A fault prediction method and electronic device

Abstract

Embodiments of the present application provide a fault prediction method and an electronic device, wherein the method comprises: obtaining a target feature parameter in data of a target component in a first time period, the target feature parameter comprising at least a feature parameter representing a use condition of the target component by a user; obtaining a return parameter representing a fault return condition of the target component in a second time period after the first time period; determining N return stages based on the return parameter and a preset threshold; and training N prediction models based on the target feature parameter, the return parameter and a specific algorithm, the N prediction models corresponding to the N return stages respectively. The fault prediction method of the embodiments of the present application has better fault prediction effect and higher accuracy when predicting the fault of the target component.

Description

Technical Field

[0001] This application relates to the field of smart devices, and in particular to a fault prediction method and an electronic device. Background Technology

[0002] As people's living standards improve, many electronic and mechanical devices, such as cars, have become increasingly common in households. As a means of transportation, cars naturally experience malfunctions and require repairs. To handle customer returns, manufacturers typically need to produce a certain number of spare parts. However, producing too many spare parts wastes resources, while producing too few prevents timely response to customer needs. Currently, there are many solutions for predicting failure rates, but their effectiveness is generally poor.

[0003] Application content

[0004] This application provides a fault prediction method with better prediction effect and higher accuracy.

[0005] To address the aforementioned technical problems, embodiments of this application provide a fault prediction method, comprising:

[0006] Obtain target feature parameters of the target component in the data of the first time period, wherein the target feature parameters include at least feature parameters characterizing the user’s usage of the target component;

[0007] Obtain repair parameters characterizing the fault repair status of the target component in a second time period following the first time period;

[0008] N rework stages are determined based on the aforementioned rework parameters and preset thresholds;

[0009] Based on the target feature parameters, rework parameters, and a specific algorithm, N prediction models are trained, and each of the N prediction models corresponds to one of the N rework stages.

[0010] As a preferred option, it also includes:

[0011] Obtain the prediction data made by the N prediction models based on the target feature parameters of the input component to be predicted;

[0012] The final prediction result is obtained by processing the N predicted data.

[0013] Preferably, the feature parameters characterizing the user's use of the target component include at least one of the following parameters:

[0014] Number of uses, duration of each use, control parameters applied to the target component, number of uses of the device containing the target component, and duration of each use;

[0015] Wherein, if the target component is a component inside a vehicle, the characteristic parameters also include the vehicle's cumulative driving mileage; or

[0016] If the target component is a rechargeable battery, the characteristic parameters also include the single charging current, single charging voltage, and electronic health of the rechargeable battery.

[0017] Preferably, the target feature parameters also include one or more of the feature parameters characterizing the environment in which the target component is used and the feature parameters characterizing the fault repair status of the target component.

[0018] Preferably, determining N rework stages based on the rework parameters and preset thresholds includes:

[0019] N thresholds are preset, including the first threshold, the second threshold, ... the Nth threshold;

[0020] Define each threshold to represent the corresponding rework stage;

[0021] If the rework parameter is defined to be within the nth threshold, then it is determined to be in the nth rework stage.

[0022] As a preferred option, it also includes:

[0023] Determine the bathtub curve based on the repair parameters;

[0024] Based on three preset thresholds for the bathtub curve, including the first threshold, the second threshold, and the third threshold;

[0025] If the rework parameters are within the first threshold, it is determined to be in the first rework stage; if they are within the second threshold, it is determined to be in the second rework stage; and if they are within the third threshold, it is determined to be in the third rework stage.

[0026] Preferably, the N prediction models trained based on the target feature parameters, the rework parameters, and a specific algorithm include:

[0027] The N prediction models are obtained by training based on the target feature parameters, rework parameters, and time series or regression algorithms.

[0028] Preferably, the step of fitting and training the N prediction models based on the target feature parameters, the rework parameters, and the time series algorithm or regression algorithm includes:

[0029] The repair stage is determined based on the aforementioned repair parameters;

[0030] Based on the target feature parameters and time series or regression algorithms, a prediction model corresponding to the rework stage is trained.

[0031] This application also provides an electronic device, which includes:

[0032] The acquisition module is used to acquire target feature parameters of the target component in historical data of a first time period, and repair parameters of the target component in a second time period after the first time period, which characterize the fault repair status of the target component. The target feature parameters include at least feature parameters characterizing the user's usage of the target component.

[0033] The first processing module is used to determine N rework stages based on the rework parameters and preset thresholds, and to obtain N prediction models by fitting and training based on the target feature parameters, rework parameters and a specific algorithm, wherein the N prediction models correspond to the N rework stages respectively.

[0034] As a preferred option, it also includes:

[0035] The second processing module is used to obtain the prediction data made by the N prediction models based on the target feature parameters of the input component to be predicted, and to perform fusion processing on the N prediction data to obtain the final prediction result.

[0036] Based on the disclosure of the above embodiments, it can be understood that the beneficial effect of the embodiments of this application is that the actual usage parameters of the target component, historical repair parameters, and repair stages determined based on historical repair parameters are used as training data to train multiple prediction models corresponding to each repair stage. These multiple prediction models are used to predict the failure rate of the target component at different stages in the future, and the prediction effect is more accurate. Attached Figure Description

[0037] Figure 1 This is a flowchart of the fault prediction method in the embodiments of this application.

[0038] Figure 2 This is a flowchart of a fault prediction method in another embodiment of this application.

[0039] Figure 3 This is a flowchart of a fault prediction method in another embodiment of this application.

[0040] Figure 4 This is a flowchart of a fault prediction method in another embodiment of this application.

[0041] Figure 5 This is a structural block diagram of the electronic device in the embodiments of this application. Detailed Implementation

[0042] The specific embodiments of this application will now be described in detail with reference to the accompanying drawings, but these are not intended to limit the scope of this application.

[0043] It should be understood that various modifications can be made to the embodiments disclosed herein. Therefore, the following description should not be considered as limiting, but merely as an example of embodiments. Other modifications within the scope and spirit of this disclosure will be apparent to those skilled in the art.

[0044] The accompanying drawings, which are included in and form part of this specification, illustrate embodiments of the present disclosure and, together with the general description of the disclosure given above and the detailed description of the embodiments given below, serve to explain the principles of the disclosure.

[0045] These and other features of this application will become apparent from the following description of preferred forms of embodiments given as non-limiting examples, with reference to the accompanying drawings.

[0046] It should also be understood that although this application has been described with reference to some specific examples, those skilled in the art can certainly implement many other equivalent forms of this application, which have the features described in the claims and are therefore all within the scope of protection defined herein.

[0047] The above and other aspects, features and advantages of this disclosure will become more apparent when taken in conjunction with the accompanying drawings and in view of the following detailed description.

[0048] Specific embodiments of the present disclosure are described thereafter with reference to the accompanying drawings; however, it should be understood that the disclosed embodiments are merely examples of the present disclosure and can be implemented in various ways. Well-known and / or repeated functions and structures are not described in detail to avoid unnecessary or redundant details that could obscure the present disclosure. Therefore, the specific structural and functional details disclosed herein are not intended to be limiting, but merely to serve as the basis and representative basis for the claims to teach those skilled in the art to use the present disclosure in a variety of substantially any suitable detailed structures.

[0049] This specification may use the phrases “in one embodiment,” “in another embodiment,” “in yet another embodiment,” or “in still another embodiment,” all of which may refer to one or more of the same or different embodiments according to this disclosure.

[0050] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0051] like Figure 1 As shown in the figure, this application provides a fault prediction method, which includes:

[0052] Obtain the target feature parameters of the target component in the data of the first time period. The target feature parameters include at least the feature parameters that characterize the user’s usage of the target component.

[0053] Obtain repair parameters characterizing the fault repair status of the target component in the second time period following the first time period;

[0054] N rework stages are determined based on rework parameters and preset thresholds;

[0055] Based on the target feature parameters, rework parameters, and specific algorithms, N prediction models are trained, and each of the N prediction models corresponds to one of the N rework stages.

[0056] For example, the target component can be a machine part, such as any component in a car, electronic device, or household appliance. When the system wants to predict the failure of the target component, it can obtain data from a past first time period through historical databases, historical information records, etc., such as relevant data of the target component in the past month, such as failure data, repair data, usage data, etc. Then, the system determines the target characteristic parameters based on the data. The target characteristic parameters in this embodiment are not unique, but they must at least include characteristic parameters of the user's usage of the target component, such as usage frequency, usage duration, and usage environment, etc. After the system determines the target characteristic parameters, it will continue to obtain the repair parameters of the target component in a second time period after the first time period. For example, if the first time period is January 2019, then the second time period can be February 2019, or February and March, etc. The repair parameters are used to characterize the repair situation of the target component due to failure, such as the specific type of failure, the number of repairs, the repair time, the time interval between two adjacent repairs, whether the repaired component was changed, the cause of the failure, etc., and the specifics are not fixed. After obtaining the rework parameters, the system can determine N rework stages based on these parameters and preset thresholds. These preset thresholds can be the estimated value or range calculated by the system based on historical big data, indicating that the target component will fail and require rework within various time periods that meet the time sequence. Based on this estimated value or range and the obtained rework parameters, the system determines the N rework stages for the target component. For example, the first rework stage is within 3 months of the target component's start of use, the second rework stage is within 3-8 months, the third rework stage is within 8-15 months, and so on. After determining each rework stage, the system establishes a model architecture corresponding to each stage. Then, based on the previously obtained target feature parameters, rework parameters, and a specific algorithm, the model architecture is trained, weights are determined, and finally, N prediction models are obtained, each used to predict the failure of the target component at each rework stage.

[0057] The beneficial effect of this application's embodiments lies in using the actual usage parameters of the target component, historical repair parameters, and repair stages determined based on the historical repair parameters as training data. This allows for the training of multiple prediction models, each corresponding to a different repair stage. These multiple prediction models are then used to predict the failure rate of the target component at different future stages, resulting in more accurate predictions. For example, if a user predicts the failure rate of a target component for the next two months, the system can determine the repair stage at the time of prediction by obtaining the actual usage time and the predicted duration of the target component. Then, it can use the prediction model corresponding to that repair stage to make predictions, resulting in better prediction performance, higher accuracy, and greater reference value.

[0058] Furthermore, such as Figure 2 As shown, the method in this embodiment further includes:

[0059] Obtain prediction data from N prediction models based on the target feature parameters of the input component to be predicted;

[0060] The final prediction result is obtained by processing N prediction data.

[0061] Specifically, when a user wants to predict the overall failure rate of a target component when it is first put into use, all models will predict the target component at each rework stage, resulting in N prediction results. The system will then perform calculations based on these N prediction results, such as calculating the average of the N results. A weighted average method can be used to calculate the average, but other algorithms for calculating averages can also be used in practice. Since this embodiment has prediction models corresponding to each rework stage, and each prediction model is trained based on a large amount of historical rework data of components of the same type as the target component, as well as user usage data, the prediction model can grasp various failure parameters and parameters that induce component failure. It fully combines theoretical and practical data, thus weighing the potential failure factors of the target component at each stage, including human error, environment, its own structure, and material properties, significantly improving the prediction accuracy and reference value of the predicted values.

[0062] Furthermore, the feature parameters characterizing the user's use of the target component described in this embodiment include at least one of the following parameters:

[0063] Number of uses, duration of each use, control parameters applied to the target component, number of uses of the device containing the target component, and duration of each use;

[0064] If the target component is a part inside a car, the characteristic parameters also include the car's cumulative driving mileage; or

[0065] If the target component is a rechargeable battery, the characteristic parameters also include the battery's single-charge current, single-charge voltage, and electronic health.

[0066] For example, taking a car part as the target component, the extracted feature parameters may include:

[0067] 1. The target component is a car rechargeable battery, and its characteristic parameters are charge and discharge information:

[0068] Charge count, single charge duration, total charge duration, single charge current, average charge current, single charge voltage, average charge voltage, SOC (a parameter characterizing battery electronic health), etc.

[0069] 2. The target components are those used during vehicle operation to achieve driving objectives, such as headlights, steering wheel, engine, drive chain, gear sets in the transmission, etc., while the characteristic parameters can be parameters related to driving behavior, including:

[0070] The data includes: cumulative number of driving trips, single driving duration, cumulative driving time, single and cumulative driving mileage, single and cumulative number of times parts are used (e.g., number of times turn signals are turned on), single and cumulative usage time of parts (e.g., turn signal usage time), and controllability parameters of parts (e.g., steering wheel rotation speed, rotation angle, etc.).

[0071] Furthermore, the target feature parameters in this embodiment also include one or more of the following: feature parameters characterizing the environment in which the target component is used, and feature parameters characterizing the fault repair status of the target component, for example:

[0072] 1. The target component is any part of the automobile, and the characteristic parameters include environmental information, specifically ambient temperature, ambient humidity, actual road conditions, etc.

[0073] 2. When extracting target parameters for a target component as fault repair parameters, this includes:

[0074] The data includes the cumulative number of times a component is used, the duration of a single and cumulative use of a component, the cumulative number of times a component is repaired, the duration of a single normal operation of a component, the interval between two consecutive failures of a component, the cumulative number of times a component is maintained, and the maintenance interval of a component.

[0075] By combining various parameters of the target component during actual use, such as environmental parameters of the environment in which it is used and usage parameters involved by the user, the user's usage habits and the overall operating environment of the component can be reflected. Therefore, the training data formed covers all aspects of the target component during use, which can effectively improve the accuracy of each weight in the model and thus improve the accuracy of the prediction results.

[0076] Furthermore, such as Figure 3 As shown, in this embodiment, determining N rework stages based on rework parameters and preset thresholds includes:

[0077] N thresholds are preset, including the first threshold, the second threshold, ... the Nth threshold;

[0078] Define each threshold to represent the corresponding rework stage;

[0079] If the rework parameters are defined to be within the nth threshold, then it is determined to be in the nth rework stage.

[0080] For example, the system can determine N thresholds corresponding to the usage time of the target component based on historical usage data, failure data, repair data, and parameters such as manufacturing materials and processes, including service life and wear rate. These thresholds include a first threshold, a second threshold, and so on up to the Nth threshold. Each threshold is then defined to represent a corresponding repair stage; for example, the first threshold corresponds to the first repair stage, the second threshold to the second repair stage, and so on. Each repair stage can represent an early failure stage, a return / exchange stage, a warranty stage, an occasional failure stage, an out-of-warranty failure stage, and so on. After defining the relationship between each threshold and each repair stage, the system can define that if repair parameters, such as the usage time of the target component, fall within the nth threshold, it is determined to be in the nth repair stage; if the usage time of the target component falls within the first threshold, it is determined to be in the first repair stage. In this way, the system can quickly and efficiently determine the repair stage of the corresponding target component, and thus quickly determine a predictive model for fault prediction.

[0081] Furthermore, such as Figure 4 As shown, in this embodiment, when the system determines N rework stages based on rework parameters and preset thresholds, it also includes:

[0082] Determine the bathtub curve based on the repair parameters;

[0083] At least three thresholds should be preset based on the bathtub curve, including a first threshold, a second threshold, and a third threshold;

[0084] If the rework parameters are within the first threshold, it is determined to be in the first rework stage; if they are within the second threshold, it is determined to be in the second rework stage; and if they are within the third threshold, it is determined to be in the third rework stage.

[0085] Specifically, the bathtub curve (failure rate curve) is characterized by its high peaks and low midpoints, exhibiting distinct stages. It refers to the predictable pattern of reliability changes in a product, or target component, throughout its entire lifecycle from initial use to eventual disposal. In this embodiment, the system calculates the corresponding bathtub curve based on rework parameters. Specifically, the system performs regression calculations on rework parameters that meet time-series requirements to obtain the bathtub curve. Then, based on the curve's characteristics and the rework parameters, at least three thresholds are preset: a first threshold, a second threshold, and a third threshold. The first threshold corresponds to one end of the curve, where the curve generally declines, indicating a decrease in the rework rate. The third threshold corresponds to the other end of the curve, where the curve generally rises, indicating an increase in the rework rate. The second threshold corresponds to the middle of the curve, indicating a consistently low rework rate. Next, based on these at least three thresholds, the system can define at least three rework stages: an early failure stage or no-reason return stage corresponding to the first threshold, an occasional failure stage corresponding to the second threshold, and an out-of-warranty failure stage corresponding to the third threshold. When setting thresholds, multiple thresholds can be set. For example, the current three regions can be further divided to obtain more thresholds.

[0086] In practical applications, the system can determine the threshold and rework stage according to the above method, or it can determine them without using the bathtub curve calculation method. For example, the threshold can be determined by manually analyzing the historical data of the target component or by big data analysis algorithms.

[0087] Furthermore, in this embodiment, when training N prediction models based on target feature parameters, revision parameters, and a specific algorithm, the following steps are included:

[0088] N prediction models are obtained by training based on target feature parameters, repair parameters, time series algorithms, and regression algorithms.

[0089] Specifically, in this embodiment, N prediction models are trained based on target feature parameters, rework parameters, and time series or regression algorithms, including:

[0090] Determine the repair stage based on the repair parameters;

[0091] Based on the target feature parameters and time series and regression algorithms, a prediction model for the corresponding repair stage is obtained through training.

[0092] In this embodiment, the prediction model for each stage is trained simultaneously based on the target component's feature parameters, rework parameters, time-series algorithms, and regression algorithms to obtain the corresponding prediction model for each stage. The time-series algorithms may include multivariate time series analysis methods (ARIMAX), which can be used to assist in predicting the number of component rework cycles. The regression algorithms may include linear regression, decision trees, and multivariate regression. As mentioned earlier, each prediction model uses a bathtub curve to determine the corresponding stage and threshold. Since the feature weights corresponding to the time-series and regression algorithms used in building the model differ for different stages of the bathtub curve (e.g., the feature weights of the regression equation are different when using a regression algorithm), and these feature weights are affected by different stages of the bathtub curve, this embodiment actually uses the bathtub curve, time-series algorithms, and regression algorithms simultaneously to train the model when determining the prediction model for each stage, thus obtaining the final prediction models.

[0093] Specifically, the prediction method of this application will be described in detail below using a specific embodiment as an example. For example, taking a car's turn signal as an example, a user wants to obtain the failure rate of the turn signal in the next year. To solve this problem, the system needs to pre-determine the repair stage of the turn signal based on historical big data, such as the number of repairs during various time periods after the turn signal leaves the factory. For example, three repair stages can be determined: early failure period / no-reason return period, occasional failure period / stable use period, and failure period beyond warranty. The time points of these three stages can be determined based on historical analysis data of car turn signals. For example, the time period of each of the three stages is 2 months, and the start time of the later time period is based on the end time of the previous time period. After determining the time period representing each repair stage, a model is built for each stage. Next, the system obtains repair data for the turn signal within 6 months of its manufacture and user usage data for the turn signal within the past 6 months. This usage data includes, for example, the duration of a single use, the cumulative duration of use, the usage environment, and the number of uses. The system then divides the obtained data into three parts, using the first two months as the time node. The data within the first two months of manufacture is used as the input data for the model corresponding to the first repair stage, and the data within the next two months is used as the output data for the model. This process continues, using the data from each repair stage and the data from the next time stage as training data to train the model for each repair stage. During training, the data from the same stage can be calculated simultaneously using bathtub curves, time series algorithms, and regression algorithms to determine the weights of the model corresponding to that stage. After training the models, users can input the turn signal data to be predicted. For example, if a user has the turn signal's repair data from the fifth month and their usage data, and wants to know the failure rate of the turn signal after the sixth month, the system, upon receiving the data, can determine the corresponding repair stage as exceeding the warranty period based on the time points in the data. Therefore, the system inputs the turn signal's fifth-month repair data and user usage data into the corresponding model that is outside the warranty period, and that model predicts the failure rate of the turn signal in the sixth month. If a user has just purchased a new turn signal and has only used it for one month, and wants to know the failure rate of the turn signal in the next six months, the system simultaneously inputs the one-month repair data and user usage data into three trained models. These three models simultaneously predict the failure rate of the turn signal, and the three prediction results are then processed, for example, by weighted averaging, to obtain the final prediction result.

[0094] like Figure 5 As shown, this application also provides an electronic device, which includes:

[0095] The acquisition module is used to acquire target feature parameters of the target component in the historical data of the first time period, and repair parameters of the target component in the second time period after the first time period, which characterize the fault repair status of the target component. The target feature parameters include at least feature parameters characterizing the user's usage of the target component.

[0096] The first processing module is used to determine N rework stages based on rework parameters and preset thresholds, and to obtain N prediction models by fitting and training based on target feature parameters, rework parameters and specific algorithms. The N prediction models correspond to the N rework stages respectively.

[0097] For example, the target component can be a machine part, such as any component in a car, electronic device, or household appliance. When the electronic device wants to predict the failure of the target component, the acquisition module can obtain data from a historical database, historical information records, etc., for a past first time period, such as relevant data of the target component in the past month, such as failure data, repair data, usage data, etc. Then, the system determines the target characteristic parameters based on the data. The target characteristic parameters in this embodiment are not unique, but they must at least include characteristic parameters of the user's usage of the target component, such as usage frequency, usage duration, and usage environment, etc. After the system determines the target characteristic parameters, it will continue to obtain the repair parameters of the target component for a second time period after the first time period. For example, if the first time period is January 2019, then the second time period can be February 2019, or February and March, etc. The repair parameters are used to characterize the situation of the target component being repaired due to failure, such as the specific type of failure, the number of repairs, the repair time, the time interval between two adjacent repairs, whether the repaired component was changed, the cause of the failure, etc., and the specifics are not fixed. After obtaining the rework parameters, the first processing module can determine N rework stages based on these parameters and preset thresholds. These preset thresholds can be the estimated value or range calculated by the system based on historical big data, indicating that the target component will fail and require rework within various time periods that meet the time sequence. The system determines the N rework stages for the target component based on this estimated value or range and the obtained rework parameters. For example, the first rework stage might be within 3 months of the target component's start of use, the second rework stage within 3-8 months, the third rework stage within 8-15 months, and so on. After determining each rework stage, the first processing module establishes a model architecture corresponding to each stage. Then, based on the previously obtained target feature parameters, rework parameters, and a specific algorithm, it trains the model architecture, determines the weights, and finally obtains N prediction models, each used to predict the failure of the target component at each rework stage.

[0098] The beneficial effect of this application's embodiments lies in using the actual usage parameters of the target component, historical repair parameters, and repair stages determined based on the historical repair parameters as training data. This allows for the training of multiple prediction models, each corresponding to a different repair stage. These multiple prediction models are then used to predict the failure rate of the target component at different future stages, resulting in more accurate predictions. For example, if a user predicts the failure rate of a target component for the next two months, the system can determine the repair stage at the time of prediction by obtaining the actual usage time and the predicted duration of the target component. Then, it can use the prediction model corresponding to that repair stage to make predictions, resulting in better prediction performance, higher accuracy, and greater reference value.

[0099] Furthermore, the electronic device in this embodiment also includes:

[0100] The second processing module is used to obtain prediction data made by N prediction models based on the target feature parameters of the input component to be predicted, and to fuse the N prediction data to obtain the final prediction result.

[0101] Specifically, when a user wants to predict the overall failure rate of a target component when it is first put into use, all models will make predictions for the target component at each rework stage, resulting in N prediction results. The second processing module will then perform calculations based on these N prediction results, such as calculating the average of the N results. Specifically, a weighted average method can be used to calculate the average, but other algorithms for calculating averages can also be used in practice; there are no specific limitations. Since this embodiment has prediction models corresponding to each rework stage, and each prediction model is trained based on a large amount of historical rework data of components of the same type as the target component, as well as user usage data, the prediction model can grasp various failure parameters and parameters that induce component failure. It fully combines theoretical and practical data, thus weighing the potential failure factors of the target component at each stage, including human error, environment, its own structure, and material properties, significantly improving the prediction accuracy and reference value of the predicted values.

[0102] Furthermore, the feature parameters characterizing the user's use of the target component described in this embodiment include at least one of the following parameters:

[0103] Number of uses, duration of each use, control parameters applied to the target component, number of uses of the device containing the target component, and duration of each use;

[0104] If the target component is a part inside a car, the characteristic parameters also include the car's cumulative driving mileage; or

[0105] If the target component is a rechargeable battery, the characteristic parameters also include the battery's single-charge current, single-charge voltage, and electronic health.

[0106] For example, taking a car part as the target component, the extracted feature parameters may include:

[0107] 1. The target component is a car rechargeable battery, and its characteristic parameters are charge and discharge information:

[0108] Charge count, single charge duration, total charge duration, single charge current, average charge current, single charge voltage, average charge voltage, SOC (a parameter characterizing battery electronic health), etc.

[0109] 2. The target components are those used during vehicle operation to achieve driving objectives, such as headlights, steering wheel, engine, drive chain, gear sets in the transmission, etc., while the characteristic parameters can be parameters related to driving behavior, including:

[0110] The data includes: cumulative number of driving trips, single driving duration, cumulative driving time, single and cumulative driving mileage, single and cumulative number of times parts are used (e.g., number of times turn signals are turned on), single and cumulative usage time of parts (e.g., turn signal usage time), and controllability parameters of parts (e.g., steering wheel rotation speed, rotation angle, etc.).

[0111] Furthermore, the target feature parameters in this embodiment also include one or more of the following: feature parameters characterizing the environment in which the target component is used, and feature parameters characterizing the fault repair status of the target component, for example:

[0112] 1. The target component is any part of the automobile, and the characteristic parameters include environmental information, specifically ambient temperature, ambient humidity, actual road conditions, etc.

[0113] 2. When extracting target parameters for a target component as fault repair parameters, this includes:

[0114] The data includes the cumulative number of times a component is used, the duration of a single and cumulative use of a component, the cumulative number of times a component is repaired, the duration of a single normal operation of a component, the interval between two consecutive failures of a component, the cumulative number of times a component is maintained, and the maintenance interval of a component.

[0115] By combining various parameters of the target component during actual use, such as environmental parameters of the environment in which it is used and usage parameters involved by the user, the resulting training data can cover all aspects of the target component during use, effectively improving the accuracy of each weight in the model, thereby improving the accuracy of the prediction results.

[0116] Furthermore, in this embodiment, the first processing module, when determining N rework stages based on rework parameters and preset thresholds, includes:

[0117] N thresholds are preset, including the first threshold, the second threshold, ... the Nth threshold;

[0118] Define each threshold to represent the corresponding rework stage;

[0119] If the rework parameters are defined to be within the nth threshold, then it is determined to be in the nth rework stage.

[0120] For example, the first processing module can determine N thresholds corresponding to the usage time of the target component based on historical usage data, fault data, repair data, and parameters such as manufacturing materials and processes, including service life and wear rate. These thresholds include a first threshold, a second threshold, ..., the Nth threshold. Each threshold is then defined to represent a corresponding repair stage; for example, the first threshold corresponds to the first repair stage, the second threshold to the second repair stage, and so on. Each repair stage can represent an early failure stage, a returnable / exchangeable stage, a warranty stage, an occasional failure stage, an out-of-warranty failure stage, etc. After the first processing module defines the relationship between each threshold and each repair stage, it can define that if repair parameters, such as the usage time of the target component, are within the nth threshold, then it is determined to be in the nth repair stage; if the usage time of the target component is within the first threshold, then it is determined to be in the first repair stage. In this way, the first processing module can quickly and efficiently determine the repair stage of the corresponding target component, and thus quickly determine a predictive model for fault prediction.

[0121] Furthermore, in this embodiment, the first processing module, when determining N rework stages based on rework parameters and preset thresholds, also includes:

[0122] Determine the bathtub curve based on the repair parameters;

[0123] At least three thresholds should be preset based on the bathtub curve, including a first threshold, a second threshold, and a third threshold;

[0124] If the rework parameters are within the first threshold, it is determined to be in the first rework stage; if they are within the second threshold, it is determined to be in the second rework stage; and if they are within the third threshold, it is determined to be in the third rework stage.

[0125] Specifically, the bathtub curve (failure rate curve) is characterized by its high peaks and low midpoints, exhibiting distinct stages. It refers to the predictable pattern of reliability changes in a product, or target component, throughout its entire lifespan from initial use to eventual disposal. In this embodiment, the first processing module calculates the corresponding bathtub curve based on return parameters. Then, based on the curve's characteristics and the return parameters, it presets at least three threshold values: a first threshold, a second threshold, and a third threshold. The first threshold corresponds to one end of the curve, where the curve generally declines, indicating a lower return rate. The third threshold corresponds to the other end of the curve, where the curve generally rises, indicating an increasing return rate. The second threshold corresponds to the middle of the curve, indicating a relatively low return rate. Next, based on these three thresholds, the system defines at least three return stages: an early failure stage or no-reason return stage corresponding to the first threshold, an occasional failure stage corresponding to the second threshold, and an out-of-warranty failure stage corresponding to the third threshold.

[0126] In practical applications, the first processing module can determine the threshold and rework stage according to the above method, or it can determine them without using the bathtub curve calculation method. For example, the threshold can be determined by manually analyzing the historical data of the target component or by big data analysis algorithms.

[0127] Furthermore, in this embodiment, the first processing module, when training N prediction models based on target feature parameters, revision parameters, and a specific algorithm, includes:

[0128] N prediction models are obtained by training based on target feature parameters, repair parameters, time series algorithms, and regression algorithms.

[0129] Specifically, in this embodiment, N prediction models are trained based on target feature parameters, rework parameters, and time series or regression algorithms, including:

[0130] Determine the repair stage based on the repair parameters;

[0131] Based on the target feature parameters and time series and regression algorithms, a prediction model for the corresponding repair stage is obtained through training.

[0132] In this embodiment, the prediction model for each stage is trained simultaneously based on the target component's feature parameters, rework parameters, time-series algorithms, and regression algorithms to obtain the corresponding prediction model for each stage. The time-series algorithms may include multivariate time series analysis methods (ARIMAX), which can be used to assist in predicting the number of component rework cycles. The regression algorithms may include linear regression, decision trees, and multivariate regression. As mentioned earlier, each prediction model uses a bathtub curve to determine the corresponding stage and threshold. Since the feature weights corresponding to the time-series and regression algorithms used in building the model differ for different stages of the bathtub curve (e.g., the feature weights of the regression equation are different when using a regression algorithm), and these feature weights are affected by different stages of the bathtub curve, this embodiment actually uses the bathtub curve, time-series algorithms, and regression algorithms simultaneously to train the model when determining the prediction model for each stage, thus obtaining the final prediction models.

[0133] Specifically, the method for fault prediction in the electronic device of this application will be described in detail below using a specific embodiment as an example. For example, taking a car's turn signal as an example, the user wants to obtain the failure rate of the turn signal in the next year. To solve this problem, the acquisition module of the electronic device needs to pre-determine the repair stage of the turn signal based on historical big data, such as the number of repairs of the turn signal in various time periods after it leaves the factory. For example, three repair stages can be determined: early failure period / no-reason return period, occasional failure period / stable use period, and failure period after warranty. The time points of these three stages can be determined based on historical analysis data of car turn signals. For example, the time period of each of the three stages is 2 months, and the start time of the later time period is based on the end time of the previous time period. After determining the time period representing each repair stage, a model is built for each stage. Then, the system obtains the repair data of the turn signal in the 6 months after it leaves the factory and the user's usage data of the turn signal in the past 6 months. The usage data includes, for example, the single usage duration, cumulative usage duration, usage environment, and number of uses of the turn signal. Next, the first processing module divides the acquired data into three parts, using the first two months as the time node. Data from the first two months after manufacturing is used as input data for the model corresponding to the first repair stage, while data from the following two months is used as output data for that model. This process continues, using data from each repair stage and the next corresponding time period as training data to train the model for each repair stage. During training, data from the same stage can be simultaneously calculated using bathtub curves, time series algorithms, and regression algorithms to determine the weights of the model corresponding to that stage. After training each model, the user can input the turn signal data to be predicted. For example, if a user has the turn signal repair data from the fifth month and their usage data, and wants to know the failure rate of the turn signal after the sixth month, the first processing module, upon receiving this data, can determine that the corresponding repair stage is beyond the warranty period based on the time points in the data. Therefore, the first processing module inputs the turn signal repair data from the fifth month and the user usage data into the model corresponding to the beyond warranty period, and the model then predicts the failure rate of the turn signal in the sixth month. If a user has just purchased a new turn signal and has only used it for a month, and wants to know the failure rate of the turn signal in the next six months, the first processing module will simultaneously input the repair data of that month and the user's usage data into three trained models. The three trained models will simultaneously predict the failure rate of the turn signal. Finally, the second processing module will perform a weighted average of the three prediction results to obtain the final prediction result.

[0134] The above embodiments are merely exemplary embodiments of this application and are not intended to limit this application. The scope of protection of this application is defined by the claims. Those skilled in the art can make various modifications or equivalent substitutions to this application within its substance and scope of protection, and such modifications or equivalent substitutions should also be considered to fall within the scope of protection of this application.

Claims

1. A fault prediction method, wherein, include: Obtain target feature parameters from the data of the target component in the first time period. The target feature parameters include at least the feature parameters characterizing the user's usage of the target component. The feature parameters characterizing the user's usage of the target component include at least one of the following parameters: number of uses, duration of a single use, control parameters applied to the target component, number of uses of the device containing the target component, and duration of a single use of the device containing the target component. Obtain repair parameters characterizing the fault repair status of the target component in a second time period following the first time period; N rework stages are determined based on the aforementioned rework parameters and preset thresholds; Based on the target feature parameters, rework parameters, and a specific algorithm, N prediction models are trained, and each of the N prediction models corresponds to one of the N rework stages. Obtain the prediction data made by the N prediction models based on the target feature parameters of the input component to be predicted; The final prediction result is obtained by processing the N predicted data. The process of determining N rework stages based on the rework parameters and preset thresholds includes: N thresholds are preset, including the first threshold, the second threshold, ... the Nth threshold; Define each threshold to represent the corresponding rework stage; If the rework parameters are defined to be within the Nth threshold, then it is determined to be in the Nth rework stage; Also includes: Determine the bathtub curve based on the repair parameters; Based on the bathtub curve, at least three thresholds are preset, including a first threshold, a second threshold, and a third threshold; If the rework parameters are within the first threshold, it is determined to be in the first rework stage; if they are within the second threshold, it is determined to be in the second rework stage; and if they are within the third threshold, it is determined to be in the third rework stage.

2. The method according to claim 1, wherein, The target feature parameters also include one or more of the feature parameters characterizing the environment in which the target component is used and the feature parameters characterizing the fault repair status of the target component.

3. The method according to claim 1, wherein, The N prediction models trained based on the target feature parameters, revision parameters, and a specific algorithm include: The N prediction models are obtained by training based on the target feature parameters, repair parameters, time series algorithm, and regression algorithm.

4. The method according to claim 3, wherein, The N prediction models obtained by training based on the target feature parameters, revision parameters, and time series or regression algorithms include: The repair stage is determined based on the aforementioned repair parameters; Based on the target feature parameters and time series or regression algorithms, a prediction model corresponding to the rework stage is trained.

5. An electronic device, wherein, include: The acquisition module is used to acquire target feature parameters of the target component in the data of a first time period, and repair parameters of the target component in a second time period after the first time period, which characterize the fault repair status of the target component. The target feature parameters include at least feature parameters characterizing the user's usage of the target component. The feature parameters characterizing the user's usage of the target component include at least one of the following parameters: number of uses, duration of a single use, control parameters applied to the target component, number of uses of the device containing the target component, and duration of a single use of the device containing the target component. The first processing module is used to determine N rework stages based on the rework parameters and preset thresholds, and to train N prediction models based on the target feature parameters, rework parameters, and a specific algorithm, wherein each of the N prediction models corresponds to one of the N rework stages; wherein determining the N rework stages based on the rework parameters and preset thresholds includes: N thresholds are preset, including the first threshold, the second threshold, ... the Nth threshold; Define each threshold to represent the corresponding rework stage; If the rework parameters are defined to be within the Nth threshold, then it is determined to be in the Nth rework stage; Also includes: Determine the bathtub curve based on the repair parameters; Based on the bathtub curve, at least three thresholds are preset, including a first threshold, a second threshold, and a third threshold; If the rework parameter is within the first threshold, it is determined to be in the first rework stage; if it is within the second threshold, it is determined to be in the second rework stage; if it is within the third threshold, it is determined to be in the third rework stage. The second processing module is used to obtain the prediction data made by the N prediction models based on the target feature parameters of the input component to be predicted, and to process the N prediction data to obtain the final prediction result.

Images