screening mechanism according to dynamic parameters such as temperature and read / write load. When the operating status changes significantly, the dimensionality reduction threshold is updated through the online screening mechanism, thereby dynamically adjusting the number of features in the fault-related feature set, making the fault detection data set more representative and maintaining a high applicability under different working conditions.

[0040] In one embodiment, if Figure 4 As shown, in step S30, the fault-related features are secondary reduced in dimension by a nonlinear dimensionality reduction algorithm to extract deep fault features, and then generate fault feature data, which specifically includes: S31: Use the t-SNE algorithm to further reduce the dimension of the fault-related features after the initial dimension reduction, and map the multi-dimensional fault-related features to a low-dimensional space.

[0041] Specifically, the t-SNE algorithm is used to further reduce the dimensionality of the fault-related features after the initial dimensionality reduction, and the multi-dimensional fault-related features are mapped to the low-dimensional space. The similarity between each fault feature data point is calculated and the adjacent relationship in the high-dimensional space is constructed. The coordinate position of each data point is gradually optimized through the transformation of the probability distribution, so that the relative position relationship between adjacent features is retained in the low-dimensional space, and finally a clustering structure is formed to highlight the spatial distribution of fault features.

[0042] S32: By setting the number of neighbors and learning rate of the t-SNE algorithm, the distribution of data in the low-dimensional space is optimized to ensure the distinguishability of fault features.

[0043] Specifically, the number of neighbors in the t-SNE algorithm is set to control the nearest neighbor relationship retained by each data point in the low-dimensional space. At the same time, the initial learning rate is set to make the dimensionality reduction convergence speed appropriate. The learning rate is gradually reduced during the iteration process to avoid overfitting. In this way, the relative position distribution of data points in the low-dimensional space is optimized, the distance between fault feature points is made clearer and more separable, and different fault features can be ensured to form significant distribution differences in the low-dimensional space.

[0044] In one embodiment, if Figure 5 As shown, in step S31, the fault-related features after preliminary dimensionality reduction are further reduced using the t-SNE algorithm to map the multi-dimensional fault-related features to a low-dimensional space, specifically including: S311: Obtain the distribution of fault-related features in the multi-dimensional space, calculate the average neighbor distance of each fault feature data point in the high-dimensional space, and then determine the neighborhood density mean, and set the neighborhood density mean as the neighborhood retention threshold.

[0045] Specifically, the distribution of fault-related features in multidimensional space is analyzed, and the average neighbor distance is obtained by calculating the distance from each data point to its neighbor data point. The neighbor distance of each data point is summed up and the average is taken to obtain the neighborhood density mean, which is used as the threshold for neighborhood preservation and used as a reference standard for maintaining neighbor relations in the subsequent dimensionality reduction process to ensure that the low-dimensional space after dimensionality reduction can better reflect the neighborhood relationship of the original space.

[0046] S312: Dynamically adjust the number of iterations of the t-SNE algorithm based on the average neighbor distance, so that the distribution of data points in the low-dimensional space is more consistent with the neighborhood relationship of the initial space.

[0047] Specifically, according to the neighbor distance information of the data points in the high-dimensional space, the number of t-SNE iterations is dynamically adjusted during the dimensionality reduction process. If the neighborhood distribution in the low-dimensional space fails to achieve the expected neighbor relationship, the number of iterations is increased to optimize the position distribution of the data points so that their neighborhood relationship is closer to the high-dimensional space. Otherwise, the number of iterations is reduced to ensure that the neighborhood relationship of the data points after dimensionality reduction is more consistent with the initial neighbor relationship in the high-dimensional space.

[0048] In one embodiment, if Figure 6 As shown, in step S40, i.e., construction of a fault detection model, specifically includes: S401: Obtain historical fault types and fault feature data corresponding to the historical fault types, and construct a feature label mapping relationship, associate each fault feature data with the corresponding fault category, and obtain a model training data set.

[0049] Specifically, the fault type information and corresponding fault feature values ​​of the solid-state drive are extracted from the historical operation data, and a label is associated with each fault feature data and the fault category to which it belongs to form a feature-label mapping relationship. The model training data set containing the correspondence between fault features and categories is sorted out so that it can be used as input data in the model training process to ensure that the model can learn the characteristic patterns under different fault categories.

[0050] S402: Iteratively train the preset machine learning model through the model training data set, and generate a fault detection model by learning the correspondence between fault feature data and fault categories.

[0051] Specifically, the sorted model training data set is input into the preset machine learning model, and the initial parameters are set for multiple rounds of iterative training. By repeatedly adjusting the parameters to optimize the model, the model gradually learns the correspondence between fault feature data and fault categories. During the training process, the error is calculated and the weight is adjusted in each round of iteration to reduce the fault detection error and improve the model accuracy, and finally a fault detection model that can identify different fault modes is generated.

[0052] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0053] In one embodiment, a solid state drive fault detection device is provided, and the solid state drive fault detection device corresponds one-to-one to the solid state drive fault detection method in the above embodiment. Figure 7 As shown, the solid state drive fault detection device includes a data acquisition module, a preliminary dimension reduction module, a deep feature extraction module, and a fault detection module. The detailed description of each functional module is as follows: A data acquisition module is used to collect the operating data of the solid-state hard disk, and pre-process the operating data to generate a standardized data set; A preliminary dimensionality reduction module is used to perform preliminary dimensionality reduction on the standardized data set through a multi-level dimensionality reduction algorithm, compress the standardized data set, and extract corresponding fault-related features; The deep feature extraction module is used to perform secondary dimension reduction on the fault-related features through a nonlinear dimension reduction algorithm, extract the corresponding deep features of the fault, and then obtain the fault feature data; The fault detection module is used to input the fault characteristic data into a preset fault detection model, and generate a fault warning message when it is confirmed that the fault characteristic data is a hard disk abnormality.

[0054] Optionally, the preliminary dimensionality reduction module specifically includes: The principal component analysis submodule is used to perform principal component decomposition on all features in the standardized data set using the principal component analysis algorithm and calculate the eigenvalue of each principal component; The feature screening submodule is used to select the principal components whose feature values ​​are higher than the preset feature threshold as the fault-related features.

[0055] Optionally, the feature screening submodule also includes: A scenario adjustment unit, used to obtain the operating scenario of the solid-state drive, and dynamically adjust the fault-related features according to the operating scenario to eliminate the noise features that appear in a specific scenario; The online screening unit is used to update the dimension reduction threshold according to the real-time operation status of the solid-state drive by utilizing the online feature screening mechanism, so as to ensure the applicability of the fault-related features under different working conditions.

[0056] Optionally, the deep feature extraction module specifically includes: The nonlinear dimension reduction submodule is used to further reduce the dimension of the fault-related features after the initial dimension reduction using the t-SNE algorithm, and map the multi-dimensional fault-related features to a low-dimensional space; The parameter optimization submodule is used to optimize the distribution of data in low-dimensional space by setting the number of neighbors and learning rate of the t-SNE algorithm to ensure the distinguishability of fault features.

[0057] Optionally, the nonlinear dimensionality reduction submodule specifically includes: A neighborhood calculation unit is used to obtain the distribution of fault-related features in the multi-dimensional space, calculate the average neighbor distance of each fault feature data point in the high-dimensional space, and then determine the neighborhood density mean, and set the neighborhood density mean as the neighborhood retention threshold; The iteration control unit is used to dynamically adjust the number of iterations of the t-SNE algorithm based on the average neighbor distance, so that the distribution of data points in the low-dimensional space is more consistent with the neighborhood relationship of the initial space.

[0058] Optionally, the construction of the fault detection model specifically includes: The training data set generation submodule is used to obtain historical fault types and fault feature data corresponding to the historical fault types, and to construct a feature label mapping relationship, and to associate each fault feature data with the corresponding fault category to obtain a model training data set; The model training submodule is used to iteratively train the preset machine learning model through the model training data set, and generate a fault detection model by learning the correspondence between fault feature data and fault categories.

[0059] For the specific definition of the solid-state hard disk fault detection device, please refer to the definition of the solid-state hard disk fault detection method above, which will not be repeated here. Each module in the above-mentioned solid-state hard disk fault detection device can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0060] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 8 The computer device includes a processor, a memory, a network interface and a database connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a solid-state hard disk fault detection method is implemented.

[0061] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the following steps are implemented: Collect the operating data of the solid-state drive, pre-process the operating data, and generate a standardized data set; Perform preliminary dimensionality reduction on the standardized data set through a multi-level dimensionality reduction algorithm, compress the standardized data set, and extract the corresponding fault-related features; The fault-related features are reduced twice by a nonlinear dimensionality reduction algorithm to extract the corresponding deep fault features, and then the fault feature data is obtained; The fault characteristic data is input into a preset fault detection model, and when it is confirmed that the fault characteristic data indicates that the hard disk is abnormal, a fault warning message is generated.

[0062] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented: Collect the operating data of the solid-state drive, pre-process the operating data, and generate a standardized data set; Perform preliminary dimensionality reduction on the standardized data set through a multi-level dimensionality reduction algorithm, compress the standardized data set, and extract the corresponding fault-related features; The fault-related features are reduced twice by a nonlinear dimensionality reduction algorithm to extract the corresponding deep fault features, and then the fault feature data is obtained; The fault characteristic data is input into a preset fault detection model, and when it is confirmed that the fault characteristic data indicates that the hard disk is abnormal, a fault warning message is generated.

[0063] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0064] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0065] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A solid state hard disk fault detection method, characterized in that: The solid state drive fault detection method comprises: Collecting operation data of the solid state drive, and preprocessing the operation data to generate a standardized data set; Performing preliminary dimensionality reduction on the standardized data set through a multi-level dimensionality reduction algorithm, compressing the standardized data set, and extracting corresponding fault-related features; Performing secondary dimensionality reduction on the fault-related features through a nonlinear dimensionality reduction algorithm to extract corresponding deep fault features, thereby obtaining fault feature data; The fault characteristic data is input into a preset fault detection model, and when it is confirmed that the fault characteristic data indicates that the hard disk is abnormal, a fault warning message is generated.

2. The solid state drive fault detection method according to claim 1, characterized in that: The method of performing preliminary dimensionality reduction on the standardized data set by using a multi-level dimensionality reduction algorithm, compressing the standardized data set, and extracting corresponding fault-related features specifically includes: Performing principal component decomposition on all features in the standardized data set using a principal component analysis algorithm, and calculating the eigenvalue of each principal component; The principal component whose characteristic value is higher than a preset characteristic threshold is selected as the fault-related characteristic.

3. The solid state drive fault detection method according to claim 2, characterized in that: The selecting the principal component having a characteristic value higher than a preset characteristic threshold as the fault-related characteristic further includes: Acquire the operation scenario of the solid state drive, and dynamically adjust the fault-related features according to the operation scenario to eliminate noise features that appear in a specific scenario; By utilizing the online feature screening mechanism, the dimension reduction threshold is updated according to the real-time operating status of the solid-state drive to ensure the applicability of the fault-related features under different working conditions.

4. The solid state drive fault detection method according to claim 1, characterized in that: The method of performing secondary dimensionality reduction on the fault-related features by using a nonlinear dimensionality reduction algorithm to extract deep fault features and then generate fault feature data specifically includes: Using the t-SNE algorithm to further reduce the dimension of the fault-related features after the initial dimension reduction, and mapping the multi-dimensional fault-related features to a low-dimensional space; By setting the number of neighbors and the learning rate of the t-SNE algorithm, the distribution of data in the low-dimensional space is optimized to ensure the distinguishability of fault features.

5. The solid state drive fault detection method according to claim 4, characterized in that: The method further performs dimensionality reduction processing on the fault-related features after preliminary dimensionality reduction using the t-SNE algorithm to map the multi-dimensional fault-related features to a low-dimensional space, specifically including: Obtaining the distribution of the fault-related features in the multidimensional space, calculating the average neighbor distance of each fault feature data point in the high-dimensional space, and then determining the neighborhood density mean, and setting the neighborhood density mean as the neighborhood retention threshold; The number of iterations of the t-SNE algorithm is dynamically adjusted based on the average neighbor distance, so that the distribution of data points in the low-dimensional space is more consistent with the neighborhood relationship of the initial space.

6. The solid state drive fault detection method according to claim 1, characterized in that: The construction of the fault detection model specifically includes: Obtaining historical fault types and fault feature data corresponding to the historical fault types, and constructing a feature label mapping relationship, associating each of the fault feature data with the corresponding fault category, and obtaining a model training data set; The preset machine learning model is iteratively trained through the model training data set, and the fault detection model is generated by learning the corresponding relationship between the fault feature data and the fault category.

7. A solid state hard disk fault detection device, characterized in that: The solid state hard disk fault detection device comprises: A data acquisition module, used to collect the operating data of the solid state drive, and pre-process the operating data to generate a standardized data set; A preliminary dimensionality reduction module, used to perform preliminary dimensionality reduction on the standardized data set through a multi-level dimensionality reduction algorithm, compress the standardized data set, and extract corresponding fault-related features; A deep feature extraction module is used to perform secondary dimensionality reduction on the fault-related features through a nonlinear dimensionality reduction algorithm, extract the corresponding deep features of the fault, and then obtain fault feature data; The fault detection module is used to input the fault characteristic data into a preset fault detection model, and generate a fault warning message when it is confirmed that the fault characteristic data is a hard disk abnormality.

8. The solid state hard disk fault detection device according to claim 7, characterized in that: The preliminary dimension reduction module specifically includes: A principal component analysis submodule is used to perform principal component decomposition on all features in the standardized data set using a principal component analysis algorithm and calculate the eigenvalue of each principal component; The feature screening submodule is used to select the principal component whose feature value is higher than a preset feature threshold as the fault-related feature.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the solid state drive failure detection method according to any one of claims 1 to 6 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the solid state drive failure detection method according to any one of claims 1 to 6 are implemented.

Images