Method and device for predicting residual service life of equipment

The equipment residual service life prediction is achieved through the support vector regression model based on restrictive quantile regression and kernel methods, and the problem of low prediction uncertainty and low accuracy in the prior art is solved, and more accurate and flexible equipment life prediction is achieved. , suitable for a wider range of application scenarios.

CN118940449APending Publication Date: 2024-11-12TSINGHUA UNIVERSITY +1
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202310518141.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-05-09
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The prior art has problems of uncertainty and low accuracy in the prediction of the remaining service life of the equipment, especially because the assumptions and prior knowledge of the degradation process are strong, resulting in a greater deviation between the predicted results and the actual results. Large, the model has low scope of application and flexibility.

Method used

Using the residual service life prediction model based on restrictive quantile regression and kernel methods, the cumulative distribution function is constructed by supporting vector regression model and random variables, and the model and the model are avoided. Assumptions and limitations of distribution forms.

Benefits of technology

It improves the accuracy and flexibility of the equipment's remaining service life prediction, can more comprehensively describe the uncertainty of prediction, expands the scope of application of the model, and provides more reliable maintenance planning.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118940449A_ABST
    Figure CN118940449A_ABST
Patent Text Reader

Abstract

The invention discloses a method and a device for predicting the residual service life of equipment. The method comprises the following steps: acquiring state data of equipment; performing quantile point estimation on the residual service life of the equipment according to the state data by using a residual service life prediction model based on restrictive quantile regression and a kernel method to obtain quantile point prediction values of the residual service life of the equipment corresponding to a plurality of quantile points respectively, the residual service life prediction model comprises a first support vector regression model, a second support vector regression model and a random variable, and each of the first support vector regression model and the second support vector regression model comprises a kernel function; and performing robust estimation of probability distribution based on the plurality of quantile point predicted values to obtain a cumulative distribution function of the residual service life of the equipment. According to the method for predicting the remaining service life of the equipment, the application range of the model can be widened, and the accuracy and flexibility of predicting the remaining service life of the equipment are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application belongs to the technical field of equipment health management, and in particular, relates to a method and device for predicting the remaining useful life of equipment. Background Art

[0002] Generally, the aging of equipment increases the risk of equipment failure and brings inevitable losses. These risks can be reduced and the failure costs can be reduced by repairing the equipment. However, frequent repairs will increase the corresponding maintenance costs. Therefore, reasonable and effective maintenance planning for equipment can minimize the total cost of failures and repairs. With the advent of the big data era and the development of sensor technology, full analysis and effective use of sensor data obtained through real-time monitoring of equipment can make a more accurate assessment of the current status of the equipment and predict its remaining service life, so that a more reliable maintenance plan can be formulated based on the remaining service life. This predictive maintenance plan can greatly improve maintenance efficiency and reduce total costs. In the process of formulating maintenance plans, the most critical step is to predict the remaining service life of the equipment.

[0003] In the prior art, a point estimate of the remaining useful life of a device is usually made based on a machine learning model, or a probability distribution estimate of the remaining useful life of a device is made based on a random process model or a Bayesian framework model.

[0004] However, in actual situations, due to the constraints of various factors, there is always uncertainty in the remaining service life prediction results, and point estimation cannot satisfy this uncertainty, resulting in low accuracy in the prediction of the remaining service life of the equipment. In addition, in order to estimate the probability distribution of the remaining service life of the equipment, it is usually necessary to make assumptions about the degradation model of the equipment or the form of the probability distribution of the remaining service life based on prior knowledge. Since the assumptions or prior knowledge about the degradation process are usually unknown, that is, the assumptions and prior knowledge are highly subjective, it may lead to a large deviation between the prediction results of the remaining service life and the actual results, resulting in a narrow scope of application of the model and low accuracy and flexibility in the prediction of the remaining service life of the equipment. Summary of the invention

[0005] The embodiments of the present application provide a method, apparatus, device, computer-readable storage medium and computer program product for predicting the remaining useful life of equipment, which can improve the scope of application of the model and improve the accuracy and flexibility of predicting the remaining useful life of equipment.

[0006] In a first aspect, an embodiment of the present application provides a method for predicting the remaining useful life of a device, the method comprising:

[0007] Get device status data;

[0008] Using a remaining useful life prediction model based on restricted quantile regression and kernel method, quantile point estimation is performed on the remaining useful life of the equipment according to the state data to obtain quantile point prediction values ​​of the remaining useful life of the equipment corresponding to multiple quantile points, respectively, wherein the remaining useful life prediction model includes a first support vector regression model, a second support vector regression model and a random variable, and both the first support vector regression model and the second support vector regression model include a kernel function;

[0009] A robust estimation of the probability distribution is performed based on the plurality of quantile point prediction values ​​to obtain a cumulative distribution function of the remaining useful life of the equipment.

[0010] In a possible implementation, before using the remaining useful life prediction model based on restricted quantile regression and kernel method to perform quantile point estimation on the remaining useful life of the equipment according to the status data, the method further includes:

[0011] An initial remaining useful life prediction model is constructed based on restricted quantile regression and kernel method, wherein the initial remaining useful life prediction model includes a first initial support vector regression model, a second initial support vector regression model and a random variable, wherein the first initial support vector regression model and the second initial support vector regression model both include a kernel function;

[0012] Obtain N historical remaining useful life data and their corresponding historical status data, where N is a positive integer;

[0013] The initial remaining useful life prediction model is trained using the N historical remaining useful life data and their respectively corresponding historical status data to obtain the remaining useful life prediction model.

[0014] In a possible implementation, the training of the initial remaining useful life prediction model using the N historical remaining useful life data and their corresponding historical status data to obtain the remaining useful life prediction model includes:

[0015] Using the N historical remaining useful life data and their corresponding historical status data, the first initial support vector regression model is estimated to obtain a first support vector regression model;

[0016] Obtaining a residual value of the first support vector regression model for each of the historical remaining useful life data and its corresponding historical status data to obtain N residual values;

[0017] Performing non-negative function fitting on the N residual values ​​and their corresponding historical remaining useful life data, determining model parameters of the second initial support vector regression model, and obtaining a second support vector regression model;

[0018] The random variable is estimated using a quantile regression model to obtain estimated values ​​of the random variable corresponding to the multiple quantile points respectively;

[0019] The remaining useful life prediction model is constructed according to the first support vector regression model, the second support vector regression model and the estimated values ​​of the random variables corresponding to the multiple quantile points.

[0020] In a possible implementation, obtaining N pieces of historical status data includes:

[0021] Acquire N historical state data vectors, each of which includes first historical state data of m dimensions, where m is a positive integer;

[0022] Normalizing the first historical state data in the N historical state data vectors to obtain N target historical state data vectors;

[0023] The N target historical state data vectors are determined as the N historical state data.

[0024] In a possible implementation, obtaining the N historical status data further includes:

[0025] For each of the target historical state data vectors, linear regression is performed on the first historical state data of the m dimensions to obtain N historical state data.

[0026] In a possible implementation, before performing linear regression on the first historical state data of the m dimensions, the method further includes:

[0027] Performing a correlation analysis on the historical remaining useful life data and the first historical status data of the m dimensions to obtain a correlation analysis result, wherein the correlation analysis results are sorted in descending order of correlation;

[0028] According to the correlation analysis result, selecting the first historical state data of the first n dimensions, where n is a positive integer;

[0029] The performing linear regression on the first historical state data of the m dimensions includes:

[0030] Perform linear regression on the first historical state data of the n dimensions.

[0031] In a possible implementation, the normalizing the first historical state data in the N historical state data vectors to obtain N target historical state data vectors includes:

[0032] Clustering the first historical state data in the N historical state data vectors according to environmental characteristics to obtain multiple categories;

[0033] The first historical state data belonging to the same category are normalized within the category to obtain N target historical state data vectors.

[0034] In a possible implementation manner, after obtaining the cumulative distribution function of the remaining useful life of the device, the method further includes:

[0035] Obtain the service life of the equipment, replacement cost after a single failure, single preventive replacement cost, and downtime cost per unit time;

[0036] The total cost per unit time within the replacement cycle of the equipment is optimized based on the service life of the equipment, the replacement cost after a single failure, the single preventive replacement cost, the downtime cost per unit time and the cumulative distribution function to determine the replacement time of the equipment. The replacement cycle is the time period between two consecutive replacement operations.

[0037] In a second aspect, an embodiment of the present application provides a device for predicting the remaining useful life of a device, the device comprising:

[0038] A first acquisition module, used to acquire status data of a device;

[0039] a first estimation module, configured to use a remaining useful life prediction model based on restricted quantile regression and a kernel method to perform quantile point estimation on the remaining useful life of the equipment according to the state data, and obtain quantile point prediction values ​​of the remaining useful life of the equipment corresponding to a plurality of quantile points, wherein the remaining useful life prediction model comprises a first support vector regression model, a second support vector regression model and a random variable, and both the first support vector regression model and the second support vector regression model comprise a kernel function;

[0040] The second estimation module is used to perform a robust estimation of the probability distribution based on the plurality of quantile point prediction values ​​to obtain a cumulative distribution function of the remaining useful life of the equipment.

[0041] In a third aspect, an embodiment of the present application provides an electronic device, the device comprising: a processor and a memory storing computer program instructions;

[0042] When the processor executes the computer program instructions, the processor implements any possible implementation method of the first aspect described above.

[0043] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, implements a method in any possible implementation method of the first aspect described above.

[0044] In a fifth aspect, an embodiment of the present application provides a computer program product. When instructions in the computer program product are executed by a processor of an electronic device, the electronic device executes a method as any possible implementation method in the first aspect above.

[0045] In the embodiment of the present application, by first using the remaining useful life prediction model based on the restricted quantile regression and kernel method, the remaining useful life of the equipment is estimated by quantile points according to the state data, and multiple quantile point prediction values ​​of the remaining useful life of the equipment are obtained, and then the probability distribution is robustly estimated based on the multiple quantile point prediction values ​​to obtain the cumulative distribution function of the remaining useful life of the equipment. The remaining useful life of the equipment can be predicted by the cumulative distribution function. Compared with predicting the remaining useful life of the equipment using point estimates, the comprehensiveness of the prediction of the remaining useful life of the equipment can be improved, that is, the uncertainty of the prediction can be characterized. In addition, since the remaining useful life prediction model is constructed based on the restricted quantile regression and kernel method, that is, no assumptions are made on the model, and no restrictions are imposed on the distribution form, therefore, by predicting the remaining useful life distribution of the equipment through the remaining useful life prediction model based on the restricted quantile regression and kernel method, the scope of application of the model can be improved, and the accuracy and flexibility of the prediction of the remaining useful life of the equipment can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solution of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0047] Figure 1 It is a flow chart of a method for predicting the remaining useful life of equipment provided in an embodiment of the present application;

[0048] Figure 2 It is a flow chart of another method for predicting the remaining useful life of equipment provided in an embodiment of the present application;

[0049] Figure 3 It is a schematic diagram of a comparison between a RUL prediction result and a true value provided in an embodiment of the present application;

[0050] Figure 4 It is a schematic diagram of a comparison between a RUL distribution estimation result and a true frequency distribution diagram provided in an embodiment of the present application;

[0051] Figure 5 It is a schematic diagram of a RUL prediction result of a single electric energy metering device throughout its life cycle provided by an embodiment of the present application;

[0052] Figure 6 It is a flow chart of another method for predicting the remaining useful life of equipment provided in an embodiment of the present application;

[0053] Figure 7 It is a structural schematic diagram of a device for predicting the remaining useful life of equipment provided in an embodiment of the present application;

[0054] Figure 8 It is a structural schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0055] The features and exemplary embodiments of various aspects of the present application will be described in detail below. In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present application, rather than to limit the present application. For those skilled in the art, the present application can be implemented without the need for some of these specific details. The following description of the embodiments is only to provide a better understanding of the present application by illustrating the examples of the present application.

[0056] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "include..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0057] As mentioned in the background technology section, the most critical step in the maintenance planning process is to predict the remaining service life of the equipment.

[0058] The data-driven method is one of the commonly used methods in the prediction of the remaining useful life of equipment. Compared with the method based on physical models, the data-driven method does not require a comprehensive understanding of the physical characteristics and failure modes of the equipment, and does not involve the establishment of a strict mathematical model in technology, thus avoiding the problem of a decrease in the accuracy of the remaining useful life prediction due to improper model selection. The models for the prediction of the remaining useful life based on data-driven mainly include statistical data-driven models and machine learning models. Among them, the output results of most machine learning algorithms are point estimates of the remaining useful life. However, in actual situations, due to the constraints of various factors, there is always uncertainty in the prediction results of the remaining useful life, and the accuracy of the prediction greatly affects the reliability of maintenance planning / maintenance decisions. Therefore, it is very necessary to characterize the uncertainty of the prediction results, give the confidence interval of the prediction, or describe the probability distribution of the predicted value in reliability research. Since the point estimate cannot describe the probability distribution, the reliability of the maintenance decision is low. In addition, models that can give a prediction of the probability distribution of the remaining useful life, such as models based on random processes and models based on the Bayesian framework, usually need to make assumptions about the degradation model of the equipment or the form of the probability distribution of the remaining useful life based on prior knowledge. Since the assumptions or prior knowledge about the degradation process are usually unknown, that is, the assumptions and prior knowledge are highly subjective, this may lead to a large deviation between the predicted results of the remaining useful life and the actual results, resulting in a narrow scope of application of the model and low accuracy and flexibility in the prediction of the remaining useful life of the equipment.

[0059] In response to the above problems, the embodiment of the present application proposes a method for predicting the probability of the remaining useful life of an equipment based on quantile regression. By processing the multidimensional sensor data of the equipment, health indicators are extracted from it, and then a quantile regression framework that avoids the intersection of quantile curves is adopted. Based on the prediction of different quantile points of the probability distribution of the health indicators, an estimate of the probability distribution of the remaining useful life can be obtained. In addition, by integrating the idea of ​​asymmetric support vector regression into the framework of quantile regression, the prediction of the probability distribution of the remaining useful life of any form can be achieved without prior knowledge. The kernel method of asymmetric support vector regression is used to achieve nonlinear prediction, and its theoretical completeness and dual method are used to efficiently solve the prediction optimization model. In existing studies, asymmetric support vector regression algorithms are often used to deal with prediction problems in the case of asymmetric loss functions. The embodiment of the present application combines it with the quantile regression framework for the estimation of probability distribution. At the same time, the method of restrictive quantile regression is adopted to avoid the problem of the intersection of the predicted quantile curves.

[0060] Based on this, in order to solve the problems of the prior art, the embodiments of the present application provide a method, apparatus, device, computer-readable storage medium and computer program product for predicting the remaining useful life of a device.

[0061] The following first introduces the method for predicting the remaining useful life of equipment provided in the embodiment of the present application.

[0062] Figure 1 The following is a flow chart showing a method for predicting the remaining useful life of a device provided in an embodiment of the present application. Figure 1 As shown, the method for predicting the remaining useful life of equipment provided in the embodiment of the present application includes the following steps:

[0063] S110, obtaining status data of the device;

[0064] S120, using a remaining useful life prediction model based on restricted quantile regression and kernel method, performing quantile point estimation on the remaining useful life of the equipment according to the state data, and obtaining quantile point prediction values ​​of the remaining useful life of the equipment corresponding to multiple quantile points, respectively, wherein the remaining useful life prediction model includes a first support vector regression model, a second support vector regression model and a random variable, and both the first support vector regression model and the second support vector regression model include a kernel function;

[0065] S130. Perform a robust estimation of the probability distribution based on the multiple quantile point prediction values ​​to obtain a cumulative distribution function of the remaining useful life of the equipment.

[0066] The method for predicting the remaining useful life of the equipment in the embodiment of the present application first uses the remaining useful life prediction model based on the restricted quantile regression and the kernel method, performs quantile point estimation on the remaining useful life of the equipment according to the state data, obtains multiple quantile point prediction values ​​of the remaining useful life of the equipment, and then performs robust estimation of the probability distribution based on the multiple quantile point prediction values ​​to obtain the cumulative distribution function of the remaining useful life of the equipment. The remaining useful life of the equipment can be predicted by the cumulative distribution function, which can improve the comprehensiveness of the prediction of the remaining useful life of the equipment compared to the prediction of the remaining useful life of the equipment using the point estimation, that is, it can characterize the uncertainty of the prediction. In addition, since the remaining useful life prediction model is constructed based on the restricted quantile regression and the kernel method, that is, no assumptions are made on the model, and no restrictions are imposed on the distribution form, therefore, by predicting the remaining useful life distribution of the equipment through the remaining useful life prediction model based on the restricted quantile regression and the kernel method, it is possible to improve the scope of application of the model and improve the accuracy and flexibility of the prediction of the remaining useful life of the equipment.

[0067] The specific implementation methods of the above steps are introduced below.

[0068] In some embodiments, in S110, the device may be, for example, an electric energy metering device. In the case where the device is an electric energy metering device, the state data may include, for example, voltage, current, power consumption, etc. Specifically, the state data may be a vector including data such as voltage, current, power consumption, etc., or may be a specific value obtained by performing linear regression on data such as voltage, current, power consumption, etc., which is not limited here.

[0069] As an example, by deploying multiple sensors on the electric energy metering device, the sensor data of the electric energy metering device in multiple dimensions can be obtained to obtain a sensor data vector. Each sensor can measure sensor data of one dimension. After normalizing the sensor data of multiple dimensions, the sensor data vector after normalization can be used as state data. In addition, linear regression can be performed on the sensor data of multiple dimensions after normalization, and the specific value obtained by the linear regression can be used as state data.

[0070] In some embodiments, in S120, the remaining useful life prediction model It can be as follows:

[0071]

[0072] in, It can be the first support vector regression model, It can be a second support vector regression model. Specifically, Among them, φ(·) is the kernel function, ω1 and b1 are the model parameters in the first support vector regression model, and ω2 and b2 are the model parameters in the second support vector regression model. In addition, It can be the random variable of the error term corresponding to the quantile point. The random variables corresponding to different quantile points can be different.

[0073] As an example, if the state data is recorded as h, then for the preset quantile points that need to be estimated By inputting h into the remaining useful life prediction model, each quantile point τ can be obtained j The corresponding quantile point prediction values ​​of the remaining useful life of the equipment Where J is the number of quantile points.

[0074] In some embodiments, in S130, for a plurality of quantile point prediction values, namely, the predicted remaining useful life (RUL) quantile points, a robust estimate of their probability distribution is obtained: for a series of estimated quantile points The cumulative distribution function (CDF) of the predicted output is:

[0075]

[0076] When J is large enough, When the selection of is dense enough, this method will obtain more and more accurate distribution estimates of RUL.

[0077] Based on this, in order to improve the reliability of the electric energy metering device, in some embodiments, after the above S130, the following steps may also be included:

[0078] Obtain the equipment usage time, replacement cost after a single failure, single preventive replacement cost, and downtime cost per unit time;

[0079] According to the equipment's service life, replacement cost after a single failure, single preventive replacement cost, unit time downtime cost and cumulative distribution function, the total unit time cost within the equipment's replacement cycle is optimized to determine the equipment's replacement time. The replacement cycle is the time period between two consecutive replacement operations.

[0080] Here, based on the prediction results of the RUL probability distribution, a recommendation on the replacement time of the energy metering device can be given. For an energy metering device with an age (duration) of t0, the replacement cost of the energy metering device after a single failure is c c , the cost of a single preventive replacement is c p , the downtime cost per unit time is c d For a given (t0, c c , c p , c d ), the time period between two consecutive replacement operations is called a replacement cycle. The following optimization problem can obtain the optimal value of the replacement time t by minimizing the total cost per unit time CR(t) in the current replacement cycle:

[0081]

[0082] Among them, t is the decision variable, that is, how long it takes to replace the electric energy metering device, F(·) is the CDF of the predicted RUL, is the complement function of CDF.

[0083] In this way, by predicting the CDF of RUL based on the above remaining service life prediction model, and then determining how long it will take to replace the electric energy metering device based on the CDF, the accuracy of the decision-making recommendations for maintenance and replacement of the electric energy metering device can be improved, and the reliability of the electric energy metering device can be improved.

[0084] In order to achieve any form of probability distribution prediction of the remaining useful life of equipment without prior knowledge, model assumptions and distribution form restrictions, as another implementation method of the present application, the present application also provides another implementation method of the method for predicting the remaining useful life of equipment, please refer to the following embodiments for details.

[0085] See also Figure 2 The method for predicting the remaining useful life of equipment provided in the embodiment of the present application may further include the following steps before S120 shown in the above embodiment:

[0086] S210, constructing an initial remaining useful life prediction model based on restricted quantile regression and kernel method, the initial remaining useful life prediction model comprising a first initial support vector regression model, a second initial support vector regression model and a random variable, the first initial support vector regression model and the second initial support vector regression model both comprising a kernel function;

[0087] S220, obtaining N historical remaining useful life data and their corresponding historical status data, where N is a positive integer;

[0088] S230: train an initial remaining useful life prediction model using N historical remaining useful life data and their corresponding historical status data to obtain a remaining useful life prediction model.

[0089] The embodiment of the present application constructs an initial remaining useful life prediction model based on restrictive quantile regression and kernel method, and trains the initial remaining useful life prediction model to obtain a remaining useful life prediction model. It can realize any form of probability distribution prediction of the remaining useful life of the equipment without prior knowledge, model assumptions and distribution form restrictions.

[0090] The specific implementation methods of the above steps are introduced below.

[0091] In some embodiments, in S210, the initial remaining useful life prediction model y(h) may be as follows:

[0092] y(h)=f(h)+s(h)e

[0093] in, is the independent variable, is the dependent variable, e is the random variable representing the error term, f(h) can be the first initial support vector regression model, and s(h) can be the second initial support vector regression model. f(·) and s(·) can be two unknown equations, representing the median and variance of the prediction result (i.e., the distribution corresponding to y(h)). Specifically, f(h) = ω1 T φ(h)+b1,s(h)=ω2 Tφ(h)+b2. Where φ(h) is the kernel function, ω1 and b1 are the model parameters to be estimated in the first initial support vector regression model, and ω2 and b2 are the model parameters to be estimated in the second initial support vector regression model. Without loss of generality, it can be assumed that the median of the e distribution is 0 and the median of |e| is 1. And for all h, s(h)>0.

[0094] In some embodiments, in S220, the historical remaining service life data may be the real remaining service life data of the device whose life cycle has ended at each time during its life cycle. The device may be, for example, an electric energy metering device. In the case where the device is an electric energy metering device, the historical status data may include, for example, voltage, current, power consumption, etc. Specifically, the historical status data may be a vector including data such as voltage, current, power consumption, etc., or may be a specific value obtained by linear regression of data such as voltage, current, power consumption, etc., which is not limited here.

[0095] Based on this, in some embodiments, the above-mentioned acquisition of N historical status data may specifically include:

[0096] Obtain N historical state data vectors, each of which includes first historical state data of m dimensions, where m is a positive integer;

[0097] Normalizing the first historical state data among the N historical state data vectors to obtain N target historical state data vectors;

[0098] The N target historical state data vectors are determined as N historical state data.

[0099] Here, by deploying multiple sensors on the electric energy metering device, the sensor data vector x of m dimensions in the entire life cycle of the electric energy metering device can be obtained. Among them, x can be a historical state data vector, and the sensor data can be the first historical state data. If the historical remaining service life data is recorded as y, the training sample data can be recorded as {(x1, y1),..., (x N ,y N )}.in, is the input vector (the eigenvector associated with the energy metering device RUL), is the true RUL of the electric energy metering device, m represents the dimension of the input feature, and N represents the total number of samples.

[0100] By normalizing the first historical state data in the N historical state data vectors, the first historical state data can be unified to the same magnitude, and N target historical state data vectors after normalization are obtained, thereby obtaining N historical state data.

[0101] Since the first historical state data of the same dimension may fluctuate greatly under different environments, in order to eliminate the impact of the environment on the data and ensure the accuracy of data normalization, in some embodiments, the above-mentioned normalization processing of the first historical state data in the N historical state data vectors to obtain N target historical state data vectors may specifically include:

[0102] Clustering the first historical state data in the N historical state data vectors according to the environmental characteristics to obtain multiple categories;

[0103] The first historical state data belonging to the same category are normalized within the category to obtain N target historical state data vectors.

[0104] Here, environmental features can also be called working condition features. The samples may be clustered according to the working condition characteristics, and based on the clustering results, the state sensor data, that is, the first historical state data belonging to the same category, may be normalized within the category.

[0105] In this way, by performing intra-class normalization processing on the first historical state data belonging to the same category, the influence of the environment on the data can be eliminated, the data can be unified to the same level, and the accuracy of data normalization can be guaranteed.

[0106] Based on this, in order to simplify the subsequent process of visualizing the prediction results, in some embodiments, the above-mentioned acquisition of N historical state data may further include:

[0107] For each target historical state data vector, linear regression is performed on the first historical state data of m dimensions to obtain N historical state data.

[0108] Here, based on the first historical status data of m dimensions, a health indicator h can be established through a linear regression model, and h is determined as the historical status data.

[0109] In this way, by performing linear regression on the first historical state data of m dimensions, specific values ​​can be obtained, which can simplify the subsequent process of visualizing the prediction results.

[0110] Based on this, in order to effectively extract key information from multi-dimensional real-time data, eliminate interference from irrelevant factors and noise, and improve the training efficiency of the model, in some embodiments, before performing linear regression on the first historical state data of m dimensions, the following may also be included:

[0111] Performing correlation analysis on the historical remaining useful life data and the first historical state data of the m dimensions to obtain correlation analysis results, and the correlation analysis results are sorted in order of correlation from strong to weak;

[0112] According to the correlation analysis results, the first historical state data of the first n dimensions are selected, where n is a positive integer;

[0113] Based on this, the linear regression of the first historical state data of m dimensions may specifically include:

[0114] Perform linear regression on the first historical state data of n dimensions.

[0115] Here, through correlation analysis, the n-dimensional first historical state data (n≤m) with the most obvious trend can be selected from the m-dimensional normalized first historical state data. Based on the extracted n-dimensional sensor data, a health indicator h can be established through a linear regression model.

[0116] In this way, by performing feature selection based on correlation analysis, it is possible to effectively extract key information from multi-dimensional real-time data, eliminate interference from irrelevant factors and noise, and improve the training efficiency of the model.

[0117] In some embodiments, in S230, the process of training the initial remaining useful life prediction model using N historical remaining useful life data and their corresponding historical status data may be to train f(·), s(·) and e τ The process of estimating separately. Among them, τ can be a set of quantile points that need to be estimated Any one of .

[0118] Based on this, in order to improve the flexibility of the model and improve the accuracy and solution efficiency of the method, in some embodiments, the above S230 may specifically include:

[0119] The first initial support vector regression model is estimated using N historical remaining useful life data and their corresponding historical status data to obtain a first support vector regression model;

[0120] Obtain the residual value of the first support vector regression model for each historical remaining useful life data and its corresponding historical status data to obtain N residual values;

[0121] Performing non-negative function fitting on the N residual values ​​and their corresponding historical remaining useful life data, determining the model parameters of the second initial support vector regression model, and obtaining the second support vector regression model;

[0122] The random variable is estimated using a quantile regression model to obtain estimated values ​​of the random variable corresponding to a plurality of quantile points;

[0123] A remaining useful life prediction model is constructed according to the first support vector regression model, the second support vector regression model and the estimated values ​​of random variables corresponding to multiple quantile points.

[0124] Here, the specific process of estimating f(·) can be as follows.

[0125] Set the regularization parameter C, the insensitive parameter ∈ and the kernel function φ(·), and use the following support vector regression model to predict f(·):

[0126]

[0127]

[0128] y i -ω1 T φ(h i )-b1≤∈+ξ i ,

[0129]

[0130] Among them, ω1 and b1 are the support vector regression model (SVR) model f(h) = ω1 T The parameters to be estimated in φ(h)+b1; represents the health index and RUL of the obtained training set samples; ξ i and is the slack variable introduced. Write down the dual problem of the above equation and solve it using the quadratic programming algorithm to get the estimate of f(·)

[0131] The specific process of estimating s(·) can be as follows.

[0132] For each training set sample i, let Represents the residual value of the training set sample obtained by the previous model. N ,|r N |)} Use a non-negative function to fit, and use the same kernel function as in the previous step:

[0133]

[0134]

[0135] |r i |-ω2 T φ(h i )-b2≤ξ i ,

[0136] ω2 T φ(h i )+b2≥0,

[0137]

[0138] Where ω2 and b2 are the SVR model s(h) = ω2 T φ(h)+b2 is the parameter to be estimated; ξ i and is the slack variable introduced. Write down the dual problem of the above equation and solve it using the quadratic programming algorithm to get an estimate of s(·)

[0139] For τ The specific process of estimating ∈(-∞, +∞) can be shown as follows.

[0140] make Select an appropriate ∈ parameter value, let The quantile regression model was used to τ Make an estimate:

[0141]

[0142] Among them, the loss function l k As shown below:

[0143]

[0144] Solving the above optimization problem, we can get an estimate of eτ

[0145] Finally, for a given input h, the τ quantile of its output value can be estimated as:

[0146]

[0147] In this way, a trained RUL prediction model can be obtained

[0148] In this way, by integrating the kernel method and dual characteristics of support vector regression into the framework of quantile regression, nonlinear prediction of remaining useful life can be achieved, which can enhance the flexibility of the model and improve the accuracy and solution efficiency of the method.

[0149] In addition, the other steps of the method in the embodiment of the present application can be seen above Figure 1 The relevant description of the illustrated embodiment will not be repeated here.

[0150] Based on this, in order to better describe the entire solution, some specific examples are given based on the above embodiments.

[0151] For example, in an embodiment of the present application, the specific process of predicting the remaining useful life of equipment using the remaining useful life prediction model based on restricted quantile regression and kernel method may include the following steps:

[0152] 1) Multi-dimensional sensor deployment to monitor the status of the electric energy metering device and obtain multi-dimensional sensor data at different times during the life cycle of the electric energy metering device;

[0153] 2) Feature selection of multi-dimensional sensor data to enable subsequent data analysis;

[0154] 3) Construct health indicators of electric energy metering devices based on multi-dimensional sensor data characteristics;

[0155] 4) Based on the relationship between the health index and the remaining service life of the electric energy metering device, a remaining service life prediction model is constructed;

[0156] 5) Using the restricted regression quantiles (RRQ) framework (restricted support vector regression framework) to predict the probability distribution of remaining useful life;

[0157] 6) Using the kernel method of ∈-support vector regression (SVR) to achieve nonlinear prediction of remaining useful life;

[0158] 7) Based on the estimation results of different quantile points of the remaining useful life distribution, a robust estimate of the probability distribution is obtained;

[0159] 8) Using training set data to train the prediction model;

[0160] 9) Evaluate the trained probability distribution prediction model through the test set;

[0161] 10) Based on the trained model, the real-time estimation of the probability distribution of the remaining service life of the electric energy metering device is realized;

[0162] 11) Based on the probability distribution prediction results of the remaining service life of the electric energy metering device, provide decision recommendations for the maintenance and replacement of the electric energy metering device.

[0163] Furthermore, the specific process of predicting the remaining useful life of equipment using the remaining useful life prediction model based on restricted quantile regression and kernel method may include steps 1 to 12.

[0164] Step 1: Before the end of the life cycle of the electric energy metering device (electric meter), multiple sensors are deployed on the electric energy metering device to obtain the sensor data vector x of m dimensions and the actual remaining service life y data of the electric energy metering device at each time point in the entire life cycle:

[0165] {(x1, y1), …, (x M ,y M )}

[0166] in is the input vector (the feature vector associated with the meter RUL), is the real RUL of the meter, m represents the dimension of the input feature, and M represents the total number of samples.

[0167] Step 2: Divide the data set into a training set and a test set. The training set is used to train the constructed RUL model, and the test set is used to verify the prediction effect. The total number of training set samples is N.

[0168] Step 3: For the training set data The samples are clustered according to the working condition characteristics, and the state sensor data are normalized within the class based on the classification results to unify the data to the same level.

[0169] Step 4: Select n types of sensor data (n≤m) with the most obvious trends from the m types of normalized sensor data through correlation analysis.

[0170] Step 5: Based on the extracted n-dimensional sensor data, a health indicator h is established through a linear regression model.

[0171] Step 6: Establish a remaining useful life prediction model based on restricted quantile regression and kernel method:

[0172] y(h)=f(h)+s(h)e.

[0173] Step 7: Train the remaining useful life prediction model to obtain a trained RUL prediction model. The specific process of model training can be found in the above description and will not be described in detail here.

[0174] In step 8, the same preliminary processing is performed on the test set data, including data normalization, feature selection, and extraction of health indicators.

[0175] In step 9, the health index extracted from the test set data is used as input and substituted into the above trained model. Then, for a given health index h, the τ quantile of its RUL can be estimated as:

[0176]

[0177] Step 10: Based on the predicted RUL quantile points, a robust estimate of its probability distribution is obtained: for a series of estimated quantile points The cumulative distribution function (CDF) of the predicted output is

[0178]

[0179] Step 11, evaluate the prediction results. Taking the degradation simulation data of the electric energy metering device as an example, the relationship between the RUL distribution prediction results, predicted values, 90% confidence interval and the actual RUL of a single electric energy metering device at a single moment can be as follows: Figure 3 As shown; a schematic diagram of the comparison between the RUL distribution prediction result of a single electric energy metering device at a single moment and the RUL frequency distribution diagram of similar electric energy metering devices in the historical data set can be shown as Figure 4 As shown; a schematic diagram of the change of the RUL predicted value, 90% confidence interval and true RUL of a single electric energy metering device over its entire life cycle with its age can be shown as Figure 5 The RUL prediction results vary with the meter life stage. The bucket evaluation results of different prediction indicators on all test sets can be shown in Table 1 (MSE: mean square error, RMSE: root mean square error, R 2 : coefficient of determination, MAE: mean absolute error, MAPE: mean absolute percentage error, 90% coverage: the proportion of predicted 90% confidence intervals covering the true value, 95% coverage: the proportion of predicted 95% confidence intervals covering the true value).

[0180] Table 1

[0181]

[0182] Step 12: Based on the RUL probability distribution prediction result, a suggestion on the time to replace the electric meter is given.

[0183] In addition, based on the training set sample data and the test machine sample data, the specific process of establishing the RUL prediction model and analyzing the prediction effect of the RUL prediction model can be as follows: Figure 6 shown.

[0184] In this way, the embodiment of the present application realizes for the first time the probability distribution prediction of the remaining service life of the electric energy metering device based on multidimensional sensor data without model assumptions and distribution form restrictions. By extracting health indicators from multidimensional sensor data and combining the kernel method of restricted quantile regression and support vector regression, the prediction of probability distribution is realized, which has the characteristics of strong applicability and high accuracy.

[0185] Based on the method for predicting the remaining useful life of equipment provided in the above embodiment, the present application also provides a specific implementation of a device for predicting the remaining useful life of equipment. Please refer to the following embodiment.

[0186] like Figure 7 As shown, the device 700 for predicting the remaining useful life of equipment provided in the embodiment of the present application includes the following modules:

[0187] A first acquisition module 710 is used to acquire status data of a device;

[0188] A first estimation module 720 is used to use a remaining useful life prediction model based on restricted quantile regression and a kernel method to perform quantile point estimation on the remaining useful life of the equipment according to the state data, and obtain quantile point prediction values ​​of the remaining useful life of the equipment corresponding to a plurality of quantile points, wherein the remaining useful life prediction model includes a first support vector regression model, a second support vector regression model and a random variable, and the first support vector regression model and the second support vector regression model both include a kernel function;

[0189] The second estimation module 730 is used to perform a robust estimation of the probability distribution based on a plurality of quantile point prediction values ​​to obtain a cumulative distribution function of the remaining useful life of the equipment.

[0190] The following is a detailed description of the device 700 for predicting the remaining service life of the above equipment, as shown below:

[0191] In some embodiments, the device 700 for predicting the remaining useful life of the equipment may further include:

[0192] A construction module is used to construct an initial remaining useful life prediction model based on restricted quantile regression and kernel method before using the remaining useful life prediction model based on restricted quantile regression and kernel method to perform quantile point estimation on the remaining useful life of the equipment according to the state data, wherein the initial remaining useful life prediction model includes a first initial support vector regression model, a second initial support vector regression model and a random variable, and the first initial support vector regression model and the second initial support vector regression model both include a kernel function;

[0193] The second acquisition module is used to acquire N historical remaining useful life data and their corresponding historical status data, where N is a positive integer;

[0194] The training module is used to train the initial remaining useful life prediction model using N historical remaining useful life data and their corresponding historical status data to obtain a remaining useful life prediction model.

[0195] In some embodiments, the training module may specifically include:

[0196] A first estimation submodule, used to estimate a first initial support vector regression model using N historical remaining useful life data and their corresponding historical status data, to obtain a first support vector regression model;

[0197] A first acquisition submodule is used to obtain a residual value of the first support vector regression model for each historical remaining useful life data and its corresponding historical status data, to obtain N residual values;

[0198] A fitting submodule is used to perform non-negative function fitting on the N residual values ​​and their corresponding historical remaining useful life data, determine the model parameters of the second initial support vector regression model, and obtain a second support vector regression model;

[0199] The second estimation submodule is used to estimate the random variable using the quantile regression model to obtain the estimated values ​​of the random variable corresponding to multiple quantile points;

[0200] A construction submodule is used to construct a remaining useful life prediction model based on the first support vector regression model, the second support vector regression model and the estimated values ​​of random variables corresponding to multiple quantile points.

[0201] In some embodiments, the second acquisition module may specifically include:

[0202] A second acquisition submodule is used to acquire N historical state data vectors, each of which includes first historical state data of m dimensions, where m is a positive integer;

[0203] A processing submodule, used for normalizing the first historical state data in the N historical state data vectors to obtain N target historical state data vectors;

[0204] The determination submodule is used to determine the N target historical state data vectors as N historical state data.

[0205] In some embodiments, the second acquisition module may further include:

[0206] The regression submodule is used to perform linear regression on the first historical state data of m dimensions for each target historical state data vector to obtain N historical state data.

[0207] In some embodiments, the second acquisition module may further include:

[0208] An analysis submodule, for performing a correlation analysis on the historical remaining useful life data and the first historical status data of the m dimensions before performing a linear regression on the first historical status data of the m dimensions, to obtain a correlation analysis result, wherein the correlation analysis results are sorted in descending order of correlation;

[0209] A selection submodule is used to select the first historical state data of the first n dimensions according to the correlation analysis result, where n is a positive integer;

[0210] Based on this, the regression submodule can specifically include:

[0211] The regression unit is used to perform linear regression on the first historical state data of n dimensions.

[0212] In some embodiments, the processing submodule may specifically include:

[0213] A clustering unit, used for clustering the first historical state data in the N historical state data vectors according to the environmental characteristics to obtain multiple categories;

[0214] The processing unit is used to perform intra-class normalization processing on the first historical state data belonging to the same category to obtain N target historical state data vectors.

[0215] In some embodiments, the device 700 for predicting the remaining useful life of the equipment may also include:

[0216] The third acquisition module is used to obtain the use time of the equipment, the replacement cost after a single failure, the single preventive replacement cost, and the downtime cost per unit time after obtaining the cumulative distribution function of the remaining service life of the equipment;

[0217] The solution module is used to optimize the total cost per unit time within the equipment replacement cycle according to the equipment usage time, replacement cost after a single failure, single preventive replacement cost, unit time downtime cost and cumulative distribution function, and determine the replacement time of the equipment. The replacement cycle is the time period between two consecutive replacement operations.

[0218] The device for predicting the remaining useful life of the equipment in the embodiment of the present application first uses the remaining useful life prediction model based on the restricted quantile regression and the kernel method to perform quantile point estimation on the remaining useful life of the equipment according to the state data, obtains multiple quantile point prediction values ​​of the remaining useful life of the equipment, and then performs robust estimation of the probability distribution based on the multiple quantile point prediction values ​​to obtain the cumulative distribution function of the remaining useful life of the equipment. The remaining useful life of the equipment can be predicted by the cumulative distribution function. Compared with predicting the remaining useful life of the equipment by using point estimation, the comprehensiveness of the prediction of the remaining useful life of the equipment can be improved, that is, the uncertainty of the prediction can be characterized. In addition, since the remaining useful life prediction model is constructed based on the restricted quantile regression and the kernel method, that is, no assumptions are made on the model and no restrictions are imposed on the distribution form, therefore, by predicting the remaining useful life distribution of the equipment by the remaining useful life prediction model based on the restricted quantile regression and the kernel method, the scope of application of the model can be improved, and the accuracy and flexibility of the prediction of the remaining useful life of the equipment can be improved.

[0219] Based on the method for predicting the remaining useful life of a device provided in the above-mentioned embodiment, the embodiment of the present application also provides a specific implementation of the electronic device. Figure 8 A schematic diagram of an electronic device 800 provided in an embodiment of the present application is shown.

[0220] The electronic device 800 may include a processor 810 and a memory 820 storing computer program instructions.

[0221] Specifically, the processor 810 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0222] The memory 820 may include a large capacity memory for data or instructions. By way of example and not limitation, the memory 820 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. In appropriate cases, the memory 820 may include a removable or non-removable (or fixed) medium. In appropriate cases, the memory 820 may be inside or outside the integrated gateway disaster recovery device. In a specific embodiment, the memory 820 is a non-volatile solid-state memory.

[0223] The memory may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk storage medium device, an optical storage medium device, a flash memory device, an electrical, optical or other physical / tangible memory storage device. Thus, typically, the memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., a memory device) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to the first aspect of the present application.

[0224] The processor 810 implements any of the methods for predicting the remaining useful life of a device in the above-mentioned embodiments by reading and executing computer program instructions stored in the memory 820 .

[0225] In one example, the electronic device 800 may further include a communication interface 830 and a bus 840. Figure 8 As shown, the processor 810, the memory 820, and the communication interface 830 are connected via a bus 840 and communicate with each other.

[0226] The communication interface 830 is mainly used to implement communication between various modules, devices, units and / or equipment in the embodiments of the present application.

[0227] Bus 840 includes hardware, software or both, and the parts of electronic equipment are coupled to each other. For example, but not limitation, bus may include accelerated graphics port (AGP) or other graphics bus, enhanced industrial standard architecture (EISA) bus, front side bus (FSB), hypertransport (HT) interconnection, industrial standard architecture (ISA) bus, infinite bandwidth interconnection, low pin count (LPC) bus, memory bus, micro channel architecture (MCA) bus, peripheral component interconnection (PCI) bus, PCI-Express (PCI-X) bus, serial advanced technology attachment (SATA) bus, video electronics standard association local (VLB) bus or other suitable bus or two or more of these combinations. In appropriate cases, bus 840 may include one or more buses. Although the present application embodiment describes and shows a specific bus, the application considers any suitable bus or interconnection.

[0228] Exemplarily, the electronic device 800 may be a mobile phone, a tablet computer, a laptop computer, a PDA, an in-vehicle electronic device, an ultra-mobile personal computer (UMPC), a netbook, or a personal digital assistant (PDA).

[0229] The electronic device can execute the method for predicting the remaining service life of the device in the embodiment of the present application, thereby realizing the combination Figures 1 to 7 A method and apparatus for predicting the remaining useful life of equipment are described.

[0230] In addition, in combination with the method for predicting the remaining service life of the equipment in the above embodiments, the embodiment of the present application can provide a computer-readable storage medium for implementation. The computer-readable storage medium stores computer program instructions; when the computer program instructions are executed by the processor, any method for predicting the remaining service life of the equipment in the above embodiments is implemented.

[0231] It should be clear that the present application is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present application is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present application.

[0232] The functional blocks shown in the above-described block diagram can be implemented as hardware, software, firmware or a combination thereof. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present application are programs or code segments that are used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link by a data signal carried in a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, optical fiber media, radio frequency (RF) links, etc. The code segment can be downloaded via a computer network such as the Internet, an intranet, etc.

[0233] It should also be noted that the exemplary embodiments mentioned in this application describe some methods or systems based on a series of steps or devices. However, this application is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiment, or in a different order from the embodiment, or several steps can be performed simultaneously.

[0234] The above reference is according to the method of the embodiment of the present application, the flow chart of the device (system) and the computer program product and / or the block diagram described various aspects of the present application.It should be understood that each square box in the flow chart and / or the block diagram and the combination of each square box in the flow chart and / or the block diagram can be realized by computer program instructions.These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer or other programmable data processing device to produce a machine so that these instructions executed by the processor of the computer or other programmable data processing device enable the realization of the function / action specified in one or more square boxes of the flow chart and / or the block diagram.Such a processor can be but is not limited to a general-purpose processor, a special-purpose processor, a special application processor or a field programmable logic circuit.It can also be understood that each square box in the block diagram and / or the flow chart and the combination of the square boxes in the block diagram and / or the flow chart can also be realized by the dedicated hardware that performs the specified function or action, or can be realized by the combination of dedicated hardware and computer instructions.

[0235] The above is only a specific implementation of the present application. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, modules and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. It should be understood that the protection scope of the present application is not limited to this. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application.

Claims

1. A method for predicting the remaining useful life of equipment, characterized in that: include: Get device status data; Using a remaining useful life prediction model based on restricted quantile regression and kernel method, quantile point estimation is performed on the remaining useful life of the equipment according to the state data to obtain quantile point prediction values ​​of the remaining useful life of the equipment corresponding to multiple quantile points, respectively, wherein the remaining useful life prediction model includes a first support vector regression model, a second support vector regression model and a random variable, and the first support vector regression model and the second support vector regression model both include a kernel function; A robust estimation of the probability distribution is performed based on the plurality of quantile point prediction values ​​to obtain a cumulative distribution function of the remaining useful life of the equipment.

2. The method according to claim 1, characterized in that Before using the remaining useful life prediction model based on restricted quantile regression and kernel method to perform quantile point estimation on the remaining useful life of the equipment according to the status data, the method further includes: An initial remaining useful life prediction model is constructed based on restricted quantile regression and kernel method, wherein the initial remaining useful life prediction model includes a first initial support vector regression model, a second initial support vector regression model and a random variable, wherein the first initial support vector regression model and the second initial support vector regression model both include a kernel function; Obtain N historical remaining useful life data and their corresponding historical status data, where N is a positive integer; The initial remaining useful life prediction model is trained using the N historical remaining useful life data and their respectively corresponding historical status data to obtain the remaining useful life prediction model.

3. The method according to claim 2, characterized in that The method of training the initial remaining useful life prediction model by using the N historical remaining useful life data and their corresponding historical status data to obtain the remaining useful life prediction model includes: Using the N historical remaining useful life data and their corresponding historical status data, the first initial support vector regression model is estimated to obtain a first support vector regression model; Obtaining a residual value of the first support vector regression model for each of the historical remaining useful life data and its corresponding historical status data to obtain N residual values; Performing non-negative function fitting on the N residual values ​​and their corresponding historical remaining useful life data, determining model parameters of the second initial support vector regression model, and obtaining a second support vector regression model; The random variable is estimated using a quantile regression model to obtain estimated values ​​of the random variable corresponding to the multiple quantile points respectively; The remaining useful life prediction model is constructed according to the first support vector regression model, the second support vector regression model and the estimated values ​​of the random variables corresponding to the multiple quantile points.

4. The method according to claim 2, characterized in that: Obtaining N pieces of historical status data, including: Acquire N historical state data vectors, each of which includes first historical state data of m dimensions, where m is a positive integer; Normalizing the first historical state data in the N historical state data vectors to obtain N target historical state data vectors; The N target historical state data vectors are determined as the N historical state data.

5. The method according to claim 4, characterized in that Acquiring N pieces of historical status data also includes: For each of the target historical state data vectors, linear regression is performed on the first historical state data of the m dimensions to obtain N historical state data.

6. The method according to claim 5, characterized in that Before performing linear regression on the first historical state data of the m dimensions, the method further includes: Performing a correlation analysis on the historical remaining useful life data and the first historical status data of the m dimensions to obtain a correlation analysis result, wherein the correlation analysis results are sorted in descending order of correlation; According to the correlation analysis result, selecting the first historical state data of the first n dimensions, where n is a positive integer; The performing linear regression on the first historical state data of the m dimensions includes: Perform linear regression on the first historical state data of the n dimensions.

7. The method according to claim 4, characterized in that The normalizing the first historical state data in the N historical state data vectors to obtain N target historical state data vectors includes: Clustering the first historical state data in the N historical state data vectors according to environmental characteristics to obtain multiple categories; The first historical state data belonging to the same category are normalized within the category to obtain N target historical state data vectors.

8. The method according to claim 1, characterized in that After obtaining the cumulative distribution function of the remaining useful life of the equipment, the method further includes: Obtain the service life of the equipment, replacement cost after a single failure, single preventive replacement cost, and downtime cost per unit time; The total cost per unit time within the replacement cycle of the equipment is optimized based on the service life of the equipment, the replacement cost after a single failure, the single preventive replacement cost, the downtime cost per unit time and the cumulative distribution function to determine the replacement time of the equipment. The replacement cycle is the time period between two consecutive replacement operations.

9. A device for predicting the remaining useful life of equipment, characterized in that: The device comprises: A first acquisition module, used to acquire status data of a device; a first estimation module, configured to use a remaining useful life prediction model based on restricted quantile regression and a kernel method to perform quantile point estimation on the remaining useful life of the equipment according to the state data, and obtain quantile point prediction values ​​of the remaining useful life of the equipment corresponding to a plurality of quantile points, wherein the remaining useful life prediction model comprises a first support vector regression model, a second support vector regression model and a random variable, and both the first support vector regression model and the second support vector regression model comprise a kernel function; The second estimation module is used to perform a robust estimation of the probability distribution based on the plurality of quantile point prediction values ​​to obtain a cumulative distribution function of the remaining useful life of the equipment.

10. An electronic device, characterized in that: The electronic device comprises: a processor and a memory storing computer program instructions; When the processor executes the computer program instructions, the method for predicting the remaining useful life of the equipment as described in any one of claims 1 to 8 is implemented.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions, and when the computer program instructions are executed by a processor, the method for predicting the remaining useful life of a device according to any one of claims 1 to 8 is implemented.

12. A computer program product, characterized in that When the instructions in the computer program product are executed by a processor of an electronic device, the electronic device executes the method for predicting the remaining useful life of a device as described in any one of claims 1 to 8.

Citation Information

Cited By

  • Product information determination method and device, electronic equipment and medium

    CN119227020A

  • Product information determination method, device, electronic device and medium

    CN119227020B