Method and device for generating aging failure prediction model of component

Abstract

The invention discloses a method and device for generating an aging failure prediction model of a component and electronic equipment, and relates to the technical field of electronic equipment. Multiple groups of state data of the component under multiple tests are adopted when the aging failure prediction model is generated; therefore, the aging failure prediction model generated by the invention can predict the residual life of the component in different working environments, and the limitation that the aging failure prediction model in the related technology can only realize aging prediction of the component in a single working environment is improved; and the accuracy of predicting the residual life of the component by the aging failure prediction model is improved.

Description

Technical Field

[0001] The present application relates to the technical field of electronic devices, and particularly to a method and device for generating an aging failure prediction model of components. Background Art

[0002] With the continuous increase in the shipment volume of server products, the failure rate of servers is increasing day by day. By studying the aging of specific components in the server, the remaining life of the components can be predicted, and the probability of component failure in the future can be predicted through the remaining life of the components, which can avoid situations such as server downtime caused by component failures.

[0003] Currently, in the related art, for the aging study of components in the server, mainly through a pre-trained aging failure prediction model to predict the aging of components to obtain the remaining life of the components. However, the aging failure prediction model in the related art can only realize the aging prediction of components in a single working environment, resulting in greater limitations in the use of the aging failure prediction model. Summary of the Invention

[0004] The present application provides a method and device for generating an aging failure prediction model of components, which can improve the limitations of the aging failure prediction model and improve the accuracy of aging failure prediction of components.

[0005] The present application provides a method for generating an aging failure prediction model of components, including:

[0006] Collecting multiple groups of state data of experimental components under multiple test trials, where the multiple test trials are used to make the experimental components in multiple abnormal working states;

[0007] Selecting multiple aging-related data from each group of state data;

[0008] For the aging-related data corresponding to each group of state data, marking the remaining life of the experimental components corresponding to the multiple aging-related data, obtaining an original sample data set composed of the aging-related data and the corresponding remaining life of the experimental components;

[0009] Dividing the original sample data set into a training sample data set and a verification sample data set;

[0010] Constructing an initial aging failure prediction model;

[0011] Training the initial aging failure prediction model based on the training sample data set to obtain a trained aging failure prediction model;

[0012] Based on the verification sample data set, verify the trained aging failure prediction model; if the verification passes, obtain the trained aging failure prediction model; if the verification fails, continue to collect a new original sample data set and continue to train the trained aging failure prediction model continuously.

[0013] This application also provides a method for predicting the aging failure of components, including:

[0014] Collect multi-source operation data of the component to be predicted, where the multi-source operation data includes multiple aging-related data;

[0015] Input the multiple aging-related data into the aging failure prediction model generated as described above to obtain the aging failure prediction result corresponding to the component to be predicted.

[0016] This application also provides a device for generating an aging failure prediction model of components, including:

[0017] A collection module for collecting multiple groups of status data of experimental components under various test trials, where the various test trials are used to make the experimental components in various abnormal working states;

[0018] A processing module for screening out multiple aging-related data from each group of status data;

[0019] The processing module is also used to mark the remaining life of the experimental components corresponding to the multiple aging-related data for each group of status data, and obtain the original sample data set composed of the aging-related data and the corresponding remaining life of the experimental components;

[0020] The processing module is also used to divide the original sample data set into a training sample data set and a verification sample data set;

[0021] The processing module is also used to construct an initial aging failure prediction model;

[0022] The processing module is also used to train the initial aging failure prediction model based on the training sample data set to obtain the trained aging failure prediction model;

[0023] The processing module is also used to verify the trained aging failure prediction model based on the verification sample data set; if the verification passes, obtain the trained aging failure prediction model; if the verification fails, continue to collect a new original sample data set and continue to train the trained aging failure prediction model continuously.

[0024] This application also provides a device for predicting the aging failure of components, including:

[0025] The acquisition module is used to acquire multi-source operation data of the component to be predicted, where the multi-source operation data includes multiple aging-related data;

[0026] The processing module is used to input multiple aging-related data into the aging failure prediction model generated as described in any one of the preceding items to obtain the aging failure prediction result corresponding to the component to be predicted.

[0027] This application also provides an electronic device, including: a memory for storing a computer program; a processor for implementing the steps of the method for generating the aging failure prediction model of any one of the above components and the steps of the method for predicting the aging failure of the component when executing the computer program.

[0028] This application also provides a computer-readable storage medium, in which a computer program is stored, and wherein the computer program implements the steps of the method for generating the aging failure prediction model of any one of the above components and the steps of the method for predicting the aging failure of the component when executed by a processor.

[0029] This application also provides a computer program product, including a computer program, and the computer program implements the steps of the method for generating the aging failure prediction model of any one of the above components and the steps of the method for predicting the aging failure of the component when executed by a processor.

[0030] Through the method and device for generating the aging failure prediction model of the component provided by this application, since multiple groups of state data of the component under multiple test trials are used when generating the aging failure prediction model, the aging failure prediction model generated by this application can predict the remaining life of the component in different working environments, improving the limitation that the aging failure prediction model in the related art can only achieve the aging prediction of the component in a single working environment, and further improving the accuracy of the aging failure prediction model in predicting the remaining life of the component. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] To more clearly illustrate the embodiments of this application, the following will briefly introduce the drawings required for the embodiments. Obviously, the drawings in the following description are only some embodiments of this application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0032] Figure 1 Scene schematic provided by this application Figure 1 ;

[0033] Figure 2 Scene process provided by this application Figure 2 ;

[0034] Figure 3Flow diagram of the method for generating the aging failure prediction model of the components provided in the embodiments of the application Figure 1 ;

[0035] Figure 4 Flow diagram of the method for generating the aging failure prediction model of the components provided in the embodiments of the application Figure 2 ;

[0036] Figure 5 Schematic diagram of data collection for the example;

[0037] Figure 6 Flow diagram of the method for generating the aging failure prediction model of the components provided in the embodiments of the application Figure 3 ;

[0038] Figure 7 Flow diagram of the method for generating the aging failure prediction model of the components provided in the embodiments of the application Figure 4 ;

[0039] Figure 8 Distribution diagram of data points of temperature data and the remaining life of the experimental components corresponding thereto for the example;

[0040] Figure 9 Flow diagram of the method for generating the aging failure prediction model of the components provided in the embodiments of the application Figure 5 ;

[0041] Figure 10 Distribution diagram of data points of ESR and the remaining life of the experimental components corresponding thereto for the example;

[0042] Figure 11 Schematic diagram of the verification result;

[0043] Figure 12 Flow diagram of the method for generating the aging failure prediction model of the components provided in the embodiments of the application Figure 6 ;

[0044] Figure 13 Flow chart of the implementation of the aging failure prediction model for the example;

[0045] Figure 14 Data processing flow chart;

[0046] Figure 15 Flow diagram of the aging failure prediction method for the components provided in the embodiments of the application Figure 1 ;

[0047] Figure 16 Schematic diagram of the component data recording flow for the example;

[0048] Figure 17Schematic structural diagram of a device for generating an aging failure prediction model of components provided by an embodiment of the present application;

[0049] Figure 18 Schematic structural diagram of an aging failure prediction device for components provided by an embodiment of the present application;

[0050] Figure 19 Schematic structural diagram of an electronic device provided by the present application. Detailed implementation manners

[0051] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the protection scope of the present application.

[0052] It should be noted that in the description of the present application, the terms "include", "comprise" or any other variant thereof are intended to cover a non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. The terms "first", "second", etc. in the present application are used to distinguish similar objects, rather than to describe a specific order or sequence.

[0053] The present application collects the state data of components under various experiments, obtains the remaining life of components corresponding to multiple aging-related data, and trains the aging failure prediction model with the remaining life of components corresponding to multiple aging-related data, so that the aging failure prediction model can accurately predict the remaining life of components in different working environments.

[0054] To enable those skilled in the art of the present technology to better understand the solutions of the present application, the present application will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners.

[0055] Combined with the specific application environment architecture or specific hardware architecture on which the execution of the method for generating an aging failure prediction model of components and the method for predicting aging failure of components depends, the specific application environment architecture or specific hardware architecture will be described herein. Refer to Figure 1 , Figure 1 Scene schematic provided by the present application Figure 1 , such as Figure 1As shown, this application mainly focuses on the research of aging failure of components in a server. Aging-related data of components can be collected through sensors, and the aging-related data of components is input into the aging failure prediction model of components. The remaining life of components is output through the aging failure prediction model of components. Figure 2 The scenario process provided by this application Figure 2 , such as Figure 2 As shown, components in a server will age due to the influence of aging factors in the operating environment. For example, affected by temperature, humidity, and vibration, the components in the server will gradually age. At this time, if there is no aging failure prediction model for components, the remaining life of components cannot be predicted. When components fail due to aging, there will be a direct failure or downtime. If there is an aging failure prediction model for components, the remaining life of components can be predicted through the aging failure prediction model according to the aging-related data of components in the actual operating environment, and an early warning can be made based on the remaining life of components to determine whether it is necessary to replace components or prepare a backup plan to deal with component failure to ensure the normal operation of the server.

[0056] Figure 3 The flow schematic of the method for generating the aging failure prediction model of components provided by the embodiment of this application Figure 1 , such as Figure 3 As shown, the embodiment of this application provides a method for generating an aging failure prediction model of components. The method is described in detail as follows:

[0057] S301: Collect multiple groups of state data of experimental components under multiple test trials, where the multiple test trials are used to make the experimental components in multiple abnormal working states.

[0058] Combined with the scenario example, the test trial can be a failures in time (FIT) test or a burn-in prevention board test. Among them, the FTI test can perform operations such as short circuit, overvoltage, and overcurrent on the experimental components, and the burn-in prevention board test can be a salt water dripping test. The FTI test or the burn-in prevention board test can make the test components work in abnormal working states, such as abnormal working states such as overvoltage, overcurrent, or high temperature. Multiple groups of state data corresponding to the experimental components under the FIT test or the burn-in prevention board test can be collected through sensors. The state data can include parameters such as the voltage, temperature, Equivalent Series Resistance (ESR), current, vibration magnitude corresponding to the experimental components, and the remaining life of the experimental components under different state data.

[0059] S302: Screen out multiple aging-related data from each group of state data.

[0060] Combined with the scenario example, the status data includes multiple data that can cause the aging and damage of test components. For example, excessive temperature, excessive vibration intensity, or excessive humidity can all cause the aging and damage of test components. These data that can cause the aging and damage of test components can be determined as aging-related data.

[0061] S303: For the aging-related data corresponding to the status data of each group, mark the remaining life of the test components corresponding to the multiple aging-related data, and obtain the original sample data set composed of the aging-related data and the corresponding remaining life of the test components.

[0062] Combined with the scenario example, the degree of aging and damage of test components caused by each type of aging-related data is different. Therefore, according to each type of aging-related data, mark the remaining life of the test components corresponding to this aging-related data. Combine each type of aging-related data and the corresponding remaining life of the test components to obtain the original sample data set.

[0063] S304: Divide the original sample data set into a training sample data set and a validation sample data set.

[0064] Combined with the scenario example, the training sample data set includes multiple types of aging-related data and the corresponding remaining life of the test components. One or more target aging-related data that meet the preset conditions can be determined from the multiple types of aging-related data. The target aging-related data and the corresponding remaining life of the test components can be determined as the validation sample data set, and the other aging-related data and the corresponding remaining life of the test components except the target aging-related data can be determined as the training sample data set. The training sample data set is used for training the model, and the validation sample data set is used to verify the trained model to determine whether the trained model can accurately predict the remaining life of the components.

[0065] S305: Construct an initial aging failure prediction model.

[0066] Combined with the scenario example, the model parameters in the initial aging failure prediction model are randomly generated. The aging failure prediction model is mainly used to predict the remaining life of the components based on the aging-related data of the components in the actual operating environment.

[0067] S306: Based on the training sample data set, train the initial aging failure prediction model to obtain the trained aging failure prediction model.

[0068] Combined with the scenario example, the initial aging failure prediction model can be trained through a machine learning algorithm, which can be the Support Vector Machine (SVM for short). Specifically, each piece of aging-related data included in the training sample dataset and the remaining life of the corresponding experimental components can be input into the initial aging failure prediction model to update the model parameters in the initial aging failure prediction model, so as to obtain the trained aging failure prediction model.

[0069] S307: Based on the validation sample dataset, validate the trained aging failure prediction model; if the validation passes, obtain the trained aging failure prediction model; if the validation fails, continue to collect a new original sample dataset and continue to train the trained aging failure prediction model.

[0070] Combined with the scenario example, the trained aging failure prediction model is validated through the validation sample dataset, and the validation result is fed back into the model to optimize the trained aging failure prediction model. Specifically, the remaining life of the experimental component corresponding to the target aging data in the validation sample dataset can be used to determine whether the remaining life of the experimental component obtained through the trained aging failure prediction model is within the correct range. If the remaining life of the experimental component obtained through the trained aging failure prediction model is within the correct range, it is determined that the validation of the trained aging failure prediction model passes, and at this time, the trained aging failure prediction model is determined as the trained aging failure prediction model. On the contrary, if the remaining life of the experimental component obtained through the trained aging failure prediction model is not within the correct range, it is determined that the validation of the trained aging failure prediction model fails. At this time, it is necessary to continue to collect a new original sample dataset and repeat the above steps until the remaining life of the experimental component obtained through the trained aging failure prediction model is within the correct range.

[0071] Based on the method provided in this application, the generated aging failure prediction model can predict the remaining life of components under different working environments, improving the limitation of the aging failure prediction model in the related technology that can only realize the aging prediction of components in a single working environment, and thus improving the accuracy of the aging failure prediction model in predicting the remaining life of components.

[0072] Optionally, the aging-related data includes temperature data, humidity data, vibration data, and equivalent series resistance value;

[0073] Correspondingly, Figure 4 The flow chart of the method for generating the aging failure prediction model of components provided by the embodiment of this application Figure 2 is shown in Figure 4 as follows. S304 includes:

[0074] S401: Determine the data set composed of temperature data, humidity data, and vibration data and the remaining life of the corresponding experimental components as the training sample data set.

[0075] Combined with the scenario example, if the temperature in the server chassis increases, it will cause the internal electronic thermal motion speed of the components to increase, migrate under the influence of the electric field, making the components in the server vulnerable and easy to break at certain positions. Long-term heating and cooling will cause the solder joints of the components in the server to loosen during thermal expansion and contraction, resulting in loosening or breakage. The high-speed rotation of the fan in the chassis will transmit vibration to the surrounding area. During the continuous vibration process, cracks and deformations will occur in the packaging and structure of the components in the server, resulting in continuous aging of the insulation layer material and causing breakdown. If the server operates in a humid area, water vapor in the air will continuously penetrate into the chassis and cause erosion. Seriously, it will lead to a short circuit, but usually, it will reduce the insulation and reliability of the components in the server under continuous erosion. Therefore, temperature data, humidity data, and vibration data can be determined as aging-related data that cause aging and damage to the components. And temperature, humidity, and vibration will cause changes in the ESR data of the components. Therefore, ESR data can also be determined as aging-related data that cause aging and damage to the components.

[0076] Figure 5 Schematic diagram of data acquisition for the example, as Figure 5 shown, collect the real-time temperature data, humidity data, and vibration data of the experimental components through a temperature sensor, a humidity sensor, and a vibration sensor respectively. The real-time ESR data of the experimental components can be obtained through ESR detection, and the real-time voltage and current during the operation of the experimental components can also be detected simultaneously.

[0077] In the FTI test or the anti-burn board test experiment, use a digital multimeter to measure the relationship data between the temperature data, humidity data, vibration data, and ESR data and the remaining life of the experimental components respectively. For example, the relationship between the temperature data and the remaining life of the experimental components can be: the remaining life corresponding to a temperature of 40 °C, or the remaining life corresponding to a temperature of 50 °C, etc.

[0078] Transmit the collected temperature data, humidity data, vibration data, ESR data, voltage and current, as well as the relationship data between the temperature data, humidity data, vibration data, ESR data and the remaining life of the experimental components to the Baseboard Management Controller (BMC). The BMC can first preprocess the collected data, including data screening to eliminate unreasonable data, and also including data normalization and other processes. Since the ESR data is affected by the temperature data, humidity data and vibration data at the same time, the ESR data can be determined as the target aging-related data. Therefore, the relationship data between the processed temperature data, humidity data and vibration data and the remaining life of the experimental components can be determined as the training sample data set.

[0079] S402: Determine the data set composed of the equivalent series resistance value and the remaining life of the corresponding experimental components as the verification sample data set.

[0080] Combined with the scenario example, similarly, the relationship data between the ESR data and the remaining life of the experimental components can be determined as the verification sample data set.

[0081] Since the ESR data is affected by the temperature data, humidity data and vibration data simultaneously in three aspects, based on the method provided in this example, determining the relationship data between the ESR data and the remaining life of the experimental components as the verification set can make the verification result of the aging failure prediction model of the components more accurate.

[0082] Optionally, Figure 6 is the flow schematic of the method for generating the aging failure prediction model of the components provided in the embodiments of the present application Figure 3 as Figure 6 shown, S305 includes:

[0083] S601: Determine the first weight corresponding to the temperature data and the initial temperature acceleration factor, the second weight corresponding to the humidity data and the initial humidity acceleration factor, and the third weight corresponding to the vibration data and the initial vibration acceleration factor.

[0084] Combined with the scenario example, the first weight can be denoted as X1, the initial temperature acceleration factor as Qv, the second weight as X2, the initial humidity acceleration factor as Qh, the third weight as X3, and the initial vibration acceleration factor as Qm.

[0085] S602: Perform weighted summation on the initial temperature acceleration factor, initial humidity acceleration factor and initial vibration acceleration factor through the corresponding first weight, second weight and third weight respectively to obtain the summation result.

[0086] Combined with the scenario example, the summation result is: Qv×X1 + Qh×X2 + Qm×X3.

[0087] S603: Multiply the sum result by the rated life of the preset experimental component to obtain a product result.

[0088] Combined with the scenario example, the rated life of the experimental component can be denoted as T, so the obtained product result is: T×(Qv×X1 + Qh×X2 + Qm×X3).

[0089] S604: Determine the product result as the aging failure prediction result of the experimental component to construct an initial aging failure prediction model.

[0090] Combined with the scenario example, the aging failure prediction result of the experimental component can be denoted as T0, so the constructed initial aging failure prediction model is:

[0091] T0 = T×(Qv×X1 + Qh×X2 + Qm×X3).

[0092] The aging failure prediction model of the component constructed based on the method provided in this example fully considers various aging-related data, and can improve the accuracy of the remaining life of the component predicted by the aging failure prediction model of the component.

[0093] Optionally, Figure 7 is a flowchart of the method for generating the aging failure prediction model of the component provided in the embodiment of the present application Figure 4 as Figure 7 shown, S306 includes:

[0094] S701: Fit the temperature data with the corresponding remaining life of the experimental component to obtain a temperature regression equation.

[0095] Combined with the scenario example, Figure 8 is a distribution diagram of data points of the temperature data of the example and the corresponding remaining life of the experimental component. Add the temperature data of each training sample dataset and the data of the corresponding remaining life of the experimental component to the coordinate system, and data points of each temperature data and the corresponding remaining life of the experimental component can be obtained to obtain Figure 8 . By fitting the regression line, the data points are fitted to obtain the temperature regression equation corresponding to the temperature data.

[0096] S702: Fit the humidity data with the corresponding remaining life of the experimental component to obtain a humidity regression equation.

[0097] Combined with the scenario example, similarly, add each humidity data in the training sample dataset and the data of the remaining life of the corresponding experimental component to the coordinate system, and data points of each humidity data and the corresponding remaining life of the experimental component can be obtained. By fitting the regression line, the data points are fitted to obtain the humidity regression equation corresponding to the humidity data.

[0098] S703: Fit the vibration data with the remaining life of the corresponding experimental component to obtain the vibration regression equation.

[0099] Combined with the scenario example, similarly, add each vibration data in the training sample dataset and the data of the remaining life of the corresponding experimental component to the coordinate system, and data points of each vibration data and the corresponding remaining life of the experimental component can be obtained. By fitting the regression line, the data points are fitted to obtain the humidity regression equation corresponding to the vibration data.

[0100] S704: Determine the first influence factor of temperature on the material properties of the experimental component, the second influence factor of humidity on the electron mobility of the experimental component, and the third influence factor of vibration on the vibration intensity of the experimental component.

[0101] Combined with the scenario example, temperature affects the material properties of the experimental component, so the magnitude of the influence of temperature on the material properties can be determined according to the actual situation and determined as the first influence factor. Similarly, humidity affects the electron mobility of the experimental component, so the magnitude of the influence of humidity on the electron mobility can be determined according to the actual situation and determined as the second influence factor. The magnitude of vibration affects the vibration intensity of the experimental component, so the magnitude of the influence of the vibration magnitude on the vibration intensity can be determined according to the actual situation and determined as the third influence factor.

[0102] S705: Apply the first influence factor to the temperature regression equation to obtain the target temperature acceleration factor.

[0103] Combined with the scenario example, the first influence factor can represent the magnitude of the influence of temperature on the material properties, and the temperature regression equation can represent the magnitude of the influence of temperature on the remaining life of the experimental component. Therefore, by applying the first influence factor to the temperature regression equation and using temperature as the conversion medium, the magnitude of the influence of the material properties on the remaining life of the experimental component can be obtained, and the magnitude of the influence of the material properties on the remaining life of the experimental component is determined as the target temperature acceleration factor.

[0104] S706: Apply the second influence factor to the humidity regression equation to obtain the target humidity acceleration factor.

[0105] Combined with the scenario example, similarly, the second influence factor can characterize the magnitude of the influence of humidity on the electron mobility, and the humidity regression equation can characterize the magnitude of the influence of humidity on the remaining life of the experimental components. Therefore, by applying the second influence factor to the humidity regression equation and using humidity as the conversion medium, the magnitude of the influence of electron mobility on the remaining life of the experimental components can be obtained, and the magnitude of the influence of electron mobility on the remaining life of the experimental components is determined as the target humidity acceleration factor.

[0106] S707: Apply the third influence factor to the vibration regression equation to obtain the target vibration acceleration factor.

[0107] Combined with the scenario example, similarly, the third influence factor can characterize the magnitude of the influence of vibration magnitude on the vibration intensity, and the vibration regression equation can characterize the magnitude of the influence of vibration on the remaining life of the experimental components. Therefore, by applying the third influence factor to the vibration regression equation and using vibration as the conversion medium, the magnitude of the influence of vibration intensity on the remaining life of the experimental components can be obtained, and the magnitude of the influence of vibration intensity on the remaining life of the experimental components is determined as the target vibration acceleration factor.

[0108] S708: Update the initial temperature acceleration factor, initial humidity acceleration factor, and initial vibration acceleration factor in the initial aging failure prediction model through the target temperature acceleration factor, target humidity acceleration factor, and target vibration acceleration factor to obtain the trained aging failure prediction model.

[0109] Combined with the scenario example, the target temperature acceleration factor can be denoted as Qv’, the target humidity acceleration factor can be denoted as Qh’, and the target vibration acceleration factor can be denoted as Qm’. Therefore, the obtained trained aging failure prediction model is:

[0110] T0 = T × (Qv’ × X1 + Qh’ × X2 + Qm’ × X3).

[0111] Based on the method provided in this example, through actual experimental data and the influence factors of various aging-related data on the components in actual situations, the obtained target acceleration factors are more in line with the actual situation, enabling the trained prediction model to more accurately predict the remaining life of the components.

[0112] Figure 9 Flow schematic of the method for generating the aging failure prediction model of the components provided in the embodiments of the present application Figure 5 , as Figure 9 shown, S307 includes:

[0113] S901: Determine the experimental aging failure prediction result corresponding to the trained aging failure prediction model.

[0114] Combined with the scenario example, Figure 10For the distribution diagram of data points of ESR as an example and the remaining life of the corresponding experimental components, adding the data of each ESR and the remaining life of the corresponding experimental components in the verification sample dataset to the coordinate system can obtain the data points of each ESR and the remaining life of the corresponding experimental components, so as to obtain Figure 10 . At the same time, determine T0 corresponding to the trained aging failure prediction model as the experimental aging failure prediction result.

[0115] S902: Determine the difference between the remaining life of the experimental component corresponding to each equivalent series resistance value in the verification sample dataset and the experimental aging failure prediction result.

[0116] Combined with the scenario example, Figure 11 For the schematic diagram of the verification result, fit the straight line corresponding to the experimental aging failure prediction result to Figure 10 to obtain Figure 11 . According to Figure 11 , the distance between each data point and the straight line corresponding to the experimental aging failure prediction result can be determined, and the difference between the remaining life of the experimental component corresponding to each equivalent series resistance value and the experimental aging failure prediction result can be determined through the distance between each data point and the straight line corresponding to the experimental aging failure prediction result. The unit of the remaining life in the figure is days, so the difference represents days.

[0117] S903: If the difference meets the preset conditions, it is determined that the verification of the trained aging failure prediction model passes; otherwise, it is determined that the verification of the trained aging failure prediction model fails.

[0118] Combined with the scenario example, the preset conditions can be that the number of differences less than the preset number of days exceeds the preset ratio, where both the preset number of days and the preset ratio can be determined according to the actual situation. For example, the preset number of days can be determined as 3 days, and the preset ratio can be determined as 95%. Therefore, when the number of differences less than 3 days exceeds 95%, it indicates that the remaining life of the experimental components predicted by the trained aging failure prediction model is relatively accurate, and it can be determined that the verification of the trained aging failure prediction model passes. Otherwise, when the number of differences less than 3 days does not exceed 95%, it indicates that the remaining life of the experimental components predicted by the trained aging failure prediction model is inaccurate, and it can be determined that the verification of the trained aging failure prediction model fails.

[0119] Based on the method provided in this example, using the data between the actual ESR data and the remaining life of the experimental components to predict the trained aging failure prediction model can make the verification result of the model more accurate.

[0120] Optionally, before continuing to collect a new original sample dataset to continue training the trained aging failure prediction model, it also includes:

[0121] Update the first weight, the second weight, and the third weight in the trained aging failure prediction model.

[0122] Combined with the scenario example, since the first weight, the second weight, and the third weight in the trained aging failure prediction model are randomly generated, the first weight, the second weight, and the third weight can be updated according to the verification situation of the trained aging failure prediction model, and then the next round of training can be carried out until the verification of the trained aging failure prediction model passes with the updated first weight, second weight, and third weight. Then, the most recently updated first weight, second weight, and third weight are determined as the final target first weight, target second weight, and target third weight. Then, the target temperature acceleration factor denoted as Qv', the target humidity acceleration factor denoted as Qh', and the target vibration acceleration factor denoted as Qm' are weighted and summed through the final target first weight, target second weight, and target third weight to obtain the trained aging failure prediction model.

[0123] Based on the method provided in this example, updating unreasonable weights can improve the training efficiency of the aging failure prediction model.

[0124] Optionally, Figure 12 is a flowchart showing the generation method of the aging failure prediction model for components provided in the embodiments of this application Figure 6 , as Figure 12 shown, further includes:

[0125] S1201: Collect the current of the experimental component at different historical moments and the adjustment relationship between the voltage and the current.

[0126] Combined with the scenario example, Figure 13 is a flowchart showing the implementation of the aging failure prediction model of the example. As Figure 13 shown, when the temperature of the component increases, material changes and intensified electronic activities will occur. When the humidity increases, it will cause corrosion of the component, resulting in a decrease in insulation ability. Vibration will cause problems such as loose solder joints and device fractures of the component. Figure 14 is a data processing flowchart. As Figure 14As shown, temperature data, humidity data, and vibration data can be collected through temperature sensors, humidity sensors, and vibration sensors to obtain aging-related data. In addition, changes in the load of server components will cause changes in the internal current and corresponding voltage. Therefore, the current of the experimental components at different historical moments and the adjustment relationship of the corresponding voltage change with the current can be collected through the load changes at different historical moments, and other voltage-related data can be obtained. The data collected above, as well as the ESR data, are transmitted to the BMC. After preprocessing the data, the BMC divides it into a verification sample data set and a training sample data set. The acceleration factors corresponding to temperature, humidity, and vibration can be calculated through the training sample data set to train the aging failure prediction model. At the same time, the trained aging failure prediction model is verified based on the verification sample data set. If the verification fails, the aging failure prediction model is continued to be trained until the verification passes, and the trained aging failure prediction model and voltage prediction model are obtained.

[0127] S1202: Construct an initial voltage prediction model.

[0128] Combined with the scenario example, for the training of the voltage prediction model, an initial voltage prediction model before training can be created first. The parameters in the initial voltage model can be randomly generated. The voltage prediction model is used to predict the corresponding voltage at different times.

[0129] S1203: Input the current of the experimental components at different historical moments and the adjustment relationship of the voltage with the current into the initial voltage prediction model to train the initial voltage prediction model to obtain a trained voltage prediction model.

[0130] Combined with the scenario example, input the appropriate currents of the experimental components and the adjustment relationship of the voltage with the current into the initial voltage prediction model to complete the training of the voltage prediction model.

[0131] Based on the method provided in this example, the purpose of constructing and training the voltage prediction model can be achieved.

[0132] Figure 15 Flow schematic of the aging failure prediction method for components provided in the embodiments of the present application Figure 1 , including:

[0133] S1501: Collect multi-source operation data of the components to be predicted, where the multi-source operation data includes multiple aging-related data.

[0134] Combined with the scenario example, the aging-related data can be temperature data, humidity data, and vibration data. The temperature data, humidity data, and vibration data of the component to be predicted can be collected in real time based on a temperature sensor, a humidity sensor, and a vibration sensor respectively, and the collected real-time temperature data, humidity data, and vibration data are determined as the multi-source operation data corresponding to the component to be predicted.

[0135] S1502: Input multiple aging-related data into the aging failure prediction model generated as described above to obtain the aging failure prediction result corresponding to the component to be predicted.

[0136] Combined with the scenario example, input the multi-source operation data corresponding to the component to be predicted into the trained aging failure prediction model. The aging failure prediction model can predict the remaining life of the component to be predicted based on the multi-source operation data corresponding to the component to be predicted, and the obtained remaining life of the component to be predicted can be determined as the aging failure prediction result corresponding to the component to be predicted.

[0137] Based on the method provided in this example, the aging failure prediction model predicts the remaining life of the component based on the multi-source operation data of the component. Since the multi-source operation data includes multiple aging-related data of the component, the accuracy of the predicted remaining life can be improved.

[0138] Similarly, the voltage prediction model can predict the voltages corresponding to different future times. Therefore, based on the voltage prediction model, a voltage prediction result can be obtained. The voltage prediction result includes the predicted voltage values corresponding to different future times, and based on the predicted voltage values corresponding to different future times, the voltage values at different future times can be adjusted in advance to the corresponding predicted voltage values.

[0139] Combined Figure 13 and Figure 14 , according to the aging failure prediction result predicted by the aging failure prediction model and the voltage prediction result predicted by the voltage prediction model, the aging failure prediction model and the voltage prediction model can be calibrated with reference to the actual life and voltage values of the actual component respectively to improve the accuracy of the prediction results corresponding to the aging failure prediction model and the voltage prediction model.

[0140] Figure 16For the component data record flow chart of the example, when aging prediction and voltage prediction of the remaining life of the component to be predicted are performed, the collected data can be divided into aging-related data and voltage-related data. Then, it can be first determined whether the component to be predicted is damaged. If the component to be predicted is not damaged, the aging test is performed on the predicted component to predict the remaining life and voltage change of the component to be predicted, and the remaining life and voltage change of the component to be predicted are recorded, and the remaining life and voltage change of the component to be predicted are recorded as the current aging data of the undamaged component. If the component to be predicted is damaged, the operating parameters corresponding to the component to be predicted are collected, including temperature data, humidity data, ESR, and voltage, and the operating parameters are recorded as the current aging data of the damaged component.

[0141] Through the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by means of software plus a necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases, the former is a better implementation manner.

[0142] Figure 17 It is a schematic structural diagram of a device for generating an aging failure prediction model of a component provided by an embodiment of the present application. As Figure 17 shown, the embodiment of the present application also provides a device for generating an aging failure prediction model of a component, including:

[0143] A collection module 171, configured to collect multiple groups of state data of an experimental component under multiple test trials, where the multiple test trials are used to make the experimental component in multiple abnormal working states;

[0144] A processing module 172, configured to screen out multiple aging-related data from each group of state data;

[0145] The processing module 172 is further configured to mark the remaining life of the experimental component corresponding to multiple aging-related data for the aging-related data corresponding to each group of state data, and obtain an original sample data set composed of the aging-related data and the corresponding remaining life of the experimental component;

[0146] The processing module 172 is further configured to divide the original sample data set into a training sample data set and a verification sample data set;

[0147] The processing module 172 is further configured to construct an initial aging failure prediction model;

[0148] The processing module 172 is further configured to train the initial aging failure prediction model based on the training sample data set to obtain a trained aging failure prediction model;

[0149] The processing module 172 is further configured to verify the trained aging failure prediction model based on the verification sample data set; if the verification is passed, the trained aging failure prediction model is obtained; if the verification fails, new original sample data sets are continuously collected to continuously train the trained aging failure prediction model.

[0150] Optionally, the processing module 172 is specifically configured to determine the data set composed of the temperature data, the humidity data, and the vibration data and the remaining life of the corresponding experimental component as the training sample data set;

[0151] The processing module 172 is specifically further configured to determine the data set composed of the equivalent series resistance value and the remaining life of the corresponding experimental component as the verification sample data set.

[0152] The processing module 172 is specifically further configured to determine the first weight and the initial temperature acceleration factor corresponding to the temperature data, the second weight and the initial humidity acceleration factor corresponding to the humidity data, and the third weight and the initial vibration acceleration factor corresponding to the vibration data;

[0153] The processing module 172 is specifically further configured to perform weighted summation on the initial temperature acceleration factor, the initial humidity acceleration factor, and the initial vibration acceleration factor through the corresponding first weight, second weight, and third weight respectively to obtain a summation result;

[0154] The processing module 172 is specifically further configured to multiply the summation result by the rated life of the preset experimental component to obtain a product result;

[0155] The processing module 172 is specifically further configured to determine the product result as the aging failure prediction result of the experimental component to construct an initial aging failure prediction model.

[0156] The processing module 172 is specifically further configured to fit the temperature data with the remaining life of the corresponding experimental component to obtain a temperature regression equation;

[0157] The processing module 172 is specifically further configured to fit the humidity data with the remaining life of the corresponding experimental component to obtain a humidity regression equation;

[0158] The processing module 172 is specifically further configured to fit the vibration data with the remaining life of the corresponding experimental component to obtain a vibration regression equation;

[0159] The processing module 172 is specifically further configured to determine the first influence factor of temperature on the material properties of the experimental component, the second influence factor of humidity on the electron mobility of the experimental component, and the third influence factor of vibration on the vibration intensity of the experimental component;

[0160] The processing module 172 is further configured to apply the first influence factor to the temperature regression equation to obtain a target temperature acceleration factor;

[0161] The processing module 172 is further configured to apply the second influence factor to the humidity regression equation to obtain a target humidity acceleration factor;

[0162] The processing module 172 is further configured to apply the third influence factor to the vibration regression equation to obtain a target vibration acceleration factor;

[0163] The processing module 172 is further configured to update the initial temperature acceleration factor, the initial humidity acceleration factor, and the initial vibration acceleration factor in the initial aging failure prediction model through the target temperature acceleration factor, the target humidity acceleration factor, and the target vibration acceleration factor, so as to obtain a trained aging failure prediction model.

[0164] The processing module 172 is further configured to update the first weight, the second weight, and the third weight in the trained aging failure prediction model.

[0165] The processing module 172 is further configured to collect the current of the experimental component at different historical moments, and the adjustment relationship between the voltage and the current;

[0166] The processing module 172 is further configured to construct an initial voltage prediction model;

[0167] The processing module 172 is further configured to input the current of the experimental component at different historical moments, and the adjustment relationship between the voltage and the current into the initial voltage prediction model to train the initial voltage prediction model, so as to obtain a trained voltage prediction model.

[0168] For the description of the features in the corresponding embodiments of the generation device of the aging failure prediction model of the component, reference may be made to the relevant descriptions in the corresponding embodiments of the generation method of the aging failure prediction model of the component, which will not be elaborated here one by one.

[0169] Figure 18 This is a schematic structural diagram of the aging failure prediction device for components provided by the embodiments of the present application. As Figure 18 shown, the embodiments of the present application further provide an aging failure prediction device for components, including:

[0170] An acquisition module 181, configured to acquire multi-source operation data of the component to be predicted, where the multi-source operation data includes a plurality of aging-related data;

[0171] A processing module 182, configured to input the plurality of aging-related data into the aging failure prediction model generated as described in any of the previous items to obtain an aging failure prediction result corresponding to the component to be predicted.

[0172] For the description of the features in the corresponding embodiments of the aging failure prediction device of components, reference can be made to the relevant descriptions in the corresponding embodiments of the aging failure prediction method of components, which will not be elaborated here one by one.

[0173] Figure 19 This is a schematic structural diagram of the electronic device provided by the present application. As Figure 19 shown, the electronic device 50 provided in this embodiment includes: at least one processor 501 and a memory 502. Optionally, the electronic device 50 further includes a communication component 503. Among them, the processor 501, the memory 502, and the communication component 503 are connected through a bus.

[0174] In the specific implementation process, at least one processor 501 executes the computer execution instructions stored in the memory 502, so that at least one processor 501 executes the above-mentioned embodiments of the method for generating an aging failure prediction model of components and the embodiments of the aging failure prediction method of components.

[0175] For the specific implementation process of the processor 501, reference can be made to the above method embodiments, and their implementation principles and technical effects are similar, which will not be elaborated here in this embodiment.

[0176] In the above embodiments, it should be understood that the processor may be a central processing unit (Central Processing Unit, abbreviated as: CPU), or other general-purpose processors, digital signal processors (Digital Signal Processor, abbreviated as: DSP), application specific integrated circuits (Application Specific Integrated Circuit, abbreviated as: ASIC), etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in combination with the application can be directly embodied as being completed by a hardware processor, or by a combination of hardware and software modules in the processor.

[0177] The memory may include a high-speed memory (Random Access Memory, RAM), and may also include a non-volatile memory (Non-volatile Memory, NVM), such as at least one disk memory.

[0178] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, the buses in the drawings of this application are not limited to only one bus or one type of bus.

[0179] An embodiment of the present application also provides a computer-readable storage medium, in which a computer program is stored. Wherein, the computer program is configured to execute the steps in any of the above XX method embodiments when running.

[0180] In an exemplary embodiment, the above computer-readable storage medium may include but is not limited to: USB flash drives, read-only memory (ROM for short), random access memory (RAM for short), external hard drives, magnetic disks, or optical discs and other various media that can store computer programs.

[0181] An embodiment of the present application also provides a computer program product. The above computer program product includes a computer program, and when the computer program is executed by a processor, it implements the steps in any of the above embodiments of the method for generating an aging failure prediction model of components and the method for predicting aging failure of components.

[0182] An embodiment of the present application also provides another computer program product, including a non-volatile computer-readable storage medium. The non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the steps in any of the above embodiments of the method for generating an aging failure prediction model of components and the method for predicting aging failure of components.

[0183] Those skilled in the art can further realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0184] The above has introduced in detail a method and device for generating an aging failure prediction model of a component provided in this application. Specific examples are used in this article to elaborate on the principle and implementation manner of this application. The description of the above embodiments is only used to help understand the method and its core idea of this application. It should be noted that for those of ordinary skill in the art, without departing from the principle of this application, several improvements and modifications can still be made to this application, and these improvements and modifications also fall within the protection scope of the claims of this application.

Claims

1. A method for generating an aging failure prediction model of a component, characterized in that, Including: Collecting multiple groups of status data of experimental components under multiple test trials, where the multiple test trials are used to make the experimental components in multiple abnormal working states; Screening out multiple aging-related data from each group of status data; For the aging-related data corresponding to each group of status data, marking the remaining life of the experimental components corresponding to the multiple aging-related data, and obtaining an original sample data set composed of the aging-related data and the corresponding remaining life of the experimental components; Dividing the original sample data set into a training sample data set and a validation sample data set; Constructing an initial aging failure prediction model; Training the initial aging failure prediction model based on the training sample data set to obtain a trained aging failure prediction model; Validating the trained aging failure prediction model based on the validation sample data set; if the validation passes, a trained aging failure prediction model is obtained; if the validation fails, continue to collect a new original sample data set and continue to train the trained aging failure prediction model.

2. The method according to claim 1, wherein The aging-related data includes temperature data, humidity data, vibration data, and equivalent series resistance value; Correspondingly, the dividing the original sample data set into a training sample data set and a validation sample data set includes: Determining the data set composed of the temperature data, humidity data, and vibration data respectively and the corresponding remaining life of the experimental components as the training sample data set; Determining the data set composed of the equivalent series resistance value and the corresponding remaining life of the experimental components as the validation sample data set.

3. The method according to claim 2, wherein The constructing an initial aging failure prediction model includes: Determining a first weight corresponding to the temperature data and an initial temperature acceleration factor, a second weight corresponding to the humidity data and an initial humidity acceleration factor, and a third weight corresponding to the vibration data and an initial vibration acceleration factor; Performing weighted summation on the initial temperature acceleration factor, initial humidity acceleration factor, and initial vibration acceleration factor through the corresponding first weight, second weight, and third weight respectively to obtain a summation result; Multiplying the summation result by the rated life of the preset experimental components to obtain a product result; Determining the product result as the aging failure prediction result of the experimental components to construct the initial aging failure prediction model.

4. The method according to claim 3, wherein The training the initial aging failure prediction model based on the training sample data set to obtain a trained aging failure prediction model includes: Fitting the temperature data with the corresponding remaining life of the experimental components to obtain a temperature regression equation; Fitting the humidity data with the corresponding remaining life of the experimental components to obtain a humidity regression equation; Fitting the vibration data with the corresponding remaining life of the experimental components to obtain a vibration regression equation; Determining a first influence factor of temperature on the material properties of the experimental components, a second influence factor of humidity on the electron mobility of the experimental components, and a third influence factor of vibration on the vibration intensity of the experimental components; Applying the first influence factor to the temperature regression equation to obtain a target temperature acceleration factor; Apply the second influence factor to the humidity regression equation to obtain a target humidity acceleration factor; Apply the third influence factor to the vibration regression equation to obtain a target vibration acceleration factor; Update the initial temperature acceleration factor, initial humidity acceleration factor, and initial vibration acceleration factor in the initial aging failure prediction model through the target temperature acceleration factor, target humidity acceleration factor, and target vibration acceleration factor to obtain a trained aging failure prediction model.

5. The method according to claim 1, wherein Before continuing to collect a new original sample data set and continue to continuously train the trained aging failure prediction model, it further includes: Update the first weight, second weight, and third weight in the trained aging failure prediction model.

6. The method according to claim 1, wherein It further includes: Collect the current of the experimental component at different historical moments and the adjustment relationship between voltage and current; Construct an initial voltage prediction model; Input the current of the experimental component at different historical moments and the adjustment relationship between voltage and current into the initial voltage prediction model to train the initial voltage prediction model to obtain a trained voltage prediction model.

7. An aging failure prediction method for components, characterized in that, It includes: Collect multi-source operation data of the component to be predicted, where the multi-source operation data includes multiple aging-related data; Input the multiple aging-related data into the aging failure prediction model generated according to any one of claims 1-6 to obtain an aging failure prediction result corresponding to the component to be predicted.

8. An apparatus for generating an aging failure prediction model of a component, characterized in that, It includes: A collection module for collecting multiple groups of state data of the experimental component under various test trials, where the various test trials are used to make the experimental component in various abnormal working states; A processing module for screening out multiple aging-related data from each group of state data; The processing module is further configured to mark the remaining life of the experimental component corresponding to the multiple aging-related data for each group of state data corresponding aging-related data to obtain an original sample data set composed of the aging-related data and the corresponding remaining life of the experimental component; The processing module is further configured to divide the original sample data set into a training sample data set and a verification sample data set; The processing module is further configured to construct an initial aging failure prediction model; The processing module is further configured to train the initial aging failure prediction model based on the training sample data set to obtain a trained aging failure prediction model; The processing module is further configured to verify the trained aging failure prediction model based on the verification sample data set; if the verification passes, a trained aging failure prediction model is obtained; if the verification fails, continue to collect a new original sample data set and continue to continuously train the trained aging failure prediction model.

9. An aging failure prediction device for a component, characterized in that, It includes: A collection module for collecting multi-source operation data of the component to be predicted, where the multi-source operation data includes multiple aging-related data; A processing module for inputting the multiple aging-related data into the aging failure prediction model generated according to any one of claims 1-6 to obtain an aging failure prediction result corresponding to the component to be predicted.

10. An electronic device, characterized in that, It includes: A memory for storing a computer program; A processor, configured to implement the steps of the method for generating an aging failure prediction model of components according to any one of claims 1 to 6 and the steps of the method for predicting aging failure of components according to claim 7 when executing the computer program.

Images