Equipment detection method and device, electronic equipment, storage medium and program product

By building specific detection models for data center equipment, using multiple operating indicator data and decision tree algorithms, the equipment deterioration problem caused by the difficulty of detecting complex factors in the existing technology is solved, and more accurate equipment status assessment and health are achieved Condition monitoring.

CN118897782APending Publication Date: 2024-11-05CHINA MOBILE INFORMATION TECHNOLOGY CO LTD +1
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202410915285.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-09
Publication Date
2024-11-05

AI Technical Summary

Technical Problem

Existing data center equipment degradation detection methods rely on a single index threshold, making it difficult to cope with equipment performance degradation caused by complex factors, affecting detection accuracy.

Method used

By building corresponding detection models for different device types, using multiple operating indicator data for training, combining decision tree algorithm and LightGBM algorithm, probability prediction values ​​are generated to determine the normal or degraded state of the device, and to calculate the characteristic importance value to explain the detection result.

Benefits of technology

It improves the accuracy of equipment detection, can reflect the operating status of the equipment more comprehensively, reduces false alarms and missed reports, and enhances the ability to evaluate the health status of the equipment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118897782A_ABST
    Figure CN118897782A_ABST
Patent Text Reader

Abstract

The invention discloses an equipment detection method and device, electronic equipment, a storage medium and a program product. The method comprises the following steps: acquiring various first operation index data of to-be-detected equipment; determining a target detection model corresponding to the device type of the to-be-detected device from a plurality of pre-trained detection models based on a preset corresponding relationship between the device type and the detection model, the detection model being obtained by training based on a training sample set constructed by a plurality of second operation index data of the corresponding device type; inputting the multiple first operation index data into a target detection model to obtain a detection result, the detection result comprising a probability prediction value of the to-be-detected device in a target state, and the target state comprising a normal state and / or a degradation state; and determining the operation state of the to-be-detected equipment according to the detection result. According to the embodiment of the invention, the accuracy of equipment detection can be effectively improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of data centers, and in particular, relates to a device detection method, device, electronic equipment, storage medium and program product. Background Art

[0002] A data center is a facility that centrally processes, stores, and transmits data, typically including computers, servers, storage devices, network equipment, power equipment, and security equipment to support the operations of an organization (websites, applications, databases, etc.).

[0003] Degradation refers to the decline in equipment performance due to various factors (such as mechanical wear, electrical aging, corrosion, etc.) during long-term use of equipment. Degradation detection can promptly detect equipment performance degradation problems, avoid service interruptions and data loss due to equipment failures, and thus ensure the stable operation of the data center.

[0004] Equipment degradation detection in data centers usually uses statistical analysis and rule engine technology to evaluate equipment status by collecting and analyzing hardware monitoring and equipment log data. This method relies on setting thresholds and rules to monitor key indicators (such as temperature, pressure, etc.) to determine whether the equipment is normal. Specifically, when an indicator exceeds the preset threshold, the system prompts that there may be a problem with the equipment. However, equipment degradation may be caused by the combined effect of multiple factors. This detection method that relies on a preset single indicator threshold is difficult to cope with more complex equipment degradation situations, which may affect the accuracy of equipment detection. Summary of the invention

[0005] The embodiments of the present invention provide a device detection method, apparatus, electronic device, storage medium and program product, which can effectively improve the accuracy of device detection.

[0006] In a first aspect, an embodiment of the present application provides a detection method for a device, where the device is a device in a data center, and the device has multiple types, and multiple device types correspond to multiple detection models one by one. The method includes:

[0007] Acquire multiple first operating indicator data of the device to be detected;

[0008] Based on the correspondence between the preset device type and the detection model, a target detection model corresponding to the device type of the device to be detected is determined from multiple pre-trained detection models, and the detection model is trained based on a training sample set constructed based on multiple second operation indicator data of the corresponding device type, and the second operation indicator data includes operation indicator data of normally operating devices and operation indicator data of faulty devices related to the device type;

[0009] Inputting a plurality of first operation indicator data into a target detection model to obtain a detection result, wherein the detection result includes a probability prediction value of the device to be detected being in a target state, wherein the target state includes a normal state and / or a degraded state;

[0010] According to the test results, determine the operating status of the equipment to be tested.

[0011] In some embodiments, the method further comprises:

[0012] For each pre-trained detection model, calculate the importance value of each feature, each feature corresponds to a second operation indicator data, and the importance value is used to characterize the importance of each feature to the detection result output by the pre-trained detection model;

[0013] After determining the operating status of the device to be detected according to the detection result, the method further includes:

[0014] According to the operating state of the device to be detected and the importance value of each feature in the target detection model, a detection result explanation for the device to be detected is generated, and the detection result explanation includes the influence degree of each feature on the operating state of the device to be detected.

[0015] In some embodiments, the detection model is trained using a decision tree algorithm, the training sample set includes multiple samples, and for each pre-trained detection model, the importance value of each feature is calculated, including:

[0016] For each pre-trained detection model, do the following:

[0017] For each node of the decision tree corresponding to the detection model, obtain the features included in the node and the number of positive and negative samples corresponding to each feature;

[0018] For each node, the Gini impurity value of the node is calculated according to the features included in the node and the number of positive and negative samples corresponding to each feature, and the weight of the node is calculated according to the Gini impurity value and the sample ratio. The sample ratio is the ratio of the number of samples on the node to the total number of samples. The Gini impurity value is used to characterize the degree of mixing of samples of different categories in the node;

[0019] For each feature, the sum of the weights of the feature on all nodes is calculated to get the importance value of the feature.

[0020] In some embodiments, determining the operating status of the device to be detected according to the detection result includes:

[0021] When the detection result is a probability prediction value that the device to be detected is in a normal state, and the probability prediction value is greater than a first preset threshold, determining that the operating state of the device to be detected is in a normal state;

[0022] When the detection result is a probability prediction value that the device to be detected is in a degraded state, and the probability prediction value is greater than a second preset threshold, determining that the operating state of the device to be detected is in a degraded state;

[0023] The first preset threshold and the second preset threshold are determined based on the operating characteristics of the device type to which the device to be detected belongs and the third operating indicator data of the device to be detected within a preset time.

[0024] In some embodiments, before inputting the plurality of first operating indicator data into the target detection model to obtain the detection result, the method further includes:

[0025] For each device type, do the following:

[0026] Build an initial detection model;

[0027] Obtaining a training sample set, the training sample set comprising a plurality of training samples, each training sample comprising a plurality of second operating indicator data of a device corresponding to the device type and a label of whether the device is in a normal state;

[0028] The initial detection model is trained using multiple training samples until the training stop condition is met to obtain a trained detection model.

[0029] In some embodiments, constructing an initial detection model includes:

[0030] It is beneficial for LightGBM algorithm to build decision tree model;

[0031] The decision tree model was determined as the initial detection model.

[0032] In a second aspect, an embodiment of the present application provides a device for detecting a device, wherein the device is a device in a data center, and the device has multiple types, and multiple device types correspond to multiple detection models one by one, and the device includes:

[0033] A first acquisition module, used to acquire a plurality of first operating indicator data of the device to be detected;

[0034] A first determination module is used to determine a target detection model corresponding to the device type of the device to be detected from a plurality of pre-trained detection models based on a correspondence between a preset device type and a detection model, the detection model being trained based on a training sample set constructed based on a plurality of second operation indicator data of the corresponding device type, the second operation indicator data including operation indicator data of a normally operating device and operation indicator data of a faulty device related to the device type;

[0035] A second determination module is used to input a plurality of first operation indicator data into a target detection model to obtain a detection result, wherein the detection result includes a probability prediction value that the device to be detected is in a target state, and the target state includes a normal state and / or a degraded state;

[0036] The third determination module is used to determine the operating status of the device to be detected according to the detection result.

[0037] In a third aspect, an embodiment of the present application provides an electronic device, the electronic device comprising:

[0038] a processor and a memory storing computer program instructions;

[0039] When the processor executes the computer program instructions, the detection method of the device as described in any one of the first aspects is implemented.

[0040] In a fourth aspect, an embodiment of the present application provides a computer storage medium having computer program instructions stored thereon, and when the computer program instructions are executed by a processor, a detection method for a device as described in any one of the first aspects is implemented.

[0041] In a fifth aspect, an embodiment of the present application provides a computer program product, which includes a computer program, and when the computer program is processed and executed, it implements the detection method of a device as described in any one of the first aspects.

[0042] The detection method, device, electronic device, storage medium and program product of the equipment of the embodiment of the present application can obtain the multi-dimensional characteristics of the equipment to be detected by collecting multiple operating indicator data of the equipment to be detected. Because different equipment types may have different operating laws and failure modes. Therefore, by constructing corresponding detection models for different equipment types, and using the second operating indicator data corresponding to different equipment types to train the obtained multiple detection models, the specific characteristics and failure modes of the corresponding equipment types can be captured respectively. After that, the health status of the equipment to be detected is evaluated by the target detection model corresponding to the equipment type of the equipment to be detected, so that the target detection model can use machine learning technology to mine the relationship between the multiple first operating indicator data of the equipment to be detected. The detection results obtained by integrating multiple first operating indicator data can more comprehensively reflect the operating status of the equipment to be detected, and effectively improve the detection accuracy of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solution of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0044] Figure 1 A schematic diagram showing a flow chart of a detection method for a device provided by an embodiment of the present application;

[0045] Figure 2 A schematic diagram showing a flow chart of a detection method for a device provided in another embodiment of the present application;

[0046] Figure 3 A flowchart of a specific implementation method of S104 provided in an embodiment of the present application is shown;

[0047] Figure 4 A flowchart of a specific implementation method for calculating the importance value of each feature provided by an embodiment of the present application is shown;

[0048] Figure 5 A schematic diagram of a node of a decision tree provided by an embodiment of the present application is shown;

[0049] Figure 6 A schematic diagram of an exemplary decision tree provided by an embodiment of the present application is shown;

[0050] Figure 7 A schematic diagram showing the structure of a detection device of a device provided by an embodiment of the present application is shown;

[0051] Figure 8 A schematic structural diagram of an electronic device provided by an embodiment of the present application is shown. DETAILED DESCRIPTION

[0052] The features and exemplary embodiments of various aspects of the present application will be described in detail below. In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present application, rather than to limit the present application. For those skilled in the art, the present application can be implemented without the need for some of these specific details. The following description of the embodiments is only to provide a better understanding of the present application by illustrating the examples of the present application.

[0053] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "include..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0054] Before describing the technical solutions provided by the embodiments of the present application, in order to facilitate the understanding of the embodiments of the present application, the present application first specifically describes the relevant technologies:

[0055] A data center is a physical facility that centrally manages and processes large amounts of data. It typically includes storage devices, cabinets, network equipment, servers, cooling systems, monitoring and management systems, security equipment, computers, power equipment, and other infrastructure to support information technology services such as cloud computing, big data analysis, website hosting, etc.

[0056] In the operation of data centers, equipment degradation is an issue that cannot be ignored. Currently, related degradation detection methods may only focus on a specific indicator, such as temperature, voltage, or operating time, and set a fixed threshold for it. When the indicator exceeds the threshold, it is judged that the device may have a degradation problem. However, equipment degradation is usually the result of the combined effect of multiple factors, which may include but are not limited to environmental factors (such as temperature, humidity), frequency of equipment use, physical wear, power fluctuations, network attacks, etc. Obviously, it is difficult for a single indicator to fully and accurately capture the overall health status of the equipment.

[0057] In order to solve the above technical problems, the embodiments of the present application provide a device detection method, apparatus, electronic device, storage medium and program product. The device detection method provided by the embodiments of the present application is first introduced below.

[0058] Figure 1 The flowchart of the device detection method provided by one embodiment of the present application is shown. The device detection method is used to detect the devices in the data center. The devices have multiple types, and the multiple device types correspond to multiple detection models one by one. Figure 1 As shown, the detection method may include the following steps: S101-S104.

[0059] S101: Acquire multiple first operating indicator data of the device to be detected.

[0060] S102: Based on the correspondence between the preset device type and the detection model, a target detection model corresponding to the device type of the device to be detected is determined from multiple pre-trained detection models, and the detection model is trained on a training sample set constructed based on multiple second operating indicator data of the corresponding device type, and the second operating indicator data includes operating indicator data of normally operating devices and operating indicator data of faulty devices related to the device type.

[0061] S103: Inputting a plurality of first operation indicator data into a target detection model to obtain a detection result, wherein the detection result includes a probability prediction value that the device to be detected is in a target state, and the target state includes a normal state and / or a degraded state.

[0062] S104: Determine the operating status of the device to be detected according to the detection result.

[0063] In an embodiment of the present application, by collecting a variety of operating indicator data of the device to be detected, the multi-dimensional characteristics of the device to be detected can be obtained. Since different equipment types may have different operating rules and failure modes. Therefore, by constructing corresponding detection models for different equipment types, and using the second operating indicator data corresponding to different equipment types for training to obtain multiple detection models, the specific characteristics and failure modes of the corresponding equipment types can be captured respectively. After that, the health status of the device to be detected is evaluated by the target detection model corresponding to the device type of the device to be detected, so that the target detection model can use machine learning technology to mine the relationship between the multiple first operating indicator data of the device to be detected. The detection results obtained by integrating multiple first operating indicator data can more comprehensively reflect the operating status of the device to be detected, and effectively improve the detection accuracy of the equipment.

[0064] In some embodiments, in S101, the device to be detected is a device in a data center. The first operating index data can be collected by a monitoring system deployed in the device to be detected or in a data center. As an example, the monitoring system may include various sensors installed on the device to be detected, such as a temperature sensor, a humidity sensor, a voltage / current sensor, etc., to achieve comprehensive detection of the device to be detected, and process the collected data in combination with data caching technology and a real-time processing algorithm to ensure that the collected data can truly reflect the operating status of the device to be detected.

[0065] In some embodiments, in S102, a database may be pre-built, the database including the correspondence between the preset device type and the corresponding detection model, so that when the device to be detected is detected, the preset correspondence may be queried through the device type of the device to be detected and the database, thereby quickly and accurately determining the target detection model that matches the device to be detected. Then, in S103, a plurality of first operation indicator data are input into the target detection model to obtain the detection result of the device to be detected.

[0066] It is worth noting that different types of equipment may obtain different first operating indicator data due to differences in their functions and architectures. The following Table 1 is an example table showing common data center equipment and their corresponding operating indicators:

[0067]

[0068] Table 1

[0069] In some embodiments, in order to more accurately capture the operating mode and abnormal characteristics unique to each device type, so as to more accurately identify potential problems or abnormal situations related to the corresponding device type. Figure 2 As shown, the detection method may further include executing the following steps for each device type: S201 to S203.

[0070] S201: Build an initial detection model.

[0071] In some embodiments, in order to effectively improve the training speed and prediction performance of the detection model, the LightGBM (Light Gradient Boosting Machine) algorithm can be used to build a decision tree model, and then the parameters of LightGBM are set, including the learning rate, the depth of the tree, the number of leaf nodes, the regularization parameter, etc. Finally, the decision tree model that has been set is determined as the initial detection model.

[0072] LightGBM adopts a leaf node splitting strategy when building a decision tree, which can reduce the amount of calculation, make it easier to handle high-dimensional sparse features, and also supports parallel training and feature parallelization, further improving training efficiency. In addition, LightGBM uses a gradient boosting algorithm to train multiple weak classifiers, each of which attempts to learn the best prediction function to reduce the residual of the current model, thereby gradually generating a decision tree model, and effectively improving the accuracy of the detection model through feature selection.

[0073] S202: Acquire a training sample set, where the training sample set includes a plurality of training samples, and each training sample includes a plurality of second operation indicator data of a device corresponding to the device type and a label indicating whether the device is in a normal state.

[0074] In some embodiments, the second operating index data includes operating index data of a normally operating device and operating index data of a faulty device. The second operating index data can be collected using the aforementioned monitoring system. For specific content, please refer to the specific content of the aforementioned monitoring system, which is not limited here. For each set of second operating index data collected, it is necessary to mark it, that is, to indicate whether the device corresponding to the data is in a normal state or a faulty state.

[0075] When training the detection model, the collected second operation indicator data needs to be cleaned, converted, and normalized for model training. Next, we will use a specific example, such as temperature data, to explain the steps of processing temperature data:

[0076] 1. Data Cleansing

[0077] During the data collection process, due to various reasons such as network jitter and device offline, there may be problems such as missing data, outliers, and duplicate data. These problems may affect the accuracy of data analysis and the effect of model training. Therefore, for missing temperature data, mean filling, median filling, interpolation and other methods can be used to supplement to ensure data integrity; for abnormal extreme values ​​caused by sensor failure or data transmission errors, the threshold method can be used to identify and replace abnormal values.

[0078] 2. Data Conversion and Normalization

[0079] Normalize the temperature data, such as converting the temperature data into numbers in the range of 0-1, to eliminate the impact of different dimensions on data analysis and model training, thereby improving the accuracy of modeling. Normalization can be achieved through linear transformation, maximum and minimum scaling, etc. Furthermore, in order to reduce the impact of noise on data analysis, the temperature data can be smoothed or filtered.

[0080] 3. Data Integration

[0081] In a data center, there is temperature data generated by multiple different devices. When performing data analysis and modeling, this data needs to be integrated. Specifically, the temperature data collected by different devices is integrated and aligned according to time to ensure that the time series of the data is aligned.

[0082] It is worth mentioning here that in order to better reflect the operating status of the equipment, it is also necessary to extract features from each group of second operating indicator data, and combine the extracted features and corresponding labels into training samples, and then use multiple training samples for model training.

[0083] In some embodiments, in order to reduce noise and redundant information in multiple training samples, ensure that the model focuses on the features that are most relevant to the target variable, and avoid overfitting and unnecessary computational overhead. Before building the initial detection model, feature selection is required. Specifically, any of the following feature selection methods can be used:

[0084] Correlation analysis: Use statistical tools such as Pearson correlation coefficient or Spearman correlation coefficient to calculate the correlation between each feature and the target variable. Select features that are highly correlated with the target variable, and avoid highly correlated features to reduce redundant information. Among the highly correlated features, only one can be retained.

[0085] ANOVA: Evaluate the variance of each feature using ANOVA methods such as the F test. Features with larger variances usually contain more information and are therefore more likely to have a positive impact on model performance. Selecting features with larger variances helps the model capture key changes in the data.

[0086] Chi-square test: When both the feature and the target variable are discrete values, the chi-square test is an effective feature selection method. By calculating the chi-square value between a feature and the target variable, we can evaluate the association between the feature and the target variable. Selecting features with larger chi-square values ​​means that these features are more strongly associated with the target variable, and are more likely to have a positive impact on model performance.

[0087] Principal Component Analysis (PCA): PCA is a dimensionality reduction technique that transforms the original feature space into a low-dimensional principal component space. In this new space, each principal component is a linear combination of the original features and is sorted according to their contribution to the data variance. By selecting the most representative principal component features, the redundant information in the data set can be reduced while retaining the main structure of the data, which in turn helps reduce the complexity of the model and improves training speed and generalization ability.

[0088] S203: Using multiple training samples to train the initial detection model until a training stop condition is met, thereby obtaining a trained detection model.

[0089] As an optional implementation method, multiple training samples can be further refined into training sets and validation sets for training and evaluating the initial detection model. In the model evaluation phase, we use the validation set to measure the performance of the trained detection model and comprehensively evaluate the effect of the detection model by calculating key indicators such as accuracy, recall, and F1 value. Subsequently, based on the evaluation results, the model is tuned. During the tuning process, a grid search of parameters can be used to optimize the corresponding detection model parameters to further enhance the prediction ability of the detection model.

[0090] In some embodiments, during the training process, the initial detection model will continuously learn based on the characteristics and labels of the training samples and adjust the model parameters of the initial detection model. When the performance of the model on the validation set reaches a preset standard (such as the accuracy reaches a certain threshold) or meets other training stop conditions (such as reaching the maximum number of iterations, the training time exceeds the preset limit, etc.), the training process ends and a trained detection model is obtained.

[0091] In an embodiment of the present application, the initial detection model is trained by multiple training samples, so that the initial detection model can learn the relationship between the features and labels of a large amount of training sample data during the training process, and thus can more accurately identify the normal state and fault state of the equipment.

[0092] In some embodiments, in order to improve the accuracy of the detection results, Figure 3 As shown, the aforementioned S104 may specifically include: S1041 to S1042.

[0093] S1041: When the detection result is a probability prediction value that the device to be detected is in a normal state, and the probability prediction value is greater than a first preset threshold, determine that the operating state of the device to be detected is in a normal state.

[0094] S1042: When the detection result is a probability prediction value that the device to be detected is in a degraded state, and the probability prediction value is greater than a second preset threshold, determine that the operating state of the device to be detected is in a degraded state.

[0095] The first preset threshold and the second preset threshold are determined based on the operating characteristics of the device type to which the device to be detected belongs and the third operating indicator data of the device to be detected within a preset time.

[0096] In some embodiments, the third operating indicator data is historical operating data of the device to be detected, which is used to analyze normal and degraded operating modes of the device to be detected and provide a basis for setting the first preset threshold and the second preset threshold.

[0097] In the embodiment of the present application, by considering the characteristics of the device type, a specific threshold can be customized for each device so that the detection is more consistent with its operating characteristics and failure mode. By comparing the probability prediction value with the customized threshold, the normal operating state and abnormal state of the device can be more accurately distinguished, thereby reducing false positives and false negatives.

[0098] In some embodiments, in order to better understand the decision-making process of the detection model, the detection method further includes:

[0099] For each pre-trained detection model, the importance value of each feature is calculated, each feature corresponds to a second operation indicator data, and the importance value is used to characterize the importance of each feature to the detection result output by the pre-trained detection model.

[0100] In some embodiments, after obtaining a trained detection model, Gini Importance can be used to calculate the importance value of each feature to explain the detection results of the detection model. The Gini Importance method is used to evaluate the importance of each feature to the prediction performance of the model, which helps to understand the decision-making process of the model and its impact on device degradation.

[0101] As a specific example, Figure 4 As shown, the importance value of each feature can be calculated through the following steps: S401 to S403.

[0102] S401: For each node of the decision tree corresponding to the detection model, obtain the features included in the node and the number of positive and negative samples corresponding to each feature.

[0103] In some embodiments, in the decision tree, each node is split according to the value of a certain feature to distinguish different categories. For each node in the decision tree, the feature used for splitting the node and the number of positive and negative samples corresponding to the feature are identified and recorded.

[0104] S402: For each node, the Gini impurity value of the node is calculated according to the features included in the node and the number of positive and negative samples corresponding to each feature, and the weight of the node is calculated according to the Gini impurity value and the sample ratio. The sample ratio is the ratio of the number of samples on the node to the total number of samples. The Gini impurity value is used to characterize the degree of mixing of samples of different categories in the node.

[0105] In some embodiments, the Gini impurity value (Gini index) refers to the expected error rate of a data item when a certain result from a set is randomly applied to the set. The lower the Gini impurity value, the higher the purity of the set, and vice versa. The sample ratio represents the ratio of the number of samples on a node to the total number of samples, which is used to consider the importance of the node when calculating the node weight. The weight of the node = Gini impurity value * (node ​​sample number / total sample number).

[0106] As an example, the Gini Index can be calculated using the following formula:

[0107]

[0108] Where n represents the total number of different categories in the node, P iIndicates the proportion of samples of the i-th category in the node.

[0109] Figure 5 A schematic diagram of a node of a decision tree provided by an embodiment of the present application is shown. Figure 5 The Gini impurity value for the temperature feature shown is calculated as follows:

[0110]

[0111] In formula (2), n represents the branch node number, i represents the number of data sample categories, P represents the probability of the child node, and P i It represents the proportion of samples of the i-th category in the node, P+ represents the number of leaf nodes whose current node is classified as True, and P- represents the number of leaf nodes whose current node is classified as False.

[0112] S403: For each feature, calculate the sum of the weights of the feature on all nodes to obtain the importance value of the feature.

[0113] Among them, the importance value reflects the contribution of the feature to the classification result in the decision tree.

[0114] In the specific implementation, first, in S401, traverse each node of the decision tree, record the features on which each node is split and the number of samples of different categories under the feature. Then, in S401, at each node, calculate using the Gini impurity formula based on the recorded number and category of samples. Then, multiply the calculated Gini impurity value by the proportion of node samples to total samples to obtain the weight of the node. Finally, in S403, for each feature, when traversing all nodes, add up the weights of the nodes containing the feature, and finally obtain the importance value of each feature.

[0115] In the embodiment of the present application, the contribution of each feature in the decision tree to the classification result can be quantified, which helps to understand which features play a key role in the model prediction, and then optimize the feature selection and improve the prediction performance and interpretability of the model.

[0116] As an optional implementation, in LightGBM, you can use the model's feature_importances method to evaluate the importance of each feature in the model's prediction. This method returns a list of features sorted by feature importance. Based on this sorting, you can select a specific number of important features as needed, such as the top 5, top 10, or more important features, and then analyze the contribution of each important feature to the model's prediction results. For each important feature, you can analyze its impact on the prediction results, for example, whether it is positively correlated or negatively correlated, and the extent of its impact.

[0117] Take the constructed detection model as an example, Figure 6 As shown, the following feature importance scores can be obtained:

[0118] Temperature: 0.45

[0119] Vibration: 0.25

[0120] Voltage: 0.15

[0121] Current: 0.10

[0122] Humidity: 0.05

[0123] Based on these feature importance scores, the top three important features, namely temperature, vibration, and voltage, were selected. Analysis of these key features revealed their contributions to the model prediction results. Specifically, temperature and vibration were identified as the two factors with the greatest impact on the prediction of device degradation, while voltage also had a certain impact, but to a lesser extent than the first two factors. Current and humidity showed relatively low influence.

[0124] Furthermore, the feature importance scores and analysis results can be integrated to provide explanations for the detection results of the detection model, for example: "For the degradation prediction of this type of equipment, it is mainly affected by temperature and vibration, while current and humidity have little impact on the detection results." Such explanations not only make the detection results more intuitive and easy to understand, but also help users or maintenance personnel quickly identify the key factors that affect the operating status of the equipment.

[0125] Based on this, after the aforementioned S104, the detection method further includes:

[0126] According to the operating state of the device to be detected and the importance value of each feature in the target detection model, a detection result explanation for the device to be detected is generated, and the detection result explanation includes the influence degree of each feature on the operating state of the device to be detected.

[0127] In the embodiment of the present application, by displaying the degree of influence of each feature on the operating status of the equipment, the detection results are made more intuitive and easy to understand, which helps users or maintenance personnel to quickly grasp the key factors of the equipment status.

[0128] Next, the detection method of the present application will be further explained with reference to a specific example.

[0129] The operation of a certain device was abnormal, with the temperature abnormally rising to 55 degrees and the vibration frequency being maintained at more than 60Hz for a long time. According to our detection model, the detection results showed that the device had a potential failure risk. By comparing the feature importance of the corresponding detection model, it can be found that temperature and vibration frequency occupy a relatively important position in degradation detection. This shows that abnormal temperature rise and abnormal vibration frequency play a decisive role in degradation detection.

[0130] It can be concluded that the high temperature state and abnormal vibration frequency of the equipment are the main factors for judging its potential failure risk. Feature importance analysis not only reveals the generation mechanism of the test results, but also enables users to monitor the operating status of the equipment more accurately based on the content and take corresponding measures to avoid equipment failure, thereby ensuring the performance of the equipment and extending its service life.

[0131] In summary, by explaining the importance of the features of the decision tree, we can clearly explain the detection results of the detection model, deeply understand how the detection model diagnoses and predicts equipment degradation, and which features have a significant impact on the detection results. These explanations not only enhance the user's understanding of the health status and fault types of the equipment, but also provide clear guidance for subsequent maintenance, inspection and possible fault repair work.

[0132] In some embodiments, in order to further increase the detection effect of the detection model, in the model iteration optimization stage, the features and the interpretability of the detection results can be optimized according to the device characteristics, and the model parameters of the detection model can be adjusted to improve the detection accuracy and generalization ability of the model.

[0133] Based on the device detection method provided in the above embodiment, the present application also provides a specific implementation of the device detection device. Please refer to the following embodiment.

[0134] See first Figure 7 The device detection device provided in the embodiment of the present application is used to detect the device of the data center. The device has multiple types, and multiple device types correspond to multiple detection models one by one. Specifically, the device may include the following modules:

[0135] The first acquisition module 701 is used to acquire a plurality of first operation indicator data of the device to be detected.

[0136] The first determination module 702 is used to determine a target detection model corresponding to the device type of the device to be detected from multiple pre-trained detection models based on the correspondence between the preset device type and the detection model, and the detection model is trained on a training sample set constructed based on multiple second operating indicator data of the corresponding device type, and the second operating indicator data includes operating indicator data of normally operating devices and operating indicator data of faulty devices related to the device type.

[0137] The second determination module 703 is used to input a plurality of first operation indicator data into the target detection model to obtain a detection result, wherein the detection result includes a probability prediction value that the device to be detected is in a target state, and the target state includes a normal state and / or a degraded state.

[0138] The third determination module 704 is used to determine the operating status of the device to be detected according to the detection result.

[0139] In the embodiment of the present application, the first acquisition module 701 is used to collect multiple operating index data of the device to be detected, so that the multi-dimensional characteristics of the device to be detected can be obtained. Because different equipment types may have different operating rules and failure modes. Therefore, by constructing corresponding detection models for different equipment types, and using the second operating index data corresponding to different equipment types to train the obtained multiple detection models, the specific characteristics and failure modes of the corresponding equipment types can be captured respectively. After that, the health status of the device to be detected is evaluated by the target detection model corresponding to the device type of the device to be detected using the first determination module 702 and the second determination module 703, so that the target detection model can use machine learning technology to mine the relationship between the multiple first operating index data of the device to be detected. The detection results obtained by integrating multiple first operating index data can more comprehensively reflect the operating status of the device to be detected, and effectively improve the detection accuracy of the device.

[0140] As an implementation of the present application, in order to more accurately capture the operating mode and abnormal characteristics unique to each device type, so as to more accurately identify potential problems or abnormal conditions related to the corresponding device type, the above-mentioned apparatus may further include:

[0141] The first building block: used to build the initial detection model.

[0142] The second acquisition module is used to acquire a training sample set, the training sample set includes multiple training samples, each training sample includes multiple second operation indicator data of a device corresponding to the device type and a label indicating whether the device is in a normal state.

[0143] The fourth determination module is used to train the initial detection model using multiple training samples until the training stop condition is met to obtain the trained detection model.

[0144] As an implementation of the present application, in order to improve the accuracy of the detection result, the third determination module 704 may further include:

[0145] The first determination submodule is used to determine that the operating state of the device to be detected is in a normal state when the detection result is a probability prediction value that the device to be detected is in a normal state and the probability prediction value is greater than a first preset threshold.

[0146] The second determination submodule is used to determine that the operating state of the device to be detected is in a degraded state when the detection result is a probability prediction value that the device to be detected is in a degraded state and the probability prediction value is greater than a second preset threshold value. The first preset threshold value and the second preset threshold value are determined based on the operating characteristics of the device type to which the device to be detected belongs and the third operating indicator data of the device to be detected within a preset time.

[0147] As an implementation of the present application, in order to better understand the decision-making process of the detection model, the detection device may also include:

[0148] The first calculation module is used to calculate the importance value of each feature for each pre-trained detection model, each feature corresponds to a second operating indicator data, and the importance value is used to characterize the importance of each feature to the detection result output by the pre-trained detection model.

[0149] As an implementation of the present application, in order to improve the accuracy of calculating the importance value of each feature, the first calculation module may specifically include:

[0150] The first acquisition unit is used to acquire, for each node of the decision tree corresponding to the detection model, the features included in the node and the number of positive and negative samples corresponding to each feature.

[0151] The first calculation unit is used to calculate the Gini impurity value of each node according to the features included in the node and the number of positive and negative samples corresponding to each feature, and calculate the weight of the node according to the Gini impurity value and the sample ratio. The sample ratio is the ratio of the number of samples on the node to the total number of samples. The Gini impurity value is used to characterize the degree of mixing of samples of different categories in the node.

[0152] The second calculation unit is used to calculate the sum of the weights of each feature on all nodes to obtain the importance value of the feature.

[0153] As an implementation of the present application, in order for users or maintenance personnel to quickly grasp the key factors of the device status, the device may also include:

[0154] The first generation module is used to generate a detection result explanation for the device to be detected according to the operating state of the device to be detected and the importance value of each feature in the target detection model, and the detection result explanation includes the influence degree of each feature on the operating state of the device to be detected.

[0155] Figure 8A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application is shown.

[0156] The electronic device may include a processor 801 and a memory 802 storing computer program instructions.

[0157] Specifically, the processor 801 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0158] The memory 802 may include a large capacity memory for data or instructions. By way of example and not limitation, the memory 802 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. In appropriate cases, the memory 802 may include a removable or non-removable (or fixed) medium. In appropriate cases, the memory 802 may be inside or outside the integrated gateway disaster recovery device. In a specific embodiment, the memory 802 is a non-volatile solid-state memory.

[0159] The memory 802 may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk storage media device, an optical storage media device, a flash memory device, an electrical, optical or other physical / tangible memory storage device. Thus, typically, the memory 802 includes one or more tangible (non-transitory) computer-readable storage media (e.g., a memory device) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to an aspect of the present disclosure.

[0160] The processor 801 implements any device detection method in the above embodiments by reading and executing computer program instructions stored in the memory 802 .

[0161] In one example, the electronic device may further include a communication interface 803 and a bus 810. Figure 8 As shown, the processor 801, the memory 802, and the communication interface 803 are connected via a bus 810 and communicate with each other.

[0162] The communication interface 803 is mainly used to implement communication between various modules, devices, units and / or equipment in the embodiments of the present application.

[0163] Bus 810 includes hardware, software or both, and the parts of electronic equipment are coupled to each other.For example, but not limitation, bus may include accelerated graphics port (AGP) or other graphics bus, enhanced industrial standard architecture (EISA) bus, front side bus (FSB), hypertransport (HT) interconnection, industrial standard architecture (ISA) bus, infinite bandwidth interconnection, low pin count (LPC) bus, memory bus, micro channel architecture (MCA) bus, peripheral component interconnection (PCI) bus, PCI-Express (PCI-X) bus, serial advanced technology attachment (SATA) bus, video electronics standard association local (VLB) bus or other suitable bus or two or more of these combinations. In appropriate cases, bus 810 may include one or more buses. Although the present application embodiment describes and shows a specific bus, the application considers any suitable bus or interconnection.

[0164] The electronic device can execute the detection method of the device in the embodiment of the present application, thereby realizing the combination Figure 1 and Figure 7 A detection method and apparatus for a device are described.

[0165] In addition, in combination with the device detection method in the above embodiment, the embodiment of the present application can provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when the computer program instructions are executed by the processor, any device detection method in the above embodiment is implemented.

[0166] In addition, in combination with the device detection method in the above embodiment, the embodiment of the present application can provide a computer program product for implementation. The computer program product includes a computer program, and when the computer program is processed and executed, it implements any of the above device detection methods.

[0167] It should be clear that the present application is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present application is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present application.

[0168] The functional blocks shown in the above block diagram can be implemented as hardware, software, firmware or a combination thereof. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present application are programs or code segments that are used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link by a data signal carried in a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, optical fiber media, radio frequency (RF) links, etc. The code segment can be downloaded via a computer network such as the Internet, an intranet, etc.

[0169] It should also be noted that the exemplary embodiments mentioned in this application describe some methods or systems based on a series of steps or devices. However, this application is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiment, or in a different order from the embodiment, or several steps can be performed simultaneously.

[0170] Aspects of the present disclosure are described above with reference to the flowchart and / or block diagram of the method, device (system) and computer program product according to the embodiment of the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device to produce a machine so that these instructions executed by the processor of the computer or other programmable data processing device enable the implementation of the function / action specified in one or more boxes of the flowchart and / or block diagram. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field programmable logic circuit. It can also be understood that each box in the block diagram and / or flowchart and the combination of boxes in the block diagram and / or flowchart can also be implemented by dedicated hardware that performs a specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions.

[0171] The above is only a specific implementation of the present application. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, modules and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. It should be understood that the protection scope of the present application is not limited to this. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in this application, and these modifications or replacements should be included in the protection scope of this application.

Claims

1. A device detection method, It is characterized in that The device is a device in a data center. The device has multiple types. The multiple device types correspond to multiple detection models one by one. The method includes: Acquire multiple first operating indicator data of the device to be detected; Based on the correspondence between the preset device type and the detection model, a target detection model corresponding to the device type of the device to be detected is determined from multiple pre-trained detection models, the detection model is trained based on a training sample set constructed based on multiple second operation indicator data of the corresponding device type, the second operation indicator data includes operation indicator data of normally operating devices and operation indicator data of faulty devices related to the device type; Inputting the plurality of first operating indicator data into the target detection model to obtain a detection result, wherein the detection result includes a probability prediction value of the device to be detected being in a target state, wherein the target state includes a normal state and / or a degraded state; According to the detection result, the operating status of the device to be detected is determined.

2. The detection method according to claim 1, It is characterized in that The method further comprises: For each pre-trained detection model, calculate the importance value of each feature, each feature corresponds to a second operation indicator data, and the importance value is used to characterize the importance of each feature to the detection result output by the pre-trained detection model; After determining the operating status of the device to be detected according to the detection result, the method further includes: According to the operating state of the device to be detected and the importance value of each feature in the target detection model, a detection result explanation for the device to be detected is generated, and the detection result explanation includes the influence degree of each feature on the operating state of the device to be detected.

3. The detection method according to claim 2, It is characterized in that The detection model is trained using a decision tree algorithm, the training sample set includes a plurality of samples, and for each pre-trained detection model, the importance value of each feature is calculated, including: For each pre-trained detection model, do the following: For each node of the decision tree corresponding to the detection model, obtain the features included in the node and the number of positive and negative samples corresponding to each feature; For each node, the Gini impurity value of the node is calculated according to the features included in the node and the number of positive and negative samples corresponding to each feature, and the weight of the node is calculated according to the Gini impurity value and the sample ratio, where the sample ratio is the ratio of the number of samples on the node to the total number of samples, and the Gini impurity value is used to characterize the degree of mixing of samples of different categories in the node; For each feature, the sum of the weights of the feature on all nodes is calculated to obtain the importance value of the feature.

4. The detection method according to any one of claims 1 to 3, It is characterized in that Determining the operating status of the device to be detected according to the detection result includes: When the detection result is a probability prediction value that the device to be detected is in a normal state, and the probability prediction value is greater than a first preset threshold, determining that the operating state of the device to be detected is in a normal state; When the detection result is a probability prediction value that the device to be detected is in a degraded state, and the probability prediction value is greater than a second preset threshold, determining that the operating state of the device to be detected is in a degraded state; The first preset threshold and the second preset threshold are determined based on the operating characteristics of the device type to which the device to be detected belongs and the third operating indicator data of the device to be detected within a preset time.

5. The detection method according to any one of claims 1 to 3, It is characterized in that Before inputting the plurality of first operation indicator data into the target detection model to obtain the detection result, the method further includes: For each device type, do the following: Build an initial detection model; Acquire a training sample set, the training sample set comprising a plurality of training samples, each of the training samples comprising a plurality of second operating indicator data of a device corresponding to the device type and a label of whether the device is in a normal state; The initial detection model is trained using the multiple training samples until a training stop condition is met to obtain a trained detection model.

6. The detection method according to claim 5, It is characterized in that The constructing of the initial detection model comprises: It is beneficial for LightGBM algorithm to build decision tree model; The decision tree model is determined as an initial detection model.

7. A detection device for an equipment, It is characterized in that The device is a device in a data center. The device has multiple types, and the multiple device types correspond to multiple detection models one by one. The apparatus includes: A first acquisition module, used to acquire a plurality of first operating indicator data of the device to be detected; A first determination module is used to determine a target detection model corresponding to the device type of the device to be detected from a plurality of pre-trained detection models based on a correspondence between preset device types and detection models, wherein the detection model is trained based on a training sample set constructed based on a plurality of second operation indicator data of the corresponding device type, wherein the second operation indicator data includes operation indicator data of a normally operating device and operation indicator data of a faulty device related to the device type; A second determination module is used to input the plurality of first operation indicator data into the target detection model to obtain a detection result, wherein the detection result includes a probability prediction value of the device to be detected being in a target state, wherein the target state includes a normal state and / or a degraded state; The third determination module is used to determine the operating status of the device to be detected according to the detection result.

8. An electronic device, It is characterized in that The device comprises: a processor and a memory storing computer program instructions; When the processor executes the computer program instructions, the detection method of the device according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores computer program instructions, and when the computer program instructions are executed by a processor, the detection method of the device according to any one of claims 1 to 6 is implemented.

10. A computer program product, It is characterized in that The invention comprises a computer program, wherein when the computer program is processed and executed, the detection method of the device according to any one of claims 1 to 6 is implemented.

Citation Information

Cited By

  • Industrial park equipment operation state management method and system

    CN119379272A

  • Equipment detection method and device, electronic equipment, medium and program product

    CN119666412A