Fault prediction method, computing device, and computer program product

Abstract

Embodiments of the present application provide a fault prediction method, a computing device, and a computer program product. The method is applied to the computing device, and includes: obtaining a first fault prediction model trained based on first sample data of a source object, the number of the first sample data being greater than or equal to a preset number; obtaining a plurality of initial sample data of a target object, the initial sample data including historical physical operation parameters of the target object and fault information corresponding to a historical physical operation parameter set, the target object and the source object being hardware of different models; determining second sample data corresponding to each initial sample data according to a feature mapping rule between the target object and the source object, the number of the second sample data being less than the preset number; training a second fault prediction model according to the first fault prediction model and the second sample data; and performing fault prediction on the target object through the second fault prediction model. The above method can improve the fault prediction accuracy.

Description

Technical Field

[0001] This application relates to the field of fault handling technology, and in particular to a fault prediction method, computing device, and computer program product. Background Technology

[0002] Fault prediction models trained based on historical operational data of an object can be used to predict object failures in the future. However, if the amount of historical operational data is insufficient, it is difficult to train a fault prediction model, resulting in low accuracy in fault prediction. Summary of the Invention

[0003] This application provides a fault prediction method, a computing device, and a computer program product to solve the technical problem of low accuracy in fault prediction.

[0004] In a first aspect, embodiments of this application provide a fault prediction method applied to a computing device, the method comprising:

[0005] Obtain the first fault prediction model. The first fault prediction model is trained based on the first sample data of the source object. The number of the first sample data is greater than or equal to the preset number.

[0006] Acquire multiple initial sample data of the target object. The initial sample data includes the historical physical operating parameters of the target object and the fault information corresponding to the historical physical operating parameter set. The target object and the source object are hardware of different models.

[0007] Based on the feature mapping rules between the target object and the source object, the second sample data corresponding to each initial sample data in multiple initial sample data is determined. The number of initial sample data and the number of second sample data are the same, and the number of second sample data is less than the preset number.

[0008] A second fault prediction model is trained based on the first fault prediction model and the second sample data.

[0009] The second fault prediction model is used to predict faults in the target object.

[0010] In the above scheme, a first fault prediction model can be obtained; multiple initial sample data of the target object can be obtained; second sample data corresponding to each initial sample data in the multiple initial sample data can be determined according to the feature mapping rules between the target object and the source object; a second fault prediction model of the target object can be trained based on the first fault prediction model and the second sample data; and fault prediction of the target object can be performed using the second fault prediction model. In the above scheme, the second fault prediction model of the target object can be trained based on the first fault prediction model of the source object, and fault prediction of the target object can be performed using the second fault prediction model, thus improving the accuracy of fault prediction.

[0011] In one possible implementation, before determining the second sample data corresponding to each initial sample data in multiple initial sample data according to the feature mapping rules between the target object and the source object, the method further includes:

[0012] Obtain the first running data and the first fault data of the source object. The first running data includes the first value range of multiple physical running parameters of the source object. The first fault data includes multiple first fault identifiers of the source object and the descriptive information of each first fault identifier. The first fault identifier is used to indicate the fault type, fault location, or fault cause of the source object.

[0013] The second operating data and the second fault data of the target object are obtained. The second operating data includes the second value range of multiple physical operating parameters of the target object. The second fault data includes multiple second fault identifiers of the target object and descriptive information of each second fault identifier. The second fault identifier is used to indicate the fault type, fault location, or fault cause of the target object.

[0014] Based on the first and second operating data, the mapping relationship of physical operating parameters is determined;

[0015] Based on the first fault data and the second fault data, determine the mapping relationship of fault information;

[0016] Based on the mapping relationship of physical operating parameters and the mapping relationship of fault information, feature mapping rules are determined.

[0017] In the above scheme, the feature mapping rules between the target object and the source object can be determined based on the first running data of the source object and the second running data of the target object, thus achieving the purpose of determining the feature mapping rules.

[0018] In one possible implementation, the mapping relationship of physical operating parameters is determined based on the first operating data and the second operating data, including:

[0019] For any physical operating parameter, the mapping rule corresponding to the physical operating parameter is determined based on the endpoint values ​​of the second value range corresponding to the physical operating parameter and the endpoint values ​​of the first value range corresponding to the physical operating parameter.

[0020] Determining the mapping relationship of physical operating parameters includes mapping rules corresponding to multiple physical operating parameters.

[0021] The above scheme can determine the mapping relationship of physical operating parameters, thus achieving the goal of determining the mapping relationship of physical operating parameters.

[0022] In one possible implementation, determining the mapping relationship of fault information based on the first fault data and the second fault data includes:

[0023] For any second fault identifier, target description information is determined in the first fault data based on the description information of the second fault identifier. The similarity between the description information of the second fault identifier and the target description information is greater than or equal to a preset threshold. A mapping relationship is established between the second fault identifier and the first fault identifier corresponding to the target description information.

[0024] Determining the mapping relationship of fault information includes the mapping relationship between multiple second fault identifiers and their corresponding first fault identifiers.

[0025] The above scheme can determine the mapping relationship of fault information, thus achieving the goal of determining the mapping relationship of fault information.

[0026] In one possible implementation, the first fault prediction model is a random forest model; based on the first fault prediction model and the second sample data, a second fault prediction model for the target object is trained, including:

[0027] The feature vector of the second sample data is obtained through the first fault prediction model. The feature vector of the second sample data is used to indicate the node path, leaf node index or node output value of the second sample data on each decision tree of the first fault prediction model.

[0028] The second fault prediction model is trained based on the feature vectors of the second sample data.

[0029] In the above scheme, when the first fault prediction model is a random forest model, the second fault prediction model can be trained based on the first fault prediction model and the second sample data, thus achieving the purpose of training the second fault prediction model.

[0030] In one possible implementation, the first fault prediction model is a neural network model; based on the first fault prediction model and the second sample data, a second fault prediction model for the target object is trained, including:

[0031] An initial fault prediction model is constructed based on the first fault prediction model. The initial fault prediction model includes a trainable output layer.

[0032] Based on the second sample data, the output layer of the initial fault prediction model is trained to obtain the second fault prediction model.

[0033] In the above scheme, when the first fault prediction model is a neural network model, the second fault prediction model can be trained based on the first fault prediction model and the second sample data, thus achieving the purpose of training the second fault prediction model.

[0034] In one possible implementation, an initial fault prediction model is constructed based on the first fault prediction model, including:

[0035] Freeze the feature extraction layer of the first fault prediction model;

[0036] Delete the output layer of the first fault prediction model;

[0037] An output layer is added after the feature extraction layer of the first fault prediction model to obtain the initial fault prediction model.

[0038] In the above scheme, when the first fault prediction model is a neural network model, the initial fault prediction model can be determined, thus achieving the purpose of determining the initial fault prediction model.

[0039] Secondly, embodiments of this application provide a fault prediction device applied to a computing device. The fault prediction device includes an acquisition module, a determination module, a training module, and a prediction module, wherein:

[0040] The acquisition module is used to acquire a first fault prediction model, which is trained based on the first sample data of the source object, and the number of the first sample data is greater than or equal to a preset number.

[0041] The acquisition module is also used to acquire multiple initial sample data of the target object. The initial sample data includes the historical physical operating parameters of the target object and the fault information corresponding to the historical physical operating parameter set. The target object and the source object are hardware of different models.

[0042] The determination module is used to determine the second sample data corresponding to each initial sample data in multiple initial sample data according to the feature mapping rules between the target object and the source object. The number of initial sample data and the number of second sample data are the same, and the number of second sample data is less than a preset number.

[0043] The training module is used to train a second fault prediction model based on the first fault prediction model and the second sample data.

[0044] The prediction module is used to predict the faults of the target object using a second fault prediction model.

[0045] In the above scheme, a first fault prediction model can be obtained; multiple initial sample data of the target object can be obtained; second sample data corresponding to each initial sample data in the multiple initial sample data can be determined according to the feature mapping rules between the target object and the source object; a second fault prediction model of the target object can be trained based on the first fault prediction model and the second sample data; and fault prediction of the target object can be performed using the second fault prediction model. In the above scheme, the second fault prediction model of the target object can be trained based on the first fault prediction model of the source object, and fault prediction of the target object can be performed using the second fault prediction model, thus improving the accuracy of fault prediction.

[0046] In one possible implementation, the acquisition module is further configured to acquire first running data and first fault data of the source object. The first running data includes a first value range of multiple physical running parameters of the source object. The first fault data includes multiple first fault identifiers of the source object and descriptive information of each first fault identifier. The first fault identifier is used to indicate the fault type, fault location, or fault cause of the source object.

[0047] The acquisition module is also used to acquire second operating data and second fault data of the target object. The second operating data includes the second value range of multiple physical operating parameters of the target object. The second fault data includes multiple second fault identifiers of the target object and descriptive information of each second fault identifier. The second fault identifier is used to indicate the fault type, fault location, or fault cause of the target object.

[0048] The determining module is also used to determine the mapping relationship of physical operating parameters based on the first operating data and the second operating data;

[0049] The determination module is also used to determine the mapping relationship of fault information based on the first fault data and the second fault data;

[0050] The determination module is also used to determine feature mapping rules based on the mapping relationship of physical operating parameters and the mapping relationship of fault information.

[0051] In the above scheme, the feature mapping rules between the target object and the source object can be determined based on the first running data of the source object and the second running data of the target object, thus achieving the purpose of determining the feature mapping rules.

[0052] In one possible implementation, the determination module is specifically used for:

[0053] For any physical operating parameter, the mapping rule corresponding to the physical operating parameter is determined based on the endpoint values ​​of the second value range corresponding to the physical operating parameter and the endpoint values ​​of the first value range corresponding to the physical operating parameter.

[0054] Determining the mapping relationship of physical operating parameters includes mapping rules corresponding to multiple physical operating parameters.

[0055] The above scheme can determine the mapping relationship of physical operating parameters, thus achieving the goal of determining the mapping relationship of physical operating parameters.

[0056] In one possible implementation, the determination module is specifically used for:

[0057] For any second fault identifier, target description information is determined in the first fault data based on the description information of the second fault identifier. The similarity between the description information of the second fault identifier and the target description information is greater than or equal to a preset threshold. A mapping relationship is established between the second fault identifier and the first fault identifier corresponding to the target description information.

[0058] Determining the mapping relationship of fault information includes the mapping relationship between multiple second fault identifiers and their corresponding first fault identifiers.

[0059] The above scheme can determine the mapping relationship of fault information, thus achieving the goal of determining the mapping relationship of fault information.

[0060] In one possible implementation, the first fault prediction model is a random forest model; the training module is specifically used for:

[0061] The feature vector of the second sample data is obtained through the first fault prediction model. The feature vector of the second sample data is used to indicate the node path, leaf node index or node output value of the second sample data on each decision tree of the first fault prediction model.

[0062] The second fault prediction model is trained based on the feature vectors of the second sample data.

[0063] In the above scheme, when the first fault prediction model is a random forest model, the second fault prediction model can be trained based on the first fault prediction model and the second sample data, thus achieving the purpose of training the second fault prediction model.

[0064] In one possible implementation, the first fault prediction model is a neural network model; the training module is specifically used for:

[0065] An initial fault prediction model is constructed based on the first fault prediction model. The initial fault prediction model includes a trainable output layer.

[0066] Based on the second sample data, the output layer of the initial fault prediction model is trained to obtain the second fault prediction model.

[0067] In the above scheme, when the first fault prediction model is a neural network model, the second fault prediction model can be trained based on the first fault prediction model and the second sample data, thus achieving the purpose of training the second fault prediction model.

[0068] In one possible implementation, the training module is specifically used for:

[0069] Freeze the feature extraction layer of the first fault prediction model;

[0070] Delete the output layer of the first fault prediction model;

[0071] An output layer is added after the feature extraction layer of the first fault prediction model to obtain the initial fault prediction model.

[0072] In the above scheme, when the first fault prediction model is a neural network model, the initial fault prediction model can be determined, thus achieving the purpose of determining the initial fault prediction model.

[0073] Thirdly, embodiments of this application provide a computing device, including: a memory and a processor;

[0074] The memory stores instructions that the computer executes;

[0075] The processor executes computer execution instructions stored in memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0076] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0077] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect. Attached Figure Description

[0078] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0079] Figure 1 This is a schematic diagram of an application scenario provided by an embodiment of this application;

[0080] Figure 2 A flowchart illustrating a fault prediction method provided in an embodiment of this application;

[0081] Figure 3 A flowchart illustrating another fault prediction method provided in an embodiment of this application;

[0082] Figure 4 A flowchart illustrating another fault prediction method provided in an embodiment of this application;

[0083] Figure 5 This is a schematic diagram of the structure of a fault prediction device provided in an embodiment of this application;

[0084] Figure 6 This is a schematic diagram of the structure of a computing device provided in an embodiment of this application. Detailed Implementation

[0085] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those of this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.

[0086] This application provides a fault prediction method, a computing device, and a computer program product to solve the technical problem of low fault prediction accuracy.

[0087] The following is combined Figure 1 The application scenarios involved in the embodiments of this application will be described.

[0088] Figure 1 This is a schematic diagram illustrating an application scenario provided by an embodiment of this application. Please refer to [link / reference]. Figure 1 By inputting the physical operation data of an object into the object's fault prediction model, fault information of the object can be obtained.

[0089] Fault prediction model: A computational model trained using machine learning or statistical analysis methods based on an object's historical operating data, used to infer fault information about the object in the future. For example, historical operating data can include the object's physical operating data over past periods, as well as fault information.

[0090] For example, the object can be a central processing unit (CPU), a graphics processing unit (GPU), or a neural processing unit (NPU).

[0091] For example, physical operation data could include the temperature, voltage, utilization rate, or power of the object during its operation.

[0092] For example, fault information about an object can be used to indicate the fault type, fault location, and fault cause.

[0093] It should be noted that the more historical operational data used when training the fault prediction model for an object, the higher the accuracy of the fault prediction model. However, for objects with a short service life, the amount of historical operational data is small, making it difficult to train a fault prediction model based on this data, thus resulting in lower fault prediction accuracy.

[0094] The technical solutions of the embodiments of this application will be described in detail below with specific examples. These specific examples can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0095] Figure 2 This is a flowchart illustrating a fault prediction method provided in an embodiment of this application. The execution subject of this method can be a computing device, or a processor or processing chip located within the computing device. For ease of understanding, the following description uses a computing device as the execution subject. Please refer to... Figure 2 The method may include:

[0096] S201. Obtain the first fault prediction model.

[0097] The first fault prediction model is trained based on the first sample data of the source object, and the number of the first sample data is greater than or equal to the preset number.

[0098] The specific size of the preset quantity can be set according to actual needs, as long as it can meet the training requirements of the first fault prediction model and the prediction accuracy of the first fault prediction model reaches the preset accuracy.

[0099] The first sample data can be the historical operational data of the source object. For example, the first sample data may include the historical physical operational parameters of the source object, as well as historical fault information.

[0100] The source object can be an object that has at least a preset number of first sample data.

[0101] In some embodiments, the source object can be a CPU with a large number of first sample data. For example, the source object can be an Intel CPU.

[0102] In this embodiment, the first fault prediction model can be a fault prediction model that has already been trained. The computing device can obtain the first fault prediction model from other electronic devices, or the computing device can obtain the first fault prediction model locally.

[0103] In some embodiments, other electronic devices can train a first fault prediction model based on the first sample data, and can store the first fault prediction model on other electronic devices or computing devices. In this way, the computing device can obtain the first fault prediction model from other electronic devices or locally.

[0104] In some embodiments, the computing device can train a first fault prediction model based on the first sample data, and can store the first fault prediction model in other electronic devices or the computing device. In this way, the computing device can obtain the first fault prediction model from other electronic devices or locally.

[0105] S202. Obtain multiple initial sample data of the target object.

[0106] The target object and the source object can be hardware of different models.

[0107] In some embodiments, the target object and the source object can be the same type of hardware on different platforms, such as a CPU, GPU, or NPU. For example, the source object can be a CPU on a first platform, and the target object can be a CPU on a second platform. The first platform can be, for example, Intel, and the second platform can be, for example, Advanced Micro Devices (AMD). It should be understood that the CPUs on different platforms may have different model numbers.

[0108] In some embodiments, the target object and the source object can be different models of CPUs, GPUs, or NPUs on the same platform. For example, the source object can be a model of CPU on a first platform, and the target object can be a different model of CPU on the first platform.

[0109] Understandably, when the target object and the source object can be the same type of hardware on different platforms, the first sample data can be sample data of the hardware corresponding to the model of the source object on the first platform, or sample data of multiple different models of hardware on the same platform, or sample data of all models of hardware on the same platform. The second sample data can be sample data of the hardware corresponding to the model of the target object on the second platform, or sample data of multiple different models of hardware on the same platform, or sample data of all models of hardware on the same platform.

[0110] In addition, the amount of historical running data for the target object is less than the preset amount.

[0111] The initial sample data can be the historical operational data of the target object. For example, the initial sample data may include the historical physical operating parameters of the target object, as well as historical fault information.

[0112] In this embodiment, the electronic device can obtain multiple initial sample data of the target object from other electronic devices, or it can obtain multiple initial sample data of the target object from local storage.

[0113] S203. Based on the feature mapping rules between the target object and the source object, determine the second sample data corresponding to each initial sample data in the multiple initial sample data.

[0114] Feature mapping rules can include mapping relationships for physical operating parameters and mapping relationships for fault information. The mapping relationship for physical operating parameters is used to transform historical physical operating parameters in the initial sample data, and the mapping relationship for fault information is used to transform historical fault information in the initial sample data.

[0115] The second sample data can be the data obtained by transforming the initial sample data according to the feature mapping rules.

[0116] Specifically, for the historical physical operating parameters in the initial sample data, the computing device can map them according to the mapping relationship of physical operating parameters; for the historical fault information in the initial sample data, the computing device can map it according to the mapping relationship of fault information. The second sample data corresponding to the initial sample data may include: parameters after mapping the historical physical operating parameters in the initial sample data, and parameters after mapping the historical fault information in the initial sample data.

[0117] The following example uses a CPU on a first platform as the source object and a CPU on a second platform as the target object to illustrate the mapping relationship of physical operating parameters and the mapping relationship of fault information:

[0118] (1) Although the CPUs of the first platform and the second platform are different, they are both processors and have the same physical operating parameters. For example, the physical operating parameters of both can include voltage, temperature, power, frequency and utilization.

[0119] However, for the same physical operating parameter, the unit of that parameter in the CPU of the first platform may differ from the unit of that parameter in the CPU of the second platform; the value range of that physical operating parameter in the CPU of the first platform may also differ from the value range of that physical operating parameter in the CPU of the second platform. The mapping relationship of physical operating parameters can be established by mapping the value of the physical operating parameter in the CPU of the second platform according to the unit and value range of the physical operating parameter in the source object. In other words, the mapping relationship of physical operating parameters can be established by mapping the value of the physical operating parameter in the target object according to the unit and value range of the physical operating parameter in the source object.

[0120] (2) The failures of the CPUs on both the first and second platforms are caused by transistor aging, physical failures, or electrical layer failures. While the reported fault identifiers differ, they represent the same meaning. The fault information mapping relationship can be used to convert the fault identifiers in the CPU of the second platform to those in the CPU of the first platform, based on the meanings they represent. In other words, the fault information mapping relationship can be used to convert the fault identifiers in the target object to those in the source object, based on the meanings they represent.

[0121] It should be noted that the methods for determining the mapping relationships of physical operating parameters and fault information can be found in [reference needed]. Figure 3 Examples are not described here.

[0122] It should be noted that S202-S203 can be executed after S201; or, S202-S203 can be executed simultaneously with S201; or, S202-S203 can be executed before S201. This embodiment does not limit the execution order of S202-S203 and S201.

[0123] S204. Based on the first fault prediction model and the second sample data, train the second fault prediction model of the target object.

[0124] In this embodiment, the first fault prediction model can be a random forest model or a neural network model. Depending on the type of the first fault prediction model, training the second fault prediction model can involve at least the following scenarios:

[0125] Case 1: The first fault prediction model is a random forest model.

[0126] In this case, the feature vector of the second sample data can be obtained through the first fault prediction model. The feature vector of the second sample data is used to indicate the node path, leaf node index or node output value of the second sample data on each decision tree of the first fault prediction model. Based on the feature vector of the second sample data, the second fault prediction model is trained.

[0127] Specifically, for any given second sample data, the second sample data can be input into the first fault prediction model. The node paths, leaf node indices, or node output values ​​on each decision tree can be obtained when the first fault prediction model processes the second sample data. The vectors corresponding to the node paths, leaf node indices, or node output values ​​on each decision tree can then be determined as the feature vectors of the second sample data. Through this method, the feature vectors of each second sample data can be obtained.

[0128] Furthermore, a new random forest model can be trained as a second fault prediction model using the feature vector of each second sample data as input.

[0129] Scenario 2: The first fault prediction model is a neural network model.

[0130] In this case, an initial fault prediction model is constructed based on the first fault prediction model, which includes a trainable output layer; the output layer of the initial fault prediction model is trained according to the second sample data to obtain the second fault prediction model.

[0131] When constructing the initial fault prediction model, the feature extraction layer of the first fault prediction model can be frozen; the output layer of the first fault prediction model can be deleted; and an output layer can be added after the feature extraction layer of the first fault prediction model to obtain the initial fault prediction model. For example, the feature extraction layer can be a convolutional layer, and the newly added output layer can be one or more fully connected layers.

[0132] Specifically, the computing device can freeze the feature extraction layer of the first fault prediction model, delete the output layer of the first fault prediction model, and then add a new output layer to obtain an initial fault prediction model. The computing device can also input second sample data into the initial fault prediction model and train the output layer of the initial fault prediction model to obtain a second fault prediction model.

[0133] In this embodiment, a second fault prediction model can be trained even with a small amount of second sample data, improving the accuracy of fault prediction for the target object. Furthermore, with a small amount of second sample data, the first fault prediction model of the source object can be effectively utilized, resulting in a shorter training cycle for the second fault prediction model.

[0134] S205. Using the second fault prediction model, perform fault prediction on the target object.

[0135] In this embodiment, after training the second fault prediction model, the physical operating parameters of the target object can be obtained in real time, and these parameters can be input into the second fault prediction model so that it can predict faults in the target object based on these parameters. For example, the physical operating parameters of the target object can be obtained through a Baseboard Management Controller (BMC).

[0136] Optionally, the computing device can send fault prediction results to the maintenance device so that the maintenance device can take appropriate fault handling measures in a timely manner based on the fault prediction results.

[0137] In the fault prediction method provided in this embodiment, the computing device can acquire a first fault prediction model; acquire multiple initial sample data of the target object; determine the second sample data corresponding to each initial sample data in the multiple initial sample data according to the feature mapping rules between the target object and the source object; train a second fault prediction model of the target object according to the first fault prediction model and the second sample data; and perform fault prediction on the target object using the second fault prediction model. In the above method, the second fault prediction model of the target object can be trained based on the first fault prediction model of the source object, and fault prediction on the target object can be performed using the second fault prediction model, thus improving the accuracy of fault prediction.

[0138] Based on any of the above embodiments, before determining the second sample data corresponding to each initial sample data in multiple initial sample data according to the feature mapping rules between the target object and the source object, it is necessary to determine the feature mapping rules between the target object and the source object. The following is in conjunction with... Figure 3 The specific methods for determining feature mapping rules are explained.

[0139] Figure 3 This is a flowchart illustrating another fault prediction method provided in an embodiment of this application. The executing entity of this method can be a computing device, or a processor or processing chip located within the computing device. For ease of understanding, the following description uses a computing device as the executing entity. Please refer to... Figure 3 The method may include:

[0140] S301, Obtain the first running data and the first fault data of the source object.

[0141] The first set of runtime data includes the first range of values ​​for multiple physical runtime parameters of the source object.

[0142] In this embodiment, multiple physical operating parameters include, but are not limited to, voltage, temperature, power, frequency, and utilization rate.

[0143] For any given physical operating parameter, the first value range of the physical operating parameter can be the range of fluctuation of the physical operating parameter during the operation of the source object. For example, the first value range of temperature can be 0℃ to 70℃.

[0144] In this embodiment, the first fault data includes multiple first fault identifiers of the source object and descriptive information of each first fault identifier. The first fault identifier is used to indicate the fault type, fault location, or fault cause of the source object.

[0145] The first fault identifier will be explained below, taking the CPU of the first platform as the source object.

[0146] (1) When the first fault identifier is used to indicate the fault type of the CPU of the first platform, the first fault identifier and its description information can be found in Table 1:

[0147] Table 1

[0148]

[0149] It should be noted that Table 1 only provides some examples of CPU failure types of the first platform and does not constitute a limitation on the technical solutions provided in this application.

[0150] (2) When the first fault identifier is used to indicate the fault location of the CPU of the first platform, the first fault identifier can be the name of the Machine Check Architecture Bank (MCA Bank). The first fault identifier and its description can be found in Table 2:

[0151] Table 2

[0152]

[0153] It should be noted that Table 2 only provides some examples of the fault locations of the CPUs of the first platform and does not constitute a limitation on the technical solutions provided in this application.

[0154] (3) When the first fault identifier is used to indicate the cause of a CPU failure on the first platform, the first fault identifier can be a Machine Check Architecture Error Code (MCA Code). The first fault identifier and its description can be found in Table 3:

[0155] Table 3

[0156]

[0157] It should be noted that Table 3 only provides some examples of the causes of CPU failure on the first platform and does not constitute a limitation on the technical solutions provided in this application.

[0158] S302, Obtain the second running data and the second fault data of the target object.

[0159] The second set of operational data includes the second range of values ​​for multiple physical operational parameters of the target object.

[0160] It should be noted that the explanations of multiple physical operating parameters can be found in S301, and will not be repeated here.

[0161] For any given physical operating parameter, the second value range of the physical operating parameter can be the range of fluctuation of the physical operating parameter during the operation of the target object. For example, the second value range of temperature can be from 0℃ to 65℃.

[0162] In this embodiment, the second fault data includes multiple second fault identifiers of the target object and descriptive information of each second fault identifier. The second fault identifiers are used to indicate the fault type, fault location, or fault cause of the target object.

[0163] The second fault identifier will be explained below using the CPU of the second platform as an example.

[0164] (1) When the second fault identifier is used to indicate the fault type of the CPU of the second platform, the second fault identifier and its description information can be found in Table 4:

[0165] Table 4

[0166]

[0167] It should be noted that Table 4 only provides some examples of CPU failure types for the second platform and does not constitute a limitation on the technical solutions provided in this application.

[0168] (2) When the second fault identifier is used to indicate the fault location of the CPU on the second platform, the second fault identifier can be the MCA Bank name. The second fault identifier and its description can be found in Table 5:

[0169] Table 5

[0170]

[0171] It should be noted that Table 5 only provides some examples of the fault locations of the CPUs of the second platform and does not constitute a limitation on the technical solutions provided in this application.

[0172] (3) When the second fault identifier is used to indicate the cause of a CPU failure on the second platform, the second fault identifier can be an MCA Code. The second fault identifier and its description can be found in Table 6:

[0173] Table 6

[0174]

[0175] It should be noted that Table 6 only provides some examples of the causes of CPU failure on the second platform and does not constitute a limitation on the technical solutions provided in this application.

[0176] It should be noted that this embodiment does not limit the execution order of S301 and S302. S301 can be executed before S302; or, S301 can be executed simultaneously with S302; or, S301 can be executed after S302.

[0177] S303. Based on the first operating data and the second operating data, determine the mapping relationship of the physical operating parameters.

[0178] In this embodiment, for any physical operating parameter, the computing device can determine the mapping rule corresponding to the physical operating parameter based on the endpoint values ​​of the second value range corresponding to the physical operating parameter and the endpoint values ​​of the first value range corresponding to the physical operating parameter. Furthermore, the computing device can determine that the mapping relationship of the physical operating parameter includes mapping rules corresponding to multiple physical operating parameters.

[0179] It should be noted that the method for determining the mapping rules corresponding to each physical operating parameter is the same. The following uses any physical operating parameter as an example to explain the method for determining the mapping rules corresponding to the physical operating parameter.

[0180] Assume the first range of values ​​for the physical operating parameters is The second range of values ​​is In other words, the minimum value of the physical operating parameters in the source object can be... The maximum value can be The minimum value of the physical operating parameters in the target object can be The maximum value can be .

[0181] When determining the mapping rules for physical operating parameters, if the units of two value ranges are different, the units of the two value ranges can be unified first. Optionally, the units of the endpoint values ​​of the second value range can be converted to be consistent with the units of the endpoint values ​​of the first value range. In this example, after converting the units of the second value range, the second value range can be... .

[0182] After unifying the units for the two value ranges, the mapping rule corresponding to the physical operating parameters can be expressed by the following formula:

[0183]

[0184] in, It can be the minimum value in the second range of values. It can be the maximum value in the second range. It can be the minimum value in the first range of values. It can be the maximum value in the first range of values. It can be any value within the second range. It can be in the first range of values ​​and The corresponding value.

[0185] S304. Based on the first fault data and the second fault data, determine the mapping relationship of the fault information.

[0186] In this embodiment, for any second fault identifier, the computing device can determine target description information from the first fault data based on the description information of the second fault identifier, wherein the similarity between the description information of the second fault identifier and the target description information is greater than or equal to a preset threshold; and establish a mapping relationship between the second fault identifier and the first fault identifier corresponding to the target description information. Further, the computing device can determine that the mapping relationship of the fault information includes mapping relationships between multiple second fault identifiers and their corresponding first fault identifiers.

[0187] The preset threshold can be set according to actual needs. For example, the preset threshold can be 99%.

[0188] Specifically, for a second fault identifier and a first fault identifier that have the same or similar descriptive information, a mapping relationship can be established between the second fault identifier and the first fault identifier.

[0189] The following example, using the source object as the CPU of the first platform and the target object as the CPU of the second platform, illustrates the mapping relationship between the second fault identifier and the first fault identifier.

[0190] (1) The first fault identifier is used to indicate the fault type of the CPU on the first platform, and the second fault identifier is used to indicate the fault type of the CPU on the second platform. The mapping relationship between the second fault identifier and the first fault identifier can be found in Table 7:

[0191] Table 7

[0192]

[0193] As shown in Table 7, there is a mapping relationship between "Uncorrected Error - Catastrophic / Fatal" and "Uncorrected system fatal error", "Uncorrected Error - SRAR" and "Uncorrected thread fatal error", "Uncorrected Error - SRAO" and "Uncorrected recoverable error", "Uncorrected Error - UCNA" and "Deferred error", and "Corrected Error" and "Corrected Error".

[0194] (2) The first fault identifier is used to indicate the fault location of the CPU on the first platform, and the second fault identifier is used to indicate the fault location of the CPU on the second platform. The mapping relationship between the second fault identifier and the first fault identifier can be found in Table 8:

[0195] Table 8

[0196]

[0197] As shown in Table 8, there is a mapping relationship between "IFU" and "IF", "DCU" and "LS", and "MLC" and "L2".

[0198] (3) The first fault identifier is used to indicate the cause of the CPU failure on the first platform, and the second fault identifier is used to indicate the cause of the CPU failure on the second platform. The mapping relationship between the second fault identifier and the first fault identifier can be found in Table 9:

[0199] Table 9

[0200]

[0201] As shown in Table 9, there is a mapping relationship between "000F 0000 0000 111L" and "0000 0001 0000 111L", "000F 0001 RRRR TTLL" and "0000 0001 RRRR TTLL", and "000F 1PPT RRRRIIIL" and "0000 1PPT RRRR IIIL".

[0202] S305. Determine the feature mapping rules, including the mapping relationship of physical operating parameters and the mapping relationship of fault information.

[0203] In the fault prediction method provided in this embodiment, the computing device can acquire first operating data and first fault data of the source object; it can acquire second operating data and second fault data of the target object; it can determine the mapping relationship of physical operating parameters based on the first and second operating data; it can determine the mapping relationship of fault information based on the first and second fault data; and it can determine the feature mapping rules, including the mapping relationship of physical operating parameters and the mapping relationship of fault information. The above method can achieve the purpose of determining the feature mapping rules between the target object and the source object.

[0204] Based on any of the above embodiments, the following is combined with Figure 4 The fault prediction method provided in the embodiments of this application will be further explained.

[0205] Figure 4 This is a flowchart illustrating another fault prediction method provided in an embodiment of this application. The executing entity of this method can be a computing device, or a processor or processing chip located within the computing device. For ease of understanding, the following description uses a computing device as the executing entity. Please refer to... Figure 4 The method may include:

[0206] S401. Obtain the first fault prediction model.

[0207] It should be noted that the specific implementation of S401 can be found in S201, and will not be repeated here.

[0208] S402. Obtain the feature mapping rules between the target object and the source object.

[0209] In some embodiments, the feature mapping rules are predetermined mapping rules, and the computing device can obtain the feature mapping rules from local or other electronic devices.

[0210] In some embodiments, the computing device may determine feature mapping rules and obtain feature mapping rules locally.

[0211] It should be noted that the method for determining the feature mapping rules can be discussed in [the relevant forum / instructions]. Figure 3 Examples are not described here.

[0212] S403. Obtain multiple initial sample data of the target object.

[0213] S404. Based on the feature mapping rules, determine the second sample data corresponding to each initial sample data in the multiple initial sample data.

[0214] S405. Based on the first fault prediction model and the second sample data, a second fault prediction model for the target object is trained.

[0215] S406. Using the second fault prediction model, perform fault prediction on the target object.

[0216] It should be noted that the specific implementation methods of S403-S406 can be found in S202-S205, and will not be repeated here.

[0217] In the fault prediction method provided in this embodiment, the computing device can acquire a first fault prediction model; acquire feature mapping rules between the target object and the source object; acquire multiple initial sample data of the target object; determine the second sample data corresponding to each initial sample data in the multiple initial sample data according to the feature mapping rules; train a second fault prediction model of the target object according to the first fault prediction model and the second sample data; and perform fault prediction on the target object using the second fault prediction model. In the above method, the second fault prediction model of the target object can be trained based on the first fault prediction model of the source object, and fault prediction on the target object can be performed using the second fault prediction model, thus improving the accuracy of fault prediction.

[0218] Figure 5 This is a schematic diagram of a fault prediction device provided in an embodiment of this application. The fault prediction device 10 can be applied to a computing device. Please refer to... Figure 5 The fault prediction device 10 may include: an acquisition module 11, a determination module 12, a training module 13, and a prediction module 14, wherein:

[0219] The acquisition module 11 is used to acquire a first fault prediction model, which is trained based on the first sample data of the source object, and the number of the first sample data is greater than or equal to a preset number.

[0220] The acquisition module 11 is also used to acquire multiple initial sample data of the target object. The initial sample data includes the historical physical operating parameters of the target object and the fault information corresponding to the historical physical operating parameter set. The target object and the source object are hardware of different models.

[0221] The determining module 12 is used to determine the second sample data corresponding to each initial sample data in multiple initial sample data according to the feature mapping rules between the target object and the source object. The number of initial sample data and the number of second sample data are the same, and the number of second sample data is less than a preset number.

[0222] The training module 13 is used to train a second fault prediction model based on the first fault prediction model and the second sample data.

[0223] The prediction module 14 is used to predict the faults of the target object using the second fault prediction model.

[0224] The fault prediction device provided in this embodiment can execute the fault prediction method executed by the computing device in any of the above method embodiments. Its implementation principle and beneficial effects are similar, and will not be described again here.

[0225] In one possible implementation, the acquisition module 11 is further configured to acquire first running data and first fault data of the source object. The first running data includes a first value range of multiple physical running parameters of the source object. The first fault data includes multiple first fault identifiers of the source object and descriptive information of each first fault identifier. The first fault identifier is used to indicate the fault type, fault location, or fault cause of the source object.

[0226] The acquisition module 11 is also used to acquire second operating data and second fault data of the target object. The second operating data includes the second value range of multiple physical operating parameters of the target object. The second fault data includes multiple second fault identifiers of the target object and descriptive information of each second fault identifier. The second fault identifier is used to indicate the fault type, fault location, or fault cause of the target object.

[0227] The determining module 12 is further configured to determine the mapping relationship of physical operating parameters based on the first operating data and the second operating data;

[0228] The determining module 12 is also used to determine the mapping relationship of fault information based on the first fault data and the second fault data;

[0229] The determination module 12 is also used to determine feature mapping rules based on the mapping relationship of physical operating parameters and the mapping relationship of fault information.

[0230] In one possible implementation, the determining module 12 is specifically used for:

[0231] For any physical operating parameter, the mapping rule corresponding to the physical operating parameter is determined based on the endpoint values ​​of the second value range corresponding to the physical operating parameter and the endpoint values ​​of the first value range corresponding to the physical operating parameter.

[0232] Determining the mapping relationship of physical operating parameters includes mapping rules corresponding to multiple physical operating parameters.

[0233] In one possible implementation, the determining module 12 is specifically used for:

[0234] For any second fault identifier, target description information is determined in the first fault data based on the description information of the second fault identifier. The similarity between the description information of the second fault identifier and the target description information is greater than or equal to a preset threshold. A mapping relationship is established between the second fault identifier and the first fault identifier corresponding to the target description information.

[0235] Determining the mapping relationship of fault information includes the mapping relationship between multiple second fault identifiers and their corresponding first fault identifiers.

[0236] In one possible implementation, the first fault prediction model is a random forest model; the training module 13 is specifically used for:

[0237] The feature vector of the second sample data is obtained through the first fault prediction model. The feature vector of the second sample data is used to indicate the node path, leaf node index or node output value of the second sample data on each decision tree of the first fault prediction model.

[0238] The second fault prediction model is trained based on the feature vectors of the second sample data.

[0239] In one possible implementation, the first fault prediction model is a neural network model; the training module 13 is specifically used for:

[0240] An initial fault prediction model is constructed based on the first fault prediction model. The initial fault prediction model includes a trainable output layer.

[0241] Based on the second sample data, the output layer of the initial fault prediction model is trained to obtain the second fault prediction model.

[0242] In one possible implementation, training module 13 is specifically used for:

[0243] Freeze the feature extraction layer of the first fault prediction model;

[0244] Delete the output layer of the first fault prediction model;

[0245] An output layer is added after the feature extraction layer of the first fault prediction model to obtain the initial fault prediction model.

[0246] The fault prediction device provided in this embodiment can execute the fault prediction method executed by the computing device in any of the above method embodiments. Its implementation principle and beneficial effects are similar, and will not be described again here.

[0247] Figure 6 This is a schematic diagram of the structure of a computing device provided in an embodiment of this application. Figure 6 As shown, the computing device 20 may include a processor 21 and a memory 22, wherein the processor 21 and the memory 22 can communicate; for example, the processor 21 and the memory 22 communicate via a communication bus 23, the memory 22 is used to store computer execution instructions, and the processor 21 is used to call the computer execution instructions in the memory to execute the fault prediction method shown in any of the above method embodiments.

[0248] Optionally, the computing device 20 may also include a communication interface, which may include a transmitter and / or a receiver.

[0249] The computing device 20 can be the computing device shown in any of the above method embodiments, and can execute the fault prediction method shown in any of the above method embodiments.

[0250] Optionally, the processor mentioned above can be a CPU, or a GPU, a Baseboard Management Controller (BMC), other general-purpose processors, a Digital Signal Processor (DSP), or an Application Specific Integrated Circuit (ASIC), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the embodiments of this application can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules in the processor.

[0251] This application provides a computer-readable storage medium storing computer-executable instructions; the computer-executable instructions are used to implement the fault prediction method as described in any of the above embodiments.

[0252] This application provides a computer program product, which includes a computer program that, when executed, causes the computer to perform the aforementioned fault prediction method.

[0253] All or part of the steps in the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a readable memory. When the program is executed, it performs the steps of the above method embodiments; and the aforementioned memory (storage medium) includes: read-only memory (ROM), RAM, flash memory, hard disk, solid-state drive, magnetic tape, floppy disk, optical disk, and any combination thereof.

[0254] This application describes embodiments of methods, apparatus (systems), and computer program products according to embodiments of this application with reference to flowchart illustrations and / or block diagrams. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processing unit of a general-purpose computer, special-purpose computer, embedded processor, or other programmable terminal device to produce a machine, such that the instructions, which execute via the processing unit of the computer or other programmable terminal device, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0255] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable terminal device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0256] These computer program instructions can also be loaded onto a computer or other programmable terminal device, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable device for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0257] Obviously, those skilled in the art can make various modifications and variations to the embodiments of this application without departing from the spirit and scope of the embodiments of this application. Therefore, if these modifications and variations to the embodiments of this application fall within the scope of the claims of this application and their equivalents, the embodiments of this application are also intended to include these modifications and variations.

[0258] In the embodiments of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. The term "or" and its variations can mean "and / or." In the embodiments of this application, the terms "first," "second," etc., are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. In the embodiments of this application, "multiple" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0259] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the invention disclosed in the specification and in practice. The embodiments of this application are intended to cover any variations, uses, or adaptations of the embodiments of this application that follow the general principles of the embodiments of this application and include common knowledge or customary technical means in the art not disclosed in the embodiments of this application.

Claims

1. A fault prediction method, characterized in that, Applied to a computing device, the method includes: Obtain a first fault prediction model, which is trained based on the first sample data of the source object, and the number of the first sample data is greater than or equal to a preset number. Acquire multiple initial sample data of the target object, the initial sample data including the historical physical operating parameters of the target object and the fault information corresponding to the set of historical physical operating parameters, wherein the target object and the source object are hardware of different models; Based on the feature mapping rules between the target object and the source object, the second sample data corresponding to each initial sample data in the plurality of initial sample data is determined. The number of initial sample data is the same as the number of second sample data, and the number of second sample data is less than the preset number. A second fault prediction model is trained based on the first fault prediction model and the second sample data. The second fault prediction model is used to predict faults in the target object.

2. The method according to claim 1, characterized in that, Before determining the second sample data corresponding to each initial sample data in the plurality of initial sample data according to the feature mapping rule between the target object and the source object, the method further includes: The first running data and the first fault data of the source object are obtained. The first running data includes a first value range of multiple physical running parameters of the source object. The first fault data includes multiple first fault identifiers of the source object and descriptive information of each first fault identifier. The first fault identifier is used to indicate the fault type, fault location, or fault cause of the source object. The second operating data and the second fault data of the target object are obtained. The second operating data includes the second value range of multiple physical operating parameters of the target object. The second fault data includes multiple second fault identifiers of the target object and descriptive information of each second fault identifier. The second fault identifier is used to indicate the fault type, fault location, or fault cause of the target object. Based on the first operating data and the second operating data, the mapping relationship of the physical operating parameters is determined; Based on the first fault data and the second fault data, determine the mapping relationship of the fault information; The feature mapping rules are determined based on the mapping relationship of the physical operating parameters and the mapping relationship of the fault information.

3. The method according to claim 2, characterized in that, Determining the mapping relationship of the physical operation parameters based on the first operation data and the second operation data includes: For any physical operating parameter, the mapping rule corresponding to the physical operating parameter is determined based on the endpoint values ​​of the second value range corresponding to the physical operating parameter and the endpoint values ​​of the first value range corresponding to the physical operating parameter. Determining the mapping relationship of the physical operating parameters includes mapping rules corresponding to multiple physical operating parameters.

4. The method according to claim 2 or 3, characterized in that, Determining the mapping relationship of the fault information based on the first fault data and the second fault data includes: For any second fault identifier, target description information is determined in the first fault data based on the description information of the second fault identifier, wherein the similarity between the description information of the second fault identifier and the target description information is greater than or equal to a preset threshold; a mapping relationship is established between the second fault identifier and the first fault identifier corresponding to the target description information. Determining the mapping relationship of the fault information includes the mapping relationship between the plurality of second fault identifiers and the corresponding first fault identifiers.

5. The method according to claim 1, characterized in that, The first fault prediction model is a random forest model; the step of training a second fault prediction model for the target object based on the first fault prediction model and the second sample data includes: The feature vector of the second sample data is obtained through the first fault prediction model. The feature vector of the second sample data is used to indicate the node path, leaf node index or node output value of the second sample data on each decision tree of the first fault prediction model. The second fault prediction model is trained based on the feature vector of the second sample data.

6. The method according to claim 1, characterized in that, The first fault prediction model is a neural network model; the step of training a second fault prediction model for the target object based on the first fault prediction model and the second sample data includes: An initial fault prediction model is constructed based on the first fault prediction model, and the initial fault prediction model includes a trainable output layer. Based on the second sample data, the output layer of the initial fault prediction model is trained to obtain the second fault prediction model.

7. The method according to claim 6, characterized in that, The construction of the initial fault prediction model based on the first fault prediction model includes: Freeze the feature extraction layer of the first fault prediction model; Delete the output layer of the first fault prediction model; An output layer is added after the feature extraction layer of the first fault prediction model to obtain the initial fault prediction model.

8. A computing device, characterized in that, include: Memory and processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method according to any one of claims 1-7.

9. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-7.

Images