WO2023115875A1 - 硬件设备维护方法、装置及电子设备 - Google Patents

硬件设备维护方法、装置及电子设备 Download PDF

Info

Publication number
WO2023115875A1
WO2023115875A1 PCT/CN2022/101794 CN2022101794W WO2023115875A1 WO 2023115875 A1 WO2023115875 A1 WO 2023115875A1 CN 2022101794 W CN2022101794 W CN 2022101794W WO 2023115875 A1 WO2023115875 A1 WO 2023115875A1
Authority
WO
WIPO (PCT)
Prior art keywords
information
fault
level
svm model
type
Prior art date
Application number
PCT/CN2022/101794
Other languages
English (en)
French (fr)
Inventor
夏敏捷
卢道和
饶俊明
郑晓腾
陈扬东
云玉柱
谢倩倩
杨振林
Original Assignee
深圳前海微众银行股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳前海微众银行股份有限公司 filed Critical 深圳前海微众银行股份有限公司
Publication of WO2023115875A1 publication Critical patent/WO2023115875A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]

Definitions

  • the embodiments of the present application relate to the technical field of equipment maintenance, and in particular, to a hardware equipment maintenance method, device, and electronic equipment.
  • the existing SVM model is limited to binary classification scenarios, that is, it can only identify two or one type of faults. Due to the wide variety of hardware devices, the corresponding number of fault types is also relatively large, resulting in the generalization of the SVM model. The performance is poor, which reduces the accuracy and efficiency of fault type identification, thereby affecting the normal operation of the server.
  • the purpose of the present application is to provide a hardware equipment maintenance method, device and electronic equipment, so as to improve the accuracy and efficiency of fault type identification.
  • the embodiment of the present application provides a hardware device maintenance method, including:
  • the status information is information corresponding to each hardware device in the server to be maintained;
  • the state information is input into the trained multi-level support vector machine SVM model, so that the multi-level SVM model performs multiple identification and classification processing on the state information to obtain a fault information set containing at least one type of fault ;
  • the state information is input into the trained multi-level SVM model, so that the multi-level SVM model performs multiple identification and classification processing on the state information to obtain a Fault information set, including:
  • the state information is input into the trained multi-level SVM model, so that the first-level SVM module in the multi-level SVM model identifies and classifies the state information to obtain the first fault type information and other faults type information;
  • the other fault type information contains more than one type of fault type, then input the other fault type information into the second-level SVM module in the multi-level SVM model for identification and classification processing, and obtain the second Fault type information and new additional fault type information;
  • the type of fault contained in the new other fault type information is still more than one type, re-execute the input of the other fault type information into the second-level SVM module in the multi-level SVM model for identification From the classification process and subsequent steps until the new other fault type information contains one type of fault, a fault information set containing at least one fault type is obtained.
  • the method further includes:
  • the non-fault information set is input into the multi-level SVM model, so that the multi-level SVM model performs identification and classification processing on the non-fault information set to obtain a level information set including at least one risk level.
  • the input of the non-fault information set into the multi-level SVM model makes the multi-level SVM model identify and classify the non-fault information set to obtain at least one risk level
  • the level information set of including:
  • the new other level information If the number of risk levels contained in the new other level information is still greater than one, re-execute the step of inputting the other level information into the fourth level SVM module in the multilevel SVM model for identification and Classification processing and subsequent steps until the number of risk levels included in the new other level information is one, and a level information set including at least one risk level is obtained.
  • the state training information set includes multiple sets of state training information, each set of state training information includes state information of a hardware device, and a target failure type corresponding to the state information of the hardware device;
  • the initial multi-level SVM model is trained according to the state information of the hardware device included in each set of state training information, and the target fault type corresponding to the state information of the hardware device, to obtain a trained multi-level SVM model.
  • the trained multi-level SVM model after obtaining the trained multi-level SVM model, it also includes:
  • the state test information set is input into the trained multi-level SVM model to perform multiple identification and classification processes to obtain the fault test information set;
  • the evaluating the accuracy of the multi-level SVM model according to the fault test information set includes:
  • Accuracy Avg(Acc i ) determines the classification accuracy rate of the multi-level SVM model
  • Acc i represents the classification accuracy of each level of SVM module in the multilevel SVM model
  • T p1 represents the number of samples that belong to the target fault type and are correctly classified into the target fault type
  • F pp represents the number of samples that do not belong to the target fault type but are incorrectly classified as the target fault type
  • Fnn represents the number of samples that belong to the target fault type but are incorrectly classified as the target fault type.
  • the number of samples classified as other fault types T n1 represents the number of samples that do not belong to the target fault type and are correctly classified
  • ⁇ i represents the accurate factor coefficient of each fault type
  • ⁇ i 1-X*X T
  • X represents all The matrix corresponding to the state test information in the state test information set
  • X T represents the transposition of X
  • i represents the number of fault types
  • F represents the recall rate of the multi-level SVM model
  • ⁇ represents the ratio of the target samples corresponding to the target fault type in the samples classified into the target fault type
  • ⁇ max ⁇ 1 , ⁇ 2 ,. .. ⁇ i ⁇
  • generating a prompt for passing the multi-level SVM model test includes:
  • a multi-level SVM model test passing prompt is generated.
  • the embodiment of the present application provides an apparatus for maintaining hardware equipment, including:
  • An acquisition module configured to acquire status information, wherein the status information is information corresponding to each hardware device in the server to be maintained;
  • a processing module configured to input the state information into the trained multi-level support vector machine SVM model, so that the multi-level SVM model performs multiple identification and classification processing on the state information to obtain at least one fault type of fault information set;
  • the processing module is further configured to perform fault maintenance on each hardware device in the server to be maintained according to the fault information set including at least one fault type.
  • an embodiment of the present application provides an electronic device, including: a processor, and a memory communicatively connected to the processor;
  • the memory stores computer-executable instructions
  • the processor executes the computer-executed instructions stored in the memory to implement the hardware device maintenance method described in the above first aspect and various possible designs of the first aspect.
  • the embodiment of the present application provides a computer-readable storage medium, the computer-readable storage medium stores computer-executable instructions, and when the processor executes the computer-executable instructions, the above-mentioned first aspect and the first aspect can be realized.
  • the hardware device maintenance method there are various possible designs of the hardware device maintenance method.
  • the embodiment of the present application provides a computer program product, including a computer program.
  • the computer program When the computer program is executed by a processor, it can realize the hardware device maintenance described in the above first aspect and various possible designs of the first aspect. method.
  • the embodiment of the present application provides a hardware device maintenance method, device, and electronic device.
  • the state information corresponding to each hardware device in the server to be maintained can be obtained first, and then the state information can be input into the multi-level training completed Identify and classify the SVM model to obtain a fault information set containing at least one fault type, and then carry out fault maintenance for each hardware device in the server to be maintained according to the fault information set containing at least one fault type, through pre-trained multiple
  • the multi-level SVM model is used to realize the identification of various fault types, which improves the versatility of the SVM model, and improves the accuracy and efficiency of fault type identification, thereby ensuring the normal operation of the server.
  • FIG. 1 is a schematic structural diagram of an application system of a hardware device maintenance method provided in an embodiment of the present application
  • FIG. 2 is a schematic flowchart of a hardware device maintenance method provided in an embodiment of the present application
  • FIG. 3 is a schematic diagram of the application of the multi-level SVM model identification process provided by the embodiment of the present application.
  • FIG. 4 is a schematic structural diagram of a device for maintaining hardware equipment provided by an embodiment of the present application.
  • FIG. 5 is a schematic diagram of a hardware structure of an electronic device provided by an embodiment of the present application.
  • the types of hardware devices used in servers are also increasing.
  • the types of hardware devices applied in the server may include memory, hard disk, processor, fan, and so on.
  • failures may occur due to excessive running time or mismatched models.
  • SVM Small Vector Machine, Support Vector Machine
  • the existing SVM model is limited to binary classification scenarios, that is, it can only identify two or one type of faults. Due to the wide variety of hardware devices, the corresponding number of fault types is also relatively large, resulting in the generalization of the SVM model. The performance is poor, which reduces the accuracy and efficiency of fault type identification, thereby affecting the normal operation of the server.
  • the present application uses a pre-trained multi-level SVM model to realize the identification of multiple fault types, which not only improves the versatility of the SVM model, but also improves the accuracy and efficiency of fault type identification. Thus, the technical effect of normal operation of the server is guaranteed.
  • FIG. 1 is a schematic structural diagram of an application system of a hardware device maintenance method provided by an embodiment of the present application.
  • the application system may include: a server 101 and a server 102 to be maintained.
  • the server to be maintained 102 can be a running device or a non-running device, and different hardware devices are deployed in the server to be maintained 102, for example, it can be a hard disk, a fan, a processor, etc., and different hardware devices can correspond to
  • the server 101 can obtain the state information corresponding to each hardware device from the server 102 to be maintained, and input the state information corresponding to different hardware devices into the pre-trained multi-level SVM model 103 for identification and classification processing, and obtain A fault information set containing at least one type of fault can then generate a fault maintenance request according to the fault information set, and send the fault maintenance request to the corresponding server 102 to be maintained.
  • the server 101 may also serve as a server to be maintained and send status information corresponding to each hardware device to other servers.
  • the server 101 may also directly acquire the status information corresponding to different hardware devices in each server to be maintained from the database.
  • FIG. 2 is a schematic flowchart of a method for maintaining a hardware device provided by an embodiment of the present application, and the method of this embodiment may be executed by the server 101 . As shown in Figure 2, the method of this embodiment may include:
  • S201 Obtain status information, where the status information is information corresponding to each hardware device in the server to be maintained.
  • the status of each hardware device in the server running the financial service may be monitored to determine whether each hardware device is running normally.
  • the state information may be information corresponding to each hardware device in the running server to be maintained, or may be information corresponding to each hardware device in the non-running server to be maintained.
  • status information can include hardware brand, model, running time, abnormal alarm log, etc.
  • S202 Input the state information into the trained multi-level support vector machine SVM model, so that the multi-level SVM model performs multiple identification and classification processing on the state information to obtain a fault information set including at least one type of fault.
  • the status information can be input into the pre-trained multi-level SVM model that can identify multiple fault types, and the multi-level SVM model can perform multiple identification and classification processing on the status information Finally, a fault information set containing at least one type of fault is obtained.
  • the state information is input into the trained multi-level SVM model, so that the multi-level SVM model performs multiple identification and classification processing on the state information, and obtains a fault information set containing at least one type of fault, specifically Can include:
  • the other fault type information contains more than one type of fault type, then input the other fault type information into the second-level SVM module in the multi-level SVM model for identification and classification processing, and obtain the second Fault type information and new additional fault type information.
  • the type of fault contained in the new other fault type information is still more than one type, re-execute the input of the other fault type information into the second-level SVM module in the multi-level SVM model for identification From the classification process and subsequent steps until the new other fault type information contains one type of fault, a fault information set containing at least one fault type is obtained.
  • the multilevel SVM model includes different SVM modules, and the more SVM modules are, the more fault types the multilevel SVM model can identify.
  • the multilevel SVM model includes two SVM modules, then the multilevel SVM model can identify three types of faults.
  • the first-level SVM module in the multi-level SVM model can first identify and classify the state information to obtain the first fault type information and other faults type information. Then judge whether the type of failure contained in other failure type information is greater than one type, if the type of failure type contained in other failure type information is less than or equal to one type (that is, one or zero), then the identification is completed and no longer executes next steps. If other fault type information contains more than one type of fault type, the other fault type information is input to the second-level SVM module in the multi-level SVM model for identification and classification processing, and the second fault type information and new other fault types are obtained. Fault type information.
  • the identification is completed, and the subsequent steps are not performed. If the types of faults contained in the new other fault type information are more than one, use the new other fault type information as other fault type information, and re-execute the second-level SVM module of inputting other fault type information into the multi-level SVM model Perform identification and classification processing and subsequent steps until the new other fault type information contains one type of fault, and obtain a fault information set containing at least one fault type.
  • Table 1 is a table of fault information. As shown in Table 1, Table 1 includes four different types of fault information.
  • the SVM model for binary classification problems, it classifies the state information containing fault information, and stops after distinguishing one of the fault types each time. The remaining undifferentiated fault types are then classified using the SVM model, and the cycle repeats until all fault types are distinguished, enabling faster discovery of fault types and increasing the number of proactive server maintenance, thereby reducing maintenance costs.
  • S203 Perform fault maintenance for each hardware device in the server to be maintained according to the fault information set including at least one fault type.
  • a corresponding fault maintenance request can be generated according to each fault type in the fault information set, and the fault maintenance request can be sent to the corresponding server to be maintained, so that the server to be maintained can automatically complete the repair of the fault according to the fault maintenance request.
  • the server to be maintained can send a maintenance request to the terminal device of the corresponding operation and maintenance personnel.
  • the server can generate a corresponding fault maintenance request according to each fault type in the fault information set, and send the fault maintenance request to the terminal device of the corresponding operation and maintenance personnel.
  • the state information corresponding to each hardware device in the server to be maintained can be obtained first, and then the state information can be input into the trained multi-level SVM model for identification and classification processing, and the faults containing at least one fault type can be obtained information set, and then carry out fault maintenance on each hardware device in the server to be maintained according to the fault information set containing at least one fault type, and realize the identification of multiple fault types through the pre-trained multi-level SVM model, which improves the SVM model. Versatility, and improve the accuracy and efficiency of fault type identification, thereby ensuring the normal operation of the server.
  • the method may further include:
  • a non-fault information set is obtained according to the state information and the fault information set including at least one type of fault.
  • the non-fault information set is input into the multi-level SVM model, so that the multi-level SVM model identifies and classifies the non-fault information set to obtain a level information set including at least one risk level.
  • the acquired state information corresponding to each hardware device in the server to be maintained may not only include the state information corresponding to the hardware device that has failed, but also include the state information corresponding to the hardware device that is running normally.
  • the status information corresponding to the hardware devices in normal operation is classified according to the risk level, which can predict the hardware devices that may fail in advance, thereby further reducing the frequency of possible server failures.
  • the information related to the fault information set in the state information can be eliminated to obtain the non-fault information set, and then the non-fault information set is input into the multi-level SVM model for identification and classification processing, and a level including at least one risk level can be obtained information set.
  • the multi-level SVM model includes different SVM modules, and the number of grades that can be divided is positively correlated with the number of SVM modules, that is, the more the number of SVM modules, the more grades that can be divided, and the concentration of grade information The more risk classes are included.
  • the non-fault information set is input into the multi-level SVM model, so that the multi-level SVM model identifies and classifies the non-fault information set to obtain a level information set including at least one risk level, Specifically can include:
  • the new other level information If the number of risk levels contained in the new other level information is still greater than one, re-execute the step of inputting the other level information into the fourth level SVM module in the multilevel SVM model for identification and Classification processing and subsequent steps until the number of risk levels included in the new other level information is one, and a level information set including at least one risk level is obtained.
  • the third-level SVM module in the multi-level SVM model can first identify and classify the non-fault information set to obtain the first-level information and other class information. Then judge whether the number of risk levels contained in other level information is greater than one, if the number of risk levels contained in other level information is less than or equal to one (that is, one or zero), then the identification is completed, no Then proceed to the next steps. If the number of risk levels contained in the other level information is greater than one, then input the other level information into the fourth level SVM module in the multilevel SVM model for identification and classification processing, and obtain the second level information and new other levels information.
  • the identification is completed, and the subsequent steps are not performed. If the number of risk levels contained in the new other level information is greater than one, use the new other level information as other level information, and re-execute the input of other level information into the fourth level SVM module in the multilevel SVM model. Identification and classification processing and subsequent steps until the number of risk levels included in the new other level information is one, and a level information set including at least one risk level is obtained.
  • FIG. 3 is an application schematic diagram of the multi-level SVM model identification process provided by the embodiment of the present application.
  • the state information can be obtained first, and then the state information can be input into the first-level SVM module for identification and classification, and the first fault type information and other fault type information can be obtained, and then the fault types contained in other fault type information can be judged Whether it is more than one type, if more than one type, input other fault type information into the second-level SVM module for identification and classification, obtain the second fault type information and new other fault type information, and the new other fault type information contains only one fault type.
  • the fault information set can be obtained according to the first fault type information, the second fault type information and new other fault type information.
  • the non-fault information set can be obtained, and then the non-fault information set can be input into the third-level SVM module for identification and classification, and the first-level information and other level information can be obtained, and then the other level information can be judged Whether the number of included risk levels is more than one, if more than one, input other level information into the fourth level SVM module for identification and classification, obtain the second level information and new other level information, and the new The number of risk classes included in other class information is one. Then obtain the level information set according to the first level information, the second level information and new other level information.
  • the method may further include:
  • the state training information set includes multiple sets of state training information, each set of state training information includes hardware device state information, and a target fault type corresponding to the hardware device state information.
  • the initial multi-level SVM model is trained according to the state information of the hardware device included in each set of state training information, and the target fault type corresponding to the state information of the hardware device, to obtain a trained multi-level SVM model.
  • the initial multi-level SVM model before using the multi-level SVM model to identify the state information and determine the fault information set, the initial multi-level SVM model can be trained through the state training information set to obtain the trained multi-level SVM model, and then Then apply the trained multi-level SVM model to identify and classify the newly acquired state information, and finally obtain a fault information set containing at least one fault type.
  • the state training information set may include multiple sets of state training information, and each set of state training information may include the state information of the hardware device and the target fault type corresponding to the state information of the hardware device.
  • the process of determining the state training information set can be specifically: collecting the historical fault data Dd in the server fault work order within the preset time period, and then performing data cleaning and integration processing on the data set Dd according to the preset data cleaning rules, In turn, incomplete and invalid data are removed. It is important to note that data cleaning here can retain duplicate data.
  • M1, M2, and M3 may represent three fault types.
  • the features to be mined can be fault time, fault components, abnormal alarm information, etc.
  • each group of state training information in the state training information set may include the state information of the hardware device that has failed and the target failure type corresponding to the state information of the hardware device that has failed , may also include status information of hardware devices without failure, and risk levels corresponding to the status information of hardware devices without failure.
  • the initial multi-level SVM model can be trained through the state training information set to obtain the trained multi-level SVM model, and then the trained multi-level SVM model can be used to identify and classify the newly acquired state information, and finally get A fault information set containing at least one type of fault, and then continue to identify and classify the status information to obtain a graded information set containing at least one risk level.
  • the method may also include:
  • the state test information set is input into the trained multi-level SVM model for multiple identification and classification processes to obtain the fault test information set.
  • the trained multi-level SVM model after obtaining the trained multi-level SVM model, in order to improve the accuracy of the subsequent application of the multi-level SVM model, the trained multi-level SVM model can be tested through the state test information set, and when the test passes After that, the multi-level SVM model is applied to identify and classify the obtained state information.
  • evaluating the accuracy of the multi-level SVM model according to the fault test information set may specifically include:
  • Accuracy Avg(Acc i ) determines the classification accuracy rate of the multi-level SVM model
  • Acc i represents the classification accuracy of each level of SVM module in the multilevel SVM model
  • T p1 represents the number of samples that belong to the target fault type and are correctly classified into the target fault type
  • F pp represents the number of samples that do not belong to the target fault type but are incorrectly classified as the target fault type
  • Fnn represents the number of samples that belong to the target fault type but are incorrectly classified as The number of samples classified as other fault types
  • T n1 represents the number of samples that do not belong to the target fault type and are correctly classified
  • ⁇ i represents the accurate factor coefficient of each fault type
  • ⁇ i 1-X*X T
  • X represents all The matrix corresponding to the state test information in the above state test information set
  • X T represents the transposition of X
  • i represents the number of fault types.
  • F represents the recall rate of the multi-level SVM model
  • ⁇ represents the ratio of the target samples corresponding to the target fault type in the samples classified into the target fault type
  • ⁇ max ⁇ 1 , ⁇ 2 ,. .. ⁇ i ⁇
  • ⁇ i can represent the exact factor coefficients of various types of faults, and can dynamically change according to the corresponding features in the iterative process according to the frequent itemsets of different features.
  • there are three fault types and an initial value is set for each fault type ⁇ i at the beginning of model training, which is 1 by default.
  • the weight contribution to the data samples in each iteration is also changing (the value gradually becomes smaller).
  • X represents the matrix corresponding to the state test information in the state test information set, but it needs to be converted into a matrix format to facilitate machine recognition.
  • X represents the matrix corresponding to the state test information in the state test information set, but it needs to be converted into a matrix format to facilitate machine recognition.
  • the article only makes a collection classification, so that it is easy to understand that the actual business classification, brand, type, faulty component, subsystem type, etc. should be considered.
  • Tp1 represents the number of samples that belong to the target fault type and are correctly classified into the target fault type, for example, the number of samples that belong to the A1 type and are also classified into the A1 class.
  • Tn1 represents the number of samples that do not belong to the target fault type and are correctly classified, for example: A4 is classified to A2 regardless of the brand.
  • Fpp represents the number of samples that do not belong to the target fault type but are misclassified as such, for example: A2 does not belong to class A1 but is classified as class A1.
  • Fnn represents the number of samples that belong to the target fault type but are misclassified to other classes, for example: belong to A4 but are classified to A1.
  • F represents the recall rate of the multi-level SVM model
  • F is a model measurement index that takes both precision and recall into account.
  • the multilevel SVM model can identify three types of faults and three risk levels.
  • Table 2 is the multi-level SVM model test result table provided by the embodiment of the present application. As shown in Table 2, after obtaining the trained multi-level SVM model, the trained multi-level SVM model can be tested through the test information set. Output the corresponding six classification results, where M k (can be M 1 , M 2 , M 3 ) indicates the type of fault, and D r (can be D 1 , D 2 , D 3 ) indicates the risk level.
  • a prompt for passing the multi-level SVM model test is generated, which may specifically include:
  • a multi-level SVM model test passing prompt is generated.
  • the precision rate threshold and the recall rate threshold can be customized and set according to the actual application scenario, and will not be limited in detail here.
  • the F determined by M k and D r can be determined separately, and then the average value of F corresponding to M k and D r can be determined as the final F.
  • Table 3 is a comparison table of algorithm classification accuracy results provided by the embodiment of the present application.
  • the multi-level SVM model can be tested, and the multi-level SVM model can be compared with the classification of existing algorithms at the current stage. Comparing the performance, it can be clearly seen that the Accuracy value obtained according to the multi-level SVM model has a certain improvement compared with the traditional Decision tree, Random forest and SVM algorithms.
  • Table 4 is the comparison table of the algorithm recall rate results provided by this application. As shown in Table 4, the F determined according to the multi-level SVM model is 0.83, which is equivalent to the performance of the traditional algorithm, and the multi-level SVM model under certain conditions does not affect the performance of each server. The classification of hardware devices works better.
  • the management of server hardware resources is more refined, which facilitates earlier discovery of problems and makes corresponding decisions, improves the reliability of the server, and can also improve the performance of the server at the bottom layer to ensure the stability of the server .
  • Fig. 4 is a schematic structural diagram of a hardware device maintenance device provided by the embodiment of this application. As shown in Fig. 4, the device provided by this embodiment may include :
  • the obtaining module 401 is configured to obtain state information, wherein the state information is information corresponding to each hardware device in the server to be maintained.
  • the processing module 402 is configured to input the state information into the trained multi-level support vector machine SVM model, so that the multi-level SVM model performs multiple identification and classification processing on the state information to obtain at least one A set of fault information for the fault type.
  • processing module 402 is further configured to:
  • the state information is input into the trained multi-level SVM model, so that the first-level SVM module in the multi-level SVM model identifies and classifies the state information to obtain the first fault type information and other faults type information.
  • the other fault type information contains more than one type of fault type, then input the other fault type information into the second-level SVM module in the multi-level SVM model for identification and classification processing, and obtain the second Fault type information and new additional fault type information.
  • the type of fault contained in the new other fault type information is still more than one type, re-execute the input of the other fault type information into the second-level SVM module in the multi-level SVM model for identification From the classification process and subsequent steps until the new other fault type information contains one type of fault, a fault information set containing at least one fault type is obtained.
  • processing module 402 is also used for:
  • a non-fault information set is obtained according to the state information and the fault information set including at least one type of fault.
  • the non-fault information set is input into the multi-level SVM model, so that the multi-level SVM model identifies and classifies the non-fault information set to obtain a level information set including at least one risk level.
  • the processing module 402 is further configured to:
  • the new other level information If the number of risk levels contained in the new other level information is still greater than one, re-execute the step of inputting the other level information into the fourth level SVM module in the multilevel SVM model for identification and Classification processing and subsequent steps until the number of risk levels included in the new other level information is one, and a level information set including at least one risk level is obtained.
  • the processing module 402 is further configured to perform fault maintenance on each hardware device in the server to be maintained according to the fault information set including at least one fault type.
  • processing module 402 is further configured to:
  • the state training information set includes multiple groups of state training information, each group of state training information includes the state information of the hardware device, and the corresponding target failure type of the state information of the hardware device.
  • the initial multi-level SVM model is trained according to the state information of the hardware device included in each set of state training information, and the target fault type corresponding to the state information of the hardware device, to obtain a trained multi-level SVM model.
  • processing module 402 is also used for:
  • the state test information set is input into the trained multi-level SVM model for multiple identification and classification processes to obtain the fault test information set.
  • processing module 402 is further configured to:
  • Acc i represents the classification accuracy of each level of SVM module in the multilevel SVM model
  • T p1 represents the number of samples that belong to the target fault type and are correctly classified as the target fault type
  • F pp represents the number of samples that do not belong to the target fault type but are incorrectly classified as the target fault type
  • Fnn represents the number of samples that belong to the target fault type but are incorrectly classified as The number of samples of other fault types
  • T n1 represents the number of samples that do not belong to the target fault type and are correctly classified
  • ⁇ i represents the accuracy factor coefficient of each type of fault
  • ⁇ i 1-X*X T
  • X represents the state The matrix corresponding to the state test information in the test information set
  • X T represents the transposition of X
  • i represents the number of fault types.
  • F represents the recall rate of the multi-level SVM model
  • ⁇ represents the ratio of the target samples corresponding to the target fault type in the samples classified into the target fault type
  • ⁇ max ⁇ 1 , ⁇ 2 ,. .. ⁇ n ⁇
  • processing module 402 is also used for:
  • a multi-level SVM model test passing prompt is generated.
  • the device provided in the embodiment of the present application can implement the method in the above embodiment as shown in FIG. 2 , and its implementation principle and technical effect are similar, and will not be repeated here.
  • FIG. 5 is a schematic diagram of a hardware structure of an electronic device provided in an embodiment of the present application.
  • a device 500 provided in this embodiment includes a processor 501 and a memory communicatively connected to the processor. Wherein, the processor 501 and the memory 502 are connected through a bus 503 .
  • the processor 501 executes the computer-executed instructions stored in the memory 502, so that the processor 501 executes the hardware device maintenance method in the foregoing method embodiments.
  • the processor can be a central processing unit (English: Central Processing Unit, referred to as: CPU), and can also be other general-purpose processors, digital signal processors (English: Digital Signal Processor, referred to as: DSP), application specific integrated circuit (English: Application Specific Integrated Circuit, referred to as: ASIC), etc.
  • a general-purpose processor may be a microprocessor, or the processor may be any conventional processor, or the like. The steps of the method disclosed in conjunction with the invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules in the processor.
  • the memory may include high-speed RAM memory, and may also include non-volatile storage NVM, such as at least one disk memory.
  • the bus can be an Industry Standard Architecture (Industry Standard Architecture, ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (Extended Industry Standard Architecture, EISA) bus, etc.
  • ISA Industry Standard Architecture
  • PCI Peripheral Component Interconnect
  • EISA Extended Industry Standard Architecture
  • the bus can be divided into address bus, data bus, control bus and so on.
  • the buses in the drawings of the present application are not limited to only one bus or one type of bus.
  • the embodiment of the present application also provides a computer-readable storage medium, where computer-executable instructions are stored in the computer-readable storage medium, and when the processor executes the computer-executable instructions, the hardware device maintenance method of the foregoing method embodiment is implemented.
  • An embodiment of the present application further provides a computer program product, including a computer program, and when the computer program is executed by a processor, the method for maintaining a hardware device as described above is implemented.
  • the above-mentioned computer-readable storage medium can be realized by any type of volatile or non-volatile storage device or their combination, such as static random access memory (SRAM), electrically erasable Programmable Read Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), Magnetic Memory, Flash Memory, Magnetic or Optical Disk.
  • SRAM static random access memory
  • EEPROM electrically erasable Programmable Read Only Memory
  • EPROM Erasable Programmable Read Only Memory
  • PROM Programmable Read Only Memory
  • ROM Read Only Memory
  • Magnetic Memory Flash Memory
  • Magnetic or Optical Disk Readable storage media can be any available media that can be accessed by a general purpose or special purpose computer.
  • An exemplary readable storage medium is coupled to the processor such the processor can read information from, and write information to, the readable storage medium.
  • the readable storage medium can also be a component of the processor.
  • the processor and the readable storage medium may be located in Application Specific Integrated Circuits (ASIC for short).
  • ASIC Application Specific Integrated Circuits
  • the processor and the readable storage medium can also exist in the device as discrete components.
  • the aforementioned program can be stored in a computer-readable storage medium.
  • the program executes the steps including the above-mentioned method embodiments; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk and other various media that can store program codes.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Artificial Intelligence (AREA)
  • Debugging And Monitoring (AREA)

Abstract

一种硬件设备维护方法、装置及电子设备，方法包括获取状态信息，其中，状态信息为待维护服务器中各硬件设备对应的信息（S201），将状态信息输入至训练完成的多级SVM模型中，使得多级SVM模型对状态信息进行多次识别与分类处理，得到包含至少一种故障类型的故障信息集（S202），根据包含至少一种故障类型的故障信息集对待维护服务器中的各硬件设备进行故障维护（S203）；进而提高了故障类型识别的准确性与效率，保证了服务器的正常运行。

Description

硬件设备维护方法、装置及电子设备
本申请要求于2021年12月24日提交中国专利局、申请号为202111603721.8、申请名称为“硬件设备维护方法、装置及电子设备”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请实施例涉及设备维护技术领域，尤其涉及一种硬件设备维护方法、装置及电子设备。
背景技术
随着计算机技术的发展，越来越多的技术应用在金融领域，传统金融业正在逐步向金融科技(Fintech)转变，设备维护技术也不例外，但由于金融行业的安全性、实时性要求，也对设备维护技术提出了更高的要求。
现有技术中，为了满足各金融业务增长的需求，服务器应用的数量越来越多，在服务器中应用的硬件设备类型也越来越多。服务器中的各硬件设备在运行过程中，可能会由于运行时间过长或者型号不匹配等原因产生故障。为了尽快排除故障，可以通过SVM(Support Vector Machine，支持向量机)模型来对获取到的硬件设备故障信息进行识别，得到不同硬件设备对应的故障类型，进而根据不同硬件设备对应的故障类型对应进行维护，保证服务器的正常运行。
然而，现有的SVM模型仅局限于二分类场景，即仅可以识别故障类型为两种或一种的情况，由于硬件设备的种类繁多，对应的故障类型数量也比较多，导致SVM模型的通用性差，降低了故障类型识别的准确性与效率，进而影响了服务器的正常运行。
发明内容
本申请的目的在于提供一种硬件设备维护方法、装置及电子设备，以提高故障类型识别的准确性与效率。
第一方面，本申请实施例提供一种硬件设备维护方法，包括：
获取状态信息，其中，所述状态信息为待维护服务器中各硬件设备对应的信息；
将所述状态信息输入至训练完成的多级支持向量机SVM模型中，使得所述多级SVM模型对所述状态信息进行多次识别与分类处理，得到包含至少一种故障类型的故障信息集；
根据所述包含至少一种故障类型的故障信息集对所述待维护服务器中的各硬件设备进行故障维护。
可选的，所述将所述状态信息输入至训练完成的多级SVM模型中，使得所述多级SVM模型对所述状态信息进行多次识别与分类处理，得到包含至少一种故障类型的故障信息集，包括：
将所述状态信息输入至训练完成的多级SVM模型中，使得所述多级SVM模型中的第一级SVM模块对所述状态信息进行识别与分类处理，得到第一故障类型信息以及其他故 障类型信息；
若所述其他故障类型信息中包含的故障类型种类大于一种，则将所述其他故障类型信息输入至所述多级SVM模型中的第二级SVM模块中进行识别与分类处理，得到第二故障类型信息以及新的其他故障类型信息；
若所述新的其他故障类型信息中包含的故障类型种类仍大于一种，则重新执行所述将所述其他故障类型信息输入至所述多级SVM模型中的第二级SVM模块中进行识别与分类处理及之后的步骤直至所述新的其他故障类型信息中包含的故障类型为一种，得到包含至少一种故障类型的故障信息集。
可选的，在所述得到包含至少一种故障类型的故障信息集之后，还包括：
根据所述状态信息以及所述包含至少一种故障类型的故障信息集，得到非故障信息集；
将所述非故障信息集输入至所述多级SVM模型中，使得所述多级SVM模型对所述非故障信息集、进行识别与分类处理，得到包含至少一个风险性等级的等级信息集。
可选的，所述将所述非故障信息集输入至所述多级SVM模型中，使得所述多级SVM模型对所述非故障信息集进行识别与分类处理，得到包含至少一个风险性等级的等级信息集，包括：
将所述非故障信息集输入至所述多级SVM模型中，使得所述多级SVM模型中的第三级SVM模块对所述非故障信息集中的各非故障信息进行识别与分类处理，得到第一等级信息以及其他等级信息；
若所述其他等级信息中包含的风险性等级数量大于一种，则将所述其他等级信息输入至所述多级SVM模型中的第四级SVM模块中进行识别与分类处理，得到第二等级信息以及新的其他等级信息；
若所述新的其他等级信息中包含的风险性等级数量仍大于一种，则重新执行所述将所述其他等级信息输入至所述多级SVM模型中的第四级SVM模块中进行识别与分类处理及之后的步骤直至所述新的其他等级信息中包含的风险性等级数量为一种，得到包含至少一个风险性等级的等级信息集。
可选的，在所述获取状态信息之前，还包括：
获取状态训练信息集，其中，所述状态训练信息集中包含多组状态训练信息，每组所述状态训练信息中包含硬件设备的状态信息，以及所述硬件设备的状态信息对应的目标故障类型；
根据每组所述状态训练信息中包含的硬件设备的状态信息，以及所述硬件设备的状态信息对应的目标故障类型对初始多级SVM模型进行训练，得到训练完成的多级SVM模型。
可选的，在所述得到训练完成的多级SVM模型之后，还包括：
获取状态测试信息集；
将所述状态测试信息集输入至训练完成的多级SVM模型中进行多次识别与分类处理，得到故障测试信息集；
根据所述故障测试信息集对所述多级SVM模型进行准确率评估，并在所述准确率评估的结果符合预设评估条件时，生成多级SVM模型测试通过提示。
可选的，所述根据所述故障测试信息集对所述多级SVM模型进行准确率评估，包括：
根据表达式：Accuracy＝Avg(Acc i)确定所述多级SVM模型的分类准确率；
其中，Acc i表示所述多级SVM模型中每级SVM模块的分类准确率，
Figure PCTCN2022101794-appb-000001
其中，T p1表示属于目标故障类型且被正确分类到目标故障类型的样本数量，F pp表示不属于目标故障类型却被错误归到目标故障类型的样本数量，Fnn表示属于目标故障类型却被错误归为其他故障类型的样本数量，T n1表示不属于目标故障类型且被正确分类的样本数量，α i表示各类故障类型的准确因子系数，α i＝1-X*X T，X表示所述状态测试信息集中状态测试信息对应的矩阵，X T表示X的转置，i表示故障类型的个数；
和/或，
根据表达式：
Figure PCTCN2022101794-appb-000002
确定所述多级SVM模型的召回率；
其中，F表示所述多级SVM模型的召回率，β表示被分为目标故障类型的样本中所述目标故障类型对应的目标样本所占的比率，β＝max{α 1，α 2，...α i}，
Figure PCTCN2022101794-appb-000003
Figure PCTCN2022101794-appb-000004
可选的，所述在所述准确率评估的结果符合预设评估条件时，生成多级SVM模型测试通过提示，包括：
在所述多级SVM模型的分类准确率超过预设准确率阈值，和/或所述多级SVM模型的召回率超过预设召回率阈值时，生成多级SVM模型测试通过提示。
第二方面，本申请实施例提供一种硬件设备维护装置，包括：
获取模块，用于获取状态信息，其中，所述状态信息为待维护服务器中各硬件设备对应的信息；
处理模块，用于将所述状态信息输入至训练完成的多级支持向量机SVM模型中，使得所述多级SVM模型对所述状态信息进行多次识别与分类处理，得到包含至少一种故障类型的故障信息集；
所述处理模块，还用于根据所述包含至少一种故障类型的故障信息集对所述待维护服务器中的各硬件设备进行故障维护。
第三方面，本申请实施例提供一种电子设备，包括：处理器，以及与所述处理器通信连接的存储器；
所述存储器存储计算机执行指令；
所述处理器执行所述存储器存储的计算机执行指令，以实现如上第一方面以及第一方面各种可能的设计所述的硬件设备维护方法。
第四方面，本申请实施例提供一种计算机可读存储介质，所述计算机可读存储介质中存储有计算机执行指令，当处理器执行所述计算机执行指令时，以实现如上第一方面以及第一方面各种可能的设计所述的硬件设备维护方法。
第五方面，本申请实施例提供一种计算机程序产品，包括计算机程序，所述计算机程序被处理器执行时，以实现如上第一方面以及第一方面各种可能的设计所述的硬件设备维护方法。
本申请实施例提供了一种硬件设备维护方法、装置及电子设备，采用上述方案后，可以先获取待维护服务器中各硬件设备对应的状态信息，然后将该状态信息输入至训练完成的多级SVM模型中进行识别与分类处理，得到包含至少一种故障类型的故障信息集，再 根据包含至少一种故障类型的故障信息集对待维护服务器中的各硬件设备进行故障维护，通过预先训练的多级SVM模型来实现对多种故障类型的识别，提高了SVM模型的通用性，且提高了故障类型识别的准确性与效率，进而保证了服务器的正常运行。
附图说明
图1为本申请实施例提供的硬件设备维护方法的应用系统的架构示意图；
图2为本申请实施例提供的硬件设备维护方法的流程示意图；
图3为本申请实施例提供的多级SVM模型识别过程的应用示意图；
图4为本申请实施例提供的硬件设备维护装置的结构示意图；
图5为本申请实施例提供的电子设备的硬件结构示意图。
具体实施方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
本申请的说明书和权利要求书及上述附图中的术语“第一”、“第二”、“第三”“第四”等(如果存在)是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。应该理解这样使用的数据在适当情况下可以互换，以便这里描述的本申请的实施例还能够包括除了图示或描述的那些实例以外的其他顺序实例。此外，术语“包括”和“具有”以及他们的任何变形，意图在于覆盖不排他的包含，例如，包含了一系列步骤或单元的过程、方法、系统、产品或设备不必限于清楚地列出的那些步骤或单元，而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它步骤或单元。
服务器应用的数量越来越多，为了更好的满足金融业务的需求，在服务器中应用的硬件设备类型也越来越多。示例性的，服务器中应用的硬件设备类型可以包含内存、硬盘、处理器、风扇等。且服务器中的各硬件设备在运行过程中，可能会由于运行时间过长或者型号不匹配等原因产生故障。为了尽快排除故障，可以通过SVM(Support Vector Machine，支持向量机)模型来对获取到的硬件设备故障信息进行识别，得到不同硬件设备对应的故障类型，进而根据不同硬件设备对应的故障类型对应进行维护，保证服务器的正常运行。然而，现有的SVM模型仅局限于二分类场景，即仅可以识别故障类型为两种或一种的情况，由于硬件设备的种类繁多，对应的故障类型数量也比较多，导致SVM模型的通用性差，降低了故障类型识别的准确性与效率，进而影响了服务器的正常运行。
基于上述技术问题，本申请通过预先训练的多级SVM模型来实现对多种故障类型的识别的方式，达到了既提高了SVM模型的通用性，又提高了故障类型识别的准确性与效率，进而保证了服务器的正常运行的技术效果。
图1为本申请实施例提供的硬件设备维护方法的应用系统的架构示意图，如图1所示，在该应用系统中，可以包括：服务器101，待维护服务器102。待维护服务器102可以为正在运行的设备或者为没有运行的设备，且待维护服务器102中部署有不同的硬件设备，示例性的，可以为硬盘、风扇、处理器等，不同的硬件设备可以对应不同的状态信息，服 务器101可以从待维护服务器102中获取各硬件设备对应的状态信息，并将不同硬件设备对应的状态信息输入至预先训练的多级SVM模型103中进行识别与分类处理，得到包含至少一种故障类型的故障信息集，然后可以根据故障信息集生成故障维护请求，并将故障维护请求发送至对应的待维护服务器102中。
此外，服务器101也可以作为待维护服务器向其他服务器发送各硬件设备对应的状态信息。服务器101还可以从数据库中直接获取各待维护服务器中不同硬件设备对应的状态信息。
下面以具体地实施例对本申请的技术方案进行详细说明。下面这几个具体的实施例可以相互结合，对于相同或相似的概念或过程可能在某些实施例不再赘述。
图2为本申请实施例提供的硬件设备维护方法的流程示意图，本实施例的方法可以由服务器101执行。如图2所示，本实施例的方法，可以包括：
S201：获取状态信息，其中，状态信息为待维护服务器中各硬件设备对应的信息。
在本实施例中，为了保证各金融业务的正常运行，可以对运行该金融业务的服务器中各硬件设备的状态进行监测，确定各硬件设备是否在正常运行。此外，在确定硬件设备出现故障时，为了提高故障维护的效率，可以先对出现的故障进行分类，然后根据确定的类别进一步的进行故障维护。
其中，状态信息可以为运行中的待维护服务器中各硬件设备对应的信息，也可以非运行中的待维护服务器中各硬件设备对应的信息。
更进一步的，状态信息可以包含硬件品牌、型号、运行时长、异常报警日志等。
S202：将状态信息输入至训练完成的多级支持向量机SVM模型中，使得多级SVM模型对状态信息进行多次识别与分类处理，得到包含至少一种故障类型的故障信息集。
在本实施例中，在得到状态信息之后，可以将状态信息输入至预先训练完成的可以识别多种故障类型的多级SVM模型中，多级SVM模型可以对状态信息进行多次识别与分类处理的过程，最终得到包含至少一种故障类型的故障信息集。
进一步的，将状态信息输入至训练完成的多级SVM模型中，使得所述多级SVM模型对所述状态信息进行多次识别与分类处理，得到包含至少一种故障类型的故障信息集，具体可以包括：
将状态信息输入至训练完成的多级SVM模型中，使得所述多级SVM模型中的第一级SVM模块对所述状态信息进行识别与分类处理，得到第一故障类型信息以及其他故障类型信息。
若所述其他故障类型信息中包含的故障类型种类大于一种，则将所述其他故障类型信息输入至所述多级SVM模型中的第二级SVM模块中进行识别与分类处理，得到第二故障类型信息以及新的其他故障类型信息。
若所述新的其他故障类型信息中包含的故障类型种类仍大于一种，则重新执行所述将所述其他故障类型信息输入至所述多级SVM模型中的第二级SVM模块中进行识别与分类处理及之后的步骤直至所述新的其他故障类型信息中包含的故障类型为一种，得到包含至少一种故障类型的故障信息集。
具体的，多级SVM模型中包括有不同的SVM模块，且SVM模块越多，多级SVM模型可以识别的故障类型越多。示例性的，若多级SVM模型中包括有两个SVM模块，则 多级SVM模型可以识别的故障类型为3种。
对应的，在将状态信息输入至训练完成的多级SVM模型中之后，多级SVM模型中的第一级SVM模块可以先对状态信息进行识别与分类处理，得到第一故障类型信息以及其他故障类型信息。然后判断其他故障类型信息中包含的故障类型种类是否大于一种，若其他故障类型信息中包含的故障类型种类小于或等于一种(即为一种或零种)，则识别完成，不再执行后续步骤。若其他故障类型信息中包含的故障类型种类大于一种，则将其他故障类型信息输入至多级SVM模型中的第二级SVM模块中进行识别与分类处理，得到第二故障类型信息以及新的其他故障类型信息。然后再判断新的其他故障类型信息中包含的故障类型种类是否大于一种，若新的其他故障类型信息中包含的故障类型种类等于一种，则识别完成，不再执行后续步骤。若新的其他故障类型信息中包含的故障类型种类大于一种，将新的其他故障类型信息作为其他故障类型信息，并重新执行将其他故障类型信息输入至多级SVM模型中的第二级SVM模块中进行识别与分类处理及之后的步骤直至所述新的其他故障类型信息中包含的故障类型为一种，得到包含至少一种故障类型的故障信息集。
示例性的，表1为故障信息表，如表1所示，在表1中包含四种不同类型的故障信息。
表1 故障信息表
Figure PCTCN2022101794-appb-000005
利用了SVM模型针对二分类问题的优势，对包含故障信息的状态信息进行分类，每次将其中一个故障类型区分出来后，即可停止。剩余未区分的故障类型再利用SVM模型进行二分类，循环往复直到所有故障类类型均被区分，能够更快的发现故障类型以及提高主动维护服务器的次数，从而降低维护成本。
S203：根据包含至少一种故障类型的故障信息集对待维护服务器中的各硬件设备进行故障维护。
在本实施例中，在得到包含至少一种故障类型的故障信息集之后，为了保证各金融业务的正常运行，可以针对故障信息集中的各故障类型尽快进行维护。
进一步的，可以根据故障信息集中的各故障类型生成对应的故障维护请求，并把该故障维护请求发送至对应的待维护服务器，使得待维护服务器可以根据该故障维护请求自动完成故障的修复。对于不能自动完成故障修复的待维护服务器，待维护服务器可以发送维护请求至对应运维人员的终端设备。获取，服务器可以根据故障信息集中的各故障类型生成对应的故障维护请求，并把该故障维护请求发送至对应运维人员的终端设备。
采用上述方案后，可以先获取待维护服务器中各硬件设备对应的状态信息，然后将该状态信息输入至训练完成的多级SVM模型中进行识别与分类处理，得到包含至少一种故 障类型的故障信息集，再根据包含至少一种故障类型的故障信息集对待维护服务器中的各硬件设备进行故障维护，通过预先训练的多级SVM模型来实现对多种故障类型的识别，提高了SVM模型的通用性，且提高了故障类型识别的准确性与效率，进而保证了服务器的正常运行。
基于图2的方法，本说明书实施例还提供了该方法的一些具体实施方案，下面进行说明。
此外，在另一实施例中，在得到包含至少一种故障类型的故障信息集之后，所述方法还可以包括：
根据所述状态信息以及所述包含至少一种故障类型的故障信息集，得到非故障信息集。
将所述非故障信息集输入至所述多级SVM模型中，使得所述多级SVM模型对所述非故障信息集进行识别与分类处理，得到包含至少一个风险性等级的等级信息集。
在本实施例中，在获取到的待维护服务器中各硬件设备对应的状态信息中除了可以包含出现故障的硬件设备对应的状态信息，还可以包含正常运行的硬件设备对应的状态信息，通过对正常运行的硬件设备对应的状态信息进行风险性等级分类，可以提前预测可能会出现故障的硬件设备，从而进一步减少了服务器可能出现故障的频次。
具体的，可以将状态信息中与故障信息集相关的信息剔除，得到非故障信息集，然后将非故障信息集输入至多级SVM模型中进行识别与分类处理，得到包含至少一个风险性等级的等级信息集。
此外，多级SVM模型中包括不同的SVM模块，且可以划分的等级个数与SVM模块的个数呈正相关，即SVM模块的个数越多，可以划分的等级个数越多，等级信息集中包含的风险性等级越多。
进一步的，将非故障信息集输入至所述多级SVM模型中，使得所述多级SVM模型对所述非故障信息集进行识别与分类处理，得到包含至少一个风险性等级的等级信息集，具体可以包括：
将所述非故障信息集输入至所述多级SVM模型中，使得所述多级SVM模型中的第三级SVM模块对所述非故障信息集中的各非故障信息进行识别与分类处理，得到第一等级信息以及其他等级信息。
若所述其他等级信息中包含的风险性等级数量大于一种，则将所述其他等级信息输入至所述多级SVM模型中的第四级SVM模块中进行识别与分类处理，得到第二等级信息以及新的其他等级信息；
若所述新的其他等级信息中包含的风险性等级数量仍大于一种，则重新执行所述将所述其他等级信息输入至所述多级SVM模型中的第四级SVM模块中进行识别与分类处理及之后的步骤直至所述新的其他等级信息中包含的风险性等级数量为一种，得到包含至少一个风险性等级的等级信息集。
具体的，在将非故障信息集输入至训练完成的多级SVM模型中之后，多级SVM模型中的第三级SVM模块可以先对非故障信息集进行识别与分类处理，得到第一等级信息以及其他等级信息。然后判断其他等级信息中包含的风险性等级的数量是否大于一种，若其他等级信息中包含的风险性等级的数量小于或等于一种(即为一种或零种)，则识别完成，不再执行后续步骤。若其他等级信息中包含的风险性等级的数量大于一种，则将其他等级 信息输入至多级SVM模型中的第四级SVM模块中进行识别与分类处理，得到第二等级信息以及新的其他等级信息。然后再判断新的其他等级信息中包含的风险性等级的数量是否大于一种，若新的其他等级信息中包含的风险性等级的数量等于一种，则识别完成，不再执行后续步骤。若新的其他等级信息中包含的风险性等级的数量大于一种，将新的其他等级信息作为其他等级信息，并重新执行将其他等级信息输入至多级SVM模型中的第四级SVM模块中进行识别与分类处理及之后的步骤直至所述新的其他等级信息中包含的风险性等级的数量为一种，得到包含至少一个风险性等级的等级信息集。
示例性的，图3为本申请实施例提供的多级SVM模型识别过程的应用示意图，如图3所示，在该实施例中，故障类型种类为三种，风险性等级的数量也为三种，则可以先获取状态信息，然后将状态信息输入至第一级SVM模块中进行识别与分类，得到第一故障类型信息和其他故障类型信息，然后判断其他故障类型信息中包含的故障类型种类是否大于一种，若大于一种，则将其他故障类型信息输入至第二级SVM模块中进行识别与分类，得到第二故障类型信息和新的其他故障类型信息，且新的其他故障类型信息中只包含一种故障类型。然后可以根据第一故障类型信息、第二故障类型信息和新的其他故障类型信息得到故障信息集。再根据状态信息和故障信息集得到非故障信息集，然后可以将非故障信息集输入至第三级SVM模块中进行识别与分类，得到第一等级信息和其他等级信息，然后判断其他等级信息中包含的风险性等级的数量是否大于一种，若大于一种，则将其他等级信息输入至第四级SVM模块中进行识别与分类，得到第二等级信息和新的其他等级信息，且新的其他等级信息中包含的风险性等级的数量为一种。再根据第一等级信息、第二等级信息和新的其他等级信息得到等级信息集。
此外，在另一实施例中，在获取状态信息之前，所述方法还可以包括：
获取状态训练信息集，其中，所述状态训练信息集中包含多组状态训练信息，每组所述状态训练信息中包含硬件设备的状态信息，以及所述硬件设备的状态信息对应的目标故障类型。
根据每组所述状态训练信息中包含的硬件设备的状态信息，以及所述硬件设备的状态信息对应的目标故障类型对初始多级SVM模型进行训练，得到训练完成的多级SVM模型。
在本实施例中，在应用多级SVM模型对状态信息进行识别，确定故障信息集之前，可以先通过状态训练信息集对初始多级SVM模型进行训练，得到训练完成的多级SVM模型，然后再应用训练完成的多级SVM模型对新获取到的状态信息进行识别与分类处理，最终得到包含至少一种故障类型的故障信息集。
其中，状态训练信息集中可以包含多组状态训练信息，每组状态训练信息中可以包含硬件设备的状态信息，以及硬件设备的状态信息对应的目标故障类型。对应的，确定状态训练信息集的过程具体可以为：收集预设时长内的服务器故障工单中的历史故障数据Dd，然后可以根据预设数据清洗规则对数据集Dd做数据清洗与集成处理，进而清除不完整的数据和无效数据。需要特别注意的是，此处数据清洗可以保留重复数据。再根据要实现的目标，消减数据集Dd中不必要的特征维度，得到聚合的规则集合Dd，然后可以根据预设故障特征挖掘规则对聚合后的规则集合Dd进行特征挖掘，归类出状态训练信息集Mk＝{M1，M2，M3…Mn},其中，n表示故障类型种类。示例性的，M1、M2、M3可以表示三种故障类型。挖掘的特征可以为故障时间、故障部件、异常报警信息等。
此外，为了进一步减少服务器出现故障的频次，状态训练信息集中的每组状态训练信息中除了可以包含出现故障的硬件设备的状态信息，以及出现故障的硬件设备的状态信息对应的目标故障类型之外，还可以包含没有出现故障的硬件设备的状态信息，以及没有出现故障的硬件设备的状态信息对应的风险性等级。可以通过该状态训练信息集对初始多级SVM模型进行训练，得到训练完成的多级SVM模型，然后再应用训练完成的多级SVM模型对新获取到的状态信息进行识别与分类处理，最终得到包含至少一种故障类型的故障信息集，然后继续对状态信息进行识别与分类处理，得到包含至少一个风险性等级的等级信息集。
另外，在得到训练完成的多级SVM模型之后，所述方法还可以包括：
获取状态测试信息集。
将所述状态测试信息集输入至训练完成的多级SVM模型中进行多次识别与分类处理，得到故障测试信息集。
根据所述故障测试信息集对所述多级SVM模型进行准确率评估，并在所述准确率评估的结果符合预设评估条件时，生成多级SVM模型测试通过提示。
在本实施例中，在得到训练完成的多级SVM模型之后，为了提高多级SVM模型后续应用的准确性，可以通过状态测试信息集对训练完成的多级SVM模型进行测试，并在测试通过之后，再应用该多级SVM模型对获取到的状态信息进行识别与分类处理。
进一步的，根据所述故障测试信息集对所述多级SVM模型进行准确率评估，具体可以包括：
根据表达式：
Accuracy＝Avg(Acc i)确定所述多级SVM模型的分类准确率，
其中，Acc i表示所述多级SVM模型中每级SVM模块的分类准确率，
Figure PCTCN2022101794-appb-000006
其中，T p1表示属于目标故障类型且被正确分类到目标故障类型的样本数量，F pp表示不属于目标故障类型却被错误归到目标故障类型的样本数量，Fnn表示属于目标故障类型却被错误归为其他故障类型的样本数量，T n1表示不属于目标故障类型且被正确分类的样本数量，α i表示各类故障类型的准确因子系数，α i＝1-X*X T，X表示所述状态测试信息集中状态测试信息对应的矩阵，X T表示X的转置，i表示故障类型的个数。
和/或根据表达式：
Figure PCTCN2022101794-appb-000007
确定所述多级SVM模型的召回率。
其中，F表示所述多级SVM模型的召回率，β表示被分为目标故障类型的样本中所述目标故障类型对应的目标样本所占的比率，β＝max{α 1，α 2，...α i}，
Figure PCTCN2022101794-appb-000008
Figure PCTCN2022101794-appb-000009
具体的，α i可以表示各类故障类型的准确因子系数，可以根据不同特征的频繁项集，在迭代过程根据对应特征动态变化。示例性的，故障类型有三个，在模型训练开始时，会对每个故障类型α i设置一个初始值，默认情况下是1。在网络训练过程中，随着数据量的不断递增，在每一次迭代的中针对数据样本的权值贡献量也在发生变化(数值逐渐变小)，当网络模型区域稳定时α i值也基本趋于稳定，即可以得到最终的系数α i，具体计算公式可以为：α i＝1-X*X T。X表示状态测试信息集中状态测试信息对应的矩阵，但是需要转 换成矩阵的格式，便于机器识别。然而，在实际考虑的时候不是单一特征，每个特征中还会包含更多，文中只是做了集合归类，便于理解实际要考虑业务分类、品牌、类型、故障部件、子系统类型等。
示例性的，继续以表1为例，Tp1表示属于目标故障类型且被正确分类到目标故障类型的样本数量，例如，属于A1类型也被分类到A1类的样本数量。Tn1表示不属于目标故障类型且被正确分类的样本数量，例如：不考虑品牌时，将A4归类到A2。Fpp表示不属于目标故障类型却被错误归为此类的样本数量，例如：A2不属于A1类却被归到A1类。Fnn表示属于该目标故障类型却被错误归为其他类的样本数量，例如：属于A4却被归类到A1。
此外，还可以根据表达式：
Figure PCTCN2022101794-appb-000010
确定多级SVM模型的召回率。
其中，F表示所述多级SVM模型的召回率，β表示被分为目标故障类型的样本中所述目标故障类型对应的目标样本所占的比率，β＝max{α 1，α 2，...α i}；
Figure PCTCN2022101794-appb-000011
Figure PCTCN2022101794-appb-000012
具体的，F为一种兼顾精确率和召回率的模型衡量指标，示例性的，多级SVM模型可以识别的故障类型种类为三种，风险性等级的数量也为三种。通过对多级SVM模型的反复训练，可以得到训练完成的多级SVM模型。表2为本申请实施例提供的多级SVM模型测试结果表，如表2所示，在得到训练完成的多级SVM模型之后，可以通过测试信息集对训练好的多级SVM模型进行测试，输出相应的六分类结果，其中，M k(可以为M 1、M 2、M 3)表示故障类型种类，D r(可以为D 1、D 2、D 3)表示风险性等级。
表2 多级SVM模型测试结果表
Figure PCTCN2022101794-appb-000013
此外，在所述准确率评估的结果符合预设评估条件时，生成多级SVM模型测试通过提示，具体可以包括：
在所述多级SVM模型的分类准确率超过预设准确率阈值，和/或所述多级SVM模型的召回率超过预设召回率阈值时，生成多级SVM模型测试通过提示。
具体的，可以在多级SVM模型的分类准确率或者多级SVM模型的召回率中的任一个符合条件时，确定多级SVM模型测试通过，也可以在多级SVM模型的分类准确率以及多级SVM模型的召回率中均符合条件时，才可以确定多级SVM模型测试通过。其中，准确率阈值和召回率阈值可以根据实际应用场景自定义进行设置，在此不再详细进行限制。
另外，在确定F时，为了提高确定的F的准确率，可以分别确定M k和D r确定的F，然后再将M k和D r对应的F的平均值确定为最终的F。
示例性的，表3为本申请实施例提供的算法分类准确率结果比对表，如表3所示，可以对多级SVM模型进行检验，将多级SVM模型与现阶段已有算法的分类性能进行对比，可以较明显的看出根据多级SVM模型得到的Accuracy值相较于传统的Decision tree、Random forest以及SVM算法有一定的提升。
表3 算法分类准确率结果比对表
算法 Accuracy
Decision tree 0.69
Random forest 0.75
SVM 0.80
多级SVM 0.85
表4为本申请提供的算法召回率结果比对表，如表4所示，根据多级SVM模型确定的F为0.83，与传统算法性能相当，且多级SVM模型在一定条件下对服务器各硬件设备的分类效果更好。
表4 算法召回率结果比对表
Figure PCTCN2022101794-appb-000014
通过对服务器硬件设备的主动管理，使服务器硬件资源管理更加精细化，便于更早期的发现问题从而做出相应的决策，提升服务器的可靠性，还可以在底层提升服务器性能，保障服务器的稳定性。
基于同样的思路，本说明书实施例还提供了上述方法对应的装置，图4为本申请实施例提供的硬件设备维护装置的结构示意图，如图4所示，本实施例提供的装置，可以包括：
获取模块401，用于获取状态信息，其中，所述状态信息为待维护服务器中各硬件设备对应的信息。
处理模块402，用于将所述状态信息输入至训练完成的多级支持向量机SVM模型中，使得所述多级SVM模型对所述状态信息进行多次识别与分类处理，得到包含至少一种故 障类型的故障信息集。
在本实施例中，所述处理模块402，还用于：
将所述状态信息输入至训练完成的多级SVM模型中，使得所述多级SVM模型中的第一级SVM模块对所述状态信息进行识别与分类处理，得到第一故障类型信息以及其他故障类型信息。
若所述其他故障类型信息中包含的故障类型种类大于一种，则将所述其他故障类型信息输入至所述多级SVM模型中的第二级SVM模块中进行识别与分类处理，得到第二故障类型信息以及新的其他故障类型信息。
若所述新的其他故障类型信息中包含的故障类型种类仍大于一种，则重新执行所述将所述其他故障类型信息输入至所述多级SVM模型中的第二级SVM模块中进行识别与分类处理及之后的步骤直至所述新的其他故障类型信息中包含的故障类型为一种，得到包含至少一种故障类型的故障信息集。
进一步的，所述处理模块402，还用于：
根据所述状态信息以及所述包含至少一种故障类型的故障信息集，得到非故障信息集。
将所述非故障信息集输入至所述多级SVM模型中，使得所述多级SVM模型对所述非故障信息集进行识别与分类处理，得到包含至少一个风险性等级的等级信息集。
所述处理模块402，还用于：
将所述非故障信息集输入至所述多级SVM模型中，使得所述多级SVM模型中的第三级SVM模块对所述非故障信息集中的各非故障信息进行识别与分类处理，得到第一等级信息以及其他等级信息。
若所述其他等级信息中包含的风险性等级数量大于一种，则将所述其他等级信息输入至所述多级SVM模型中的第四级SVM模块中进行识别与分类处理，得到第二等级信息以及新的其他等级信息。
若所述新的其他等级信息中包含的风险性等级数量仍大于一种，则重新执行所述将所述其他等级信息输入至所述多级SVM模型中的第四级SVM模块中进行识别与分类处理及之后的步骤直至所述新的其他等级信息中包含的风险性等级数量为一种，得到包含至少一个风险性等级的等级信息集。
所述处理模块402，还用于根据所述包含至少一种故障类型的故障信息集对所述待维护服务器中的各硬件设备进行故障维护。
此外，在另一实施例中，所述处理模块402，还用于：
获取状态训练信息集，其中，所述状态训练信息集中包含多组状态训练信息，每组所 述状态训练信息中包含硬件设备的状态信息，以及所述硬件设备的状态信息对应的目标故障类型。
根据每组所述状态训练信息中包含的硬件设备的状态信息，以及所述硬件设备的状态信息对应的目标故障类型对初始多级SVM模型进行训练，得到训练完成的多级SVM模型。
此外，所述处理模块402，还用于：
获取状态测试信息集。
将所述状态测试信息集输入至训练完成的多级SVM模型中进行多次识别与分类处理，得到故障测试信息集。
根据所述故障测试信息集对所述多级SVM模型进行准确率评估，并在所述准确率评估的结果符合预设评估条件时，生成多级SVM模型测试通过提示。
具体的，所述处理模块402，还用于：
根据表达式：Accuracy＝Avg(Acc i)确定所述多级SVM模型的分类准确率，
其中，Acc i表示所述多级SVM模型中每级SVM模块的分类准确率，
Figure PCTCN2022101794-appb-000015
T p1表示属于目标故障类型且被正确分类到目标故障类型的样本数量，F pp表示不属于目标故障类型却被错误归到目标故障类型的样本数量，Fnn表示属于目标故障类型却被错误归为其他故障类型的样本数量，T n1表示不属于目标故障类型且被正确分类的样本数量；α i表示各类故障类型的准确因子系数，α i＝1-X*X T，X表示所述状态测试信息集中状态测试信息对应的矩阵，X T表示X的转置，i表示故障类型的个数。
和/或，根据表达式：
Figure PCTCN2022101794-appb-000016
确定所述多级SVM模型的召回率。
其中，F表示所述多级SVM模型的召回率，β表示被分为目标故障类型的样本中所述目标故障类型对应的目标样本所占的比率，β＝max{α 1，α 2，...α n}，
Figure PCTCN2022101794-appb-000017
Figure PCTCN2022101794-appb-000018
此外，所述处理模块402，还用于：
在所述多级SVM模型的分类准确率超过预设准确率阈值，和/或所述多级SVM模型的召回率超过预设召回率阈值时，生成多级SVM模型测试通过提示。
本申请实施例提供的装置，可以实现上述如图2所示的实施例的方法，其实现原理和技术效果类似，此处不再赘述。
图5为本申请实施例提供的电子设备的硬件结构示意图，如图5所示，本实施例提供的设备500包括：处理器501，以及与所述处理器通信连接的存储器。其中，处理器501、存储器502通过总线503连接。
在具体实现过程中，处理器501执行所述存储器502存储的计算机执行指令，使得处理器501执行上述方法实施例中的硬件设备维护方法。
处理器501的具体实现过程可参见上述方法实施例，其实现原理和技术效果类似，本实施例此处不再赘述。
在上述的图5所示的实施例中，应理解，处理器可以是中央处理单元(英文：Central Processing Unit，简称：CPU)，还可以是其他通用处理器、数字信号处理器(英文：Digital Signal Processor，简称：DSP)、专用集成电路(英文：Application Specific Integrated Circuit，简称：ASIC)等。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。结合发明所公开的方法的步骤可以直接体现为硬件处理器执行完成，或者用处理器中的硬件及软件模块组合执行完成。
存储器可能包含高速RAM存储器，也可能还包括非易失性存储NVM，例如至少一个磁盘存储器。
总线可以是工业标准体系结构(Industry Standard Architecture，ISA)总线、外部设备互连(Peripheral Component Interconnect，PCI)总线或扩展工业标准体系结构(Extended Industry Standard Architecture，EISA)总线等。总线可以分为地址总线、数据总线、控制总线等。为便于表示，本申请附图中的总线并不限定仅有一根总线或一种类型的总线。
本申请实施例还提供一种计算机可读存储介质，所述计算机可读存储介质中存储有计算机执行指令，当处理器执行所述计算机执行指令时，实现上述方法实施例的硬件设备维护方法。
本申请实施例还提供一种计算机程序产品，包括计算机程序，所述计算机程序被处理器执行时，实现如上所述的硬件设备维护方法。
上述的计算机可读存储介质，上述可读存储介质可以是由任何类型的易失性或非易失性存储设备或者它们的组合实现，如静态随机存取存储器(SRAM)，电可擦除可编程只读存储器(EEPROM)，可擦除可编程只读存储器(EPROM)，可编程只读存储器(PROM)，只读存储器(ROM)，磁存储器，快闪存储器，磁盘或光盘。可读存储介质可以是通用或专用计算机能够存取的任何可用介质。
一种示例性的可读存储介质耦合至处理器，从而使处理器能够从该可读存储介质读取信息，且可向该可读存储介质写入信息。当然，可读存储介质也可以是处理器的组成部分。处理器和可读存储介质可以位于专用集成电路(Application Specific Integrated Circuits，简称：ASIC)中。当然，处理器和可读存储介质也可以作为分立组件存在于设备中。
本领域普通技术人员可以理解：实现上述各方法实施例的全部或部分步骤可以通过程 序指令相关的硬件来完成。前述的程序可以存储于一计算机可读取存储介质中。该程序在执行时，执行包括上述各方法实施例的步骤；而前述的存储介质包括：ROM、RAM、磁碟或者光盘等各种可以存储程序代码的介质。
最后应说明的是：以上各实施例仅用以说明本申请的技术方案，而非对其限制；尽管参照前述各实施例对本申请进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分或者全部技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本申请各实施例技术方案的范围。

Claims (12)

  1. 一种硬件设备维护方法，其特征在于，包括：
    获取状态信息，其中，所述状态信息为待维护服务器中各硬件设备对应的信息；
    将所述状态信息输入至训练完成的多级支持向量机SVM模型中，使得所述多级SVM模型对所述状态信息进行多次识别与分类处理，得到包含至少一种故障类型的故障信息集；
    根据所述包含至少一种故障类型的故障信息集对所述待维护服务器中的各硬件设备进行故障维护。
  2. 根据权利要求1所述的方法，其特征在于，所述将所述状态信息输入至训练完成的多级SVM模型中，使得所述多级SVM模型对所述状态信息进行多次识别与分类处理，得到包含至少一种故障类型的故障信息集，包括：
    将所述状态信息输入至训练完成的多级SVM模型中，使得所述多级SVM模型中的第一级SVM模块对所述状态信息进行识别与分类处理，得到第一故障类型信息以及其他故障类型信息；
    若所述其他故障类型信息中包含的故障类型种类大于一种，则将所述其他故障类型信息输入至所述多级SVM模型中的第二级SVM模块中进行识别与分类处理，得到第二故障类型信息以及新的其他故障类型信息；
    若所述新的其他故障类型信息中包含的故障类型种类仍大于一种，则重新执行所述将所述其他故障类型信息输入至所述多级SVM模型中的第二级SVM模块中进行识别与分类处理及之后的步骤直至所述新的其他故障类型信息中包含的故障类型为一种，得到包含至少一种故障类型的故障信息集。
  3. 根据权利要求2所述的方法，其特征在于，在所述得到包含至少一种故障类型的故障信息集之后，还包括：
    根据所述状态信息以及所述包含至少一种故障类型的故障信息集，得到非故障信息集；
    将所述非故障信息集输入至所述多级SVM模型中，使得所述多级SVM模型对所述非故障信息集进行识别与分类处理，得到包含至少一个风险性等级的等级信息集。
  4. 根据权利要求3所述的方法，其特征在于，所述将所述非故障信息集输入至所述多级SVM模型中，使得所述多级SVM模型对所述非故障信息集进行识别与分类处理，得到包含至少一个风险性等级的等级信息集，包括：
    将所述非故障信息集输入至所述多级SVM模型中，使得所述多级SVM模型中的第三级SVM模块对所述非故障信息集中的各非故障信息进行识别与分类处理，得到第一等级信息以及其他等级信息；
    若所述其他等级信息中包含的风险性等级数量大于一种，则将所述其他等级信息输入至所述多级SVM模型中的第四级SVM模块中进行识别与分类处理，得到第二等级信息以及新的其他等级信息；
    若所述新的其他等级信息中包含的风险性等级数量仍大于一种，则重新执行所述将所述其他等级信息输入至所述多级SVM模型中的第四级SVM模块中进行识别与分类处理及之后的步骤直至所述新的其他等级信息中包含的风险性等级数量为一种，得到包含至少一个风险性等级的等级信息集。
  5. 根据权利要求1-4任一项所述的方法，其特征在于，在所述获取状态信息之前，还包括：
    获取状态训练信息集，其中，所述状态训练信息集中包含多组状态训练信息，每 组所述状态训练信息中包含硬件设备的状态信息，以及所述硬件设备的状态信息对应的目标故障类型；
    根据每组所述状态训练信息中包含的硬件设备的状态信息，以及所述硬件设备的状态信息对应的目标故障类型对初始多级SVM模型进行训练，得到训练完成的多级SVM模型。
  6. 根据权利要求5所述的方法，其特征在于，在所述得到训练完成的多级SVM模型之后，还包括：
    获取状态测试信息集；
    将所述状态测试信息集输入至训练完成的多级SVM模型中进行多次识别与分类处理，得到故障测试信息集；
    根据所述故障测试信息集对所述多级SVM模型进行准确率评估，并在所述准确率评估的结果符合预设评估条件时，生成多级SVM模型测试通过提示。
  7. 根据权利要求6所述的方法，其特征在于，所述根据所述故障测试信息集对所述多级SVM模型进行准确率评估，包括：
    根据表达式：
    Accuracy＝Avg(Acc i)确定所述多级SVM模型的分类准确率，
    其中，Acc i表示所述多级SVM模型中每级SVM模块的分类准确率，
    Figure PCTCN2022101794-appb-100001
    其中，T p1表示属于目标故障类型且被正确分类到目标故障类型的样本数量，F pp表示不属于目标故障类型却被错误归到目标故障类型的样本数量，Fnn表示属于目标故障类型却被错误归为其他故障类型的样本数量，T n1表示不属于目标故障类型且被正确分类的样本数量，α i表示各类故障类型的准确因子系数，α i＝1-X*X T，X表示所述状态测试信息集中状态测试信息对应的矩阵，X T表示X的转置，i表示故障类型的个数；
    和/或，
    根据表达式：
    Figure PCTCN2022101794-appb-100002
    确定所述多级SVM模型的召回率；
    其中，F表示所述多级SVM模型的召回率，β表示被分为目标故障类型的样本中所述目标故障类型对应的目标样本所占的比率，β＝max{α 1，α 2，...α i}，
    Figure PCTCN2022101794-appb-100003
    Figure PCTCN2022101794-appb-100004
  8. 根据权利要求7所述的方法，其特征在于，所述在所述准确率评估的结果符合预设评估条件时，生成多级SVM模型测试通过提示，包括：
    在所述多级SVM模型的分类准确率超过预设准确率阈值，和/或所述多级SVM模型的召回率超过预设召回率阈值时，生成多级SVM模型测试通过提示。
  9. 一种硬件设备维护装置，其特征在于，包括：
    获取模块，用于获取状态信息，其中，所述状态信息为待维护服务器中各硬件设备对应的信息；
    处理模块，用于将所述状态信息输入至训练完成的多级支持向量机SVM模型中，使得所述多级SVM模型对所述状态信息进行多次识别与分类处理，得到包含至少一种故障类型的故障信息集；
    所述处理模块，还用于根据所述包含至少一种故障类型的故障信息集对所述待维护服务器中的各硬件设备进行故障维护。
  10. 一种电子设备，其特征在于，包括：处理器，以及与所述处理器通信连接的存储器；
    所述存储器存储计算机执行指令；
    所述处理器执行所述存储器存储的计算机执行指令，以实现如权利要求1至8任一项所述的硬件设备维护方法。
  11. 一种计算机可读存储介质，其特征在于，所述计算机可读存储介质中存储有计算机执行指令，当处理器执行所述计算机执行指令时，实现如权利要求1至8任一项所述的硬件设备维护方法。
  12. 一种计算机程序产品，包括计算机程序，其特征在于，所述计算机程序被处理器执行时实现如权利要求1至8任一项所述的硬件设备维护方法。
PCT/CN2022/101794 2021-12-24 2022-06-28 硬件设备维护方法、装置及电子设备 WO2023115875A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202111603721.8A CN114298209A (zh) 2021-12-24 2021-12-24 硬件设备维护方法、装置及电子设备
CN202111603721.8 2021-12-24

Publications (1)

Publication Number Publication Date
WO2023115875A1 true WO2023115875A1 (zh) 2023-06-29

Family

ID=80968764

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/101794 WO2023115875A1 (zh) 2021-12-24 2022-06-28 硬件设备维护方法、装置及电子设备

Country Status (2)

Country Link
CN (1) CN114298209A (zh)
WO (1) WO2023115875A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116703284A (zh) * 2023-08-03 2023-09-05 八爪鱼人工智能科技(常熟)有限公司 应用于冷库管理系统的故障识别方法及人工智能服务器
CN118540208A (zh) * 2024-07-19 2024-08-23 成都凌亚科技有限公司 一种基于srio总线的诊断及修正方法

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114298209A (zh) * 2021-12-24 2022-04-08 深圳前海微众银行股份有限公司 硬件设备维护方法、装置及电子设备

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107563451A (zh) * 2017-09-18 2018-01-09 河海大学 一种泵站稳态工况下运行状态识别方法
CN107644127A (zh) * 2017-09-08 2018-01-30 西安工程大学 基于imabc优化支持向量机的变压器故障诊断方法
CN110717505A (zh) * 2019-08-20 2020-01-21 河海大学 一种基于支持向量机的泵站故障分类方法
CN111008648A (zh) * 2019-11-12 2020-04-14 国网湖南省电力有限公司 一种纯光纤电子式电流互感器故障辨识方法、系统及介质
WO2020260649A1 (fr) * 2019-06-26 2020-12-30 Thales Dispositif radar pour la détection d'un comportement de référence de cibles pistées; procédé et produit programme d'ordinateur associés
CN114298209A (zh) * 2021-12-24 2022-04-08 深圳前海微众银行股份有限公司 硬件设备维护方法、装置及电子设备

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107644127A (zh) * 2017-09-08 2018-01-30 西安工程大学 基于imabc优化支持向量机的变压器故障诊断方法
CN107563451A (zh) * 2017-09-18 2018-01-09 河海大学 一种泵站稳态工况下运行状态识别方法
WO2020260649A1 (fr) * 2019-06-26 2020-12-30 Thales Dispositif radar pour la détection d'un comportement de référence de cibles pistées; procédé et produit programme d'ordinateur associés
CN110717505A (zh) * 2019-08-20 2020-01-21 河海大学 一种基于支持向量机的泵站故障分类方法
CN111008648A (zh) * 2019-11-12 2020-04-14 国网湖南省电力有限公司 一种纯光纤电子式电流互感器故障辨识方法、系统及介质
CN114298209A (zh) * 2021-12-24 2022-04-08 深圳前海微众银行股份有限公司 硬件设备维护方法、装置及电子设备

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
SHANG QIANMING, YANG YE; WANG XIAO; CAO ZHAO; DENG XIAOGUANG: "Application of SVM-based 2PTMC Multi-classification Algorithm in Fault Diagnosis of Marine Diesel Engine", CHINA SHIPREPAIR, vol. 32, no. 2, 1 April 2019 (2019-04-01), pages 30 - 33, XP093074023, ISSN: 1001-8328, DOI: 10.13352/j.issn.1001-8328.2019.02.009 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116703284A (zh) * 2023-08-03 2023-09-05 八爪鱼人工智能科技(常熟)有限公司 应用于冷库管理系统的故障识别方法及人工智能服务器
CN116703284B (zh) * 2023-08-03 2023-10-17 八爪鱼人工智能科技(常熟)有限公司 应用于冷库管理系统的故障识别方法及人工智能服务器
CN118540208A (zh) * 2024-07-19 2024-08-23 成都凌亚科技有限公司 一种基于srio总线的诊断及修正方法

Also Published As

Publication number Publication date
CN114298209A (zh) 2022-04-08

Similar Documents

Publication Publication Date Title
WO2023115875A1 (zh) 硬件设备维护方法、装置及电子设备
US20220147405A1 (en) Automatically scalable system for serverless hyperparameter tuning
WO2019238109A1 (zh) 一种故障根因分析的方法及装置
CN104021264B (zh) 一种缺陷预测方法及装置
WO2020147317A1 (zh) 一种网络异常行为确定方法、装置、设备及可读存储介质
US10789225B2 (en) Column weight calculation for data deduplication
WO2022227388A1 (zh) 日志异常检测模型训练方法、装置及设备
US11860721B2 (en) Utilizing automatic labelling, prioritizing, and root cause analysis machine learning models and dependency graphs to determine recommendations for software products
CN110991474A (zh) 一种机器学习建模平台
WO2023093100A1 (zh) 一种api网关异常调用识别的方法、装置、设备及产品
WO2020082734A1 (zh) 文本情感识别方法、装置、电子设备及计算机非易失性可读存储介质
WO2023056723A1 (zh) 故障诊断的方法、装置、电子设备及存储介质
CN108600000A (zh) 一种故障预测方法、服务器和计算机存储介质
WO2023116111A1 (zh) 一种磁盘故障预测方法及装置
CN112685324B (zh) 一种生成测试方案的方法及系统
CN114298176A (zh) 一种欺诈用户检测方法、装置、介质及电子设备
WO2022231770A1 (en) Automatic triaging of diagnostics failures
CN111327480B (zh) 移动边缘环境下的Web服务多元QoS监控方法
CN113704389A (zh) 一种数据评估方法、装置、计算机设备及存储介质
CN112949697A (zh) 一种管道异常的确认方法、装置及计算机可读存储介质
CN115907972B (zh) 基于双重自步学习的不平衡征信数据风险评估方法及系统
CN111783883A (zh) 一种异常数据的检测方法及装置
WO2022098729A1 (en) Technologies for unsupervised data classification with topological methods
CN113641823A (zh) 文本分类模型训练、文本分类方法、装置、设备及介质
CN115048345A (zh) 异常日志检测方法、装置、电子设备和存储介质

Legal Events

Date Code Title Description
NENP Non-entry into the national phase

Ref country code: DE