Product residual life prediction method and device, electronic equipment and storage medium

Abstract

Embodiments of the present application provide a product residual life prediction method and device, electronic equipment and storage medium. The method comprises: obtaining degradation data of a product, and extracting a degradation curve according to the degradation data; performing degradation modeling on the degradation curve according to a Wiener process to obtain a degradation model, and determining offline parameters according to the degradation model; dividing the degradation curve into a training set, a test set and a validation set, and performing long short-term neural network training; establishing a state transition equation and a measurement equation according to the Wiener process, and sampling from the offline parameters; predicting a system state at an n-th moment according to a long short-term memory neural network, recursively performing the state transition equation to the n-th moment according to a particle parameter at a current moment, and synthesizing a particle weight in a recursive process as a particle weight at the current moment; performing normalization according to the particle weight to obtain a residual life probability density of the product at the current moment; and performing parameter updating according to real state data obtained at each moment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of failure prediction and health management, and particularly relates to a product residual life prediction method and device, an electronic device and a storage medium. BACKGROUND

[0002] With the rapid development of modern industry and the information age, a new generation of industrial revolution has swept the world, and the speed of industrial modernization is gradually accelerating. Whether it is a product or equipment, it becomes more and more complex with the improvement of people's demand, and its failure mode is also changing. For some potential failure modes, people are difficult to distinguish their failure through senses, and simple detection often leads to performance degradation in the overhaul of expensive and precise equipment. Based on this, in recent years, failure prediction and health management (PHM) has become more and more extensive in both academia and industrial applications. The PHM method allows the reliability of the system to be evaluated and the residual life to be predicted under the actual life cycle conditions of the system, so as to predict the time and place where the failure will occur, thereby eliminating the risk of the system.

[0003] Although neural networks have made outstanding achievements in residual life prediction in recent years, neural networks cannot give the failure probability density for prediction, which indirectly leads to the loss of some decision-making methods for product users. At present, the neural network method has gradually become the mainstream in residual life prediction, but since the neural network method is a "black box", the existing research cannot explain the real principle of the neural network, which also leads to the fact that the neural network method cannot obtain the probability density of product failure, and the prediction accuracy is low. SUMMARY

[0004] The main purpose of the embodiments of the present application is to propose a product residual life prediction method and device, an electronic device and a storage medium, which can combine the prediction ability of neural networks with particle filtering, not only can give the failure probability density, but also has higher prediction accuracy than ordinary particle filtering.

[0005] To achieve the above purpose, a first aspect of the embodiments of the present application proposes a product residual life prediction method, which comprises:

[0006] Obtaining degradation data of a product, and extracting a degradation curve according to the degradation data;

[0007] Performing degradation modeling on the degradation curve according to a Wiener process to obtain a degradation model, and determining offline parameters according to the degradation model;

[0008] Dividing the degradation curve into a training set, a test set and a validation set, and performing long short-term neural network training;

[0009] State transition equations and measurement equations are established based on the Wiener process, and samples are taken from the offline parameters;

[0010] Based on the prediction of the system state at time n using the long short-term memory neural network, the state transition equation is recursively applied to time n based on the particle parameters at the current time, and the particle weights in the recursion process are combined to obtain the particle weights at the current time.

[0011] Normalize based on particle weights to obtain the probability density of the product's remaining lifetime at the current moment;

[0012] The parameters are updated based on the actual state data acquired at each moment.

[0013] In some embodiments, the degradation model is:

[0014] Where X(t) k ) for t k The device degradation state is defined by time step, k is the number of observations; λ is the drift parameter, used to characterize the degradation rate; μ(τ; θ) represents the time-varying nonlinear function of parameter θ, used to describe the nonlinearity of the device degradation state; B(t k ) represents standard Brownian motion, and σ is the diffusion parameter, which are used to describe the time-varying random fluctuations and the degree of fluctuations in the degradation process, respectively.

[0015] The value of the fixed parameter θ, and the distribution of the random parameters λ and σ are obtained by a two-stage method.

[0016] In some embodiments, establishing the state transition equations and measurement equations based on the Wiener process includes:

[0017] Observation sequences will be used The parameter estimation problem is treated as a recursive Bayesian filtering problem, estimating the posterior distribution.

[0018]

[0019] in, For the end t k The online degradation state sequence at any given moment. For t k The estimated model parameters are obtained at each time step; the offline parameters are updated using the observation sequence to construct a state-space model.

[0020]

[0021] in,

[0022] In some embodiments, the step of predicting the system state at time n based on a long short-term memory neural network and recursively applying the state transition equation to time n based on the particle parameters at the current time includes:

[0023] The degradation sequence predicted by the neural network is defined as... Keep The parameters remain unchanged, and then based on the formula Recursion can be made up to t k+n At each step of the recursion, the weights are calculated.

[0024] In some embodiments, the sum of particle weights during the recursive process is the particle weight at the current moment, including:

[0025] Note t k The weighted particle set of parameters at time t is M is the number of particles.

[0026]

[0027] in, For in t k-1 The predicted particle state at time t k The weights assigned to the likelihood between the actual states of the system at different times are... N represents the number of steps the neural network takes to predict the state, which is used to transform the system's true state into the state predicted by the long short-term neural network.

[0028] In some embodiments, the step of normalizing according to particle weights to obtain the probability density of the product's remaining lifetime at the current moment includes:

[0029] Based on the principle of particle filtering, at t k Available at any time

[0030]

[0031]

[0032] in,

[0033]

[0034] For m = 1, 2, ..., M, normalize according to the weights to obtain...

[0035]

[0036] Therefore, in t k The formula for calculating the probability density of the remaining lifespan of a product that is constantly predicting degradation is as follows:

[0037]

[0038] In some embodiments, after normalization based on particle weights, the following is included:

[0039] Particles with larger weight values ​​are copied, while particles with smaller weight values ​​are discarded. An effective sampling size is introduced, defined as:

[0040]

[0041] Where, N eff The smaller the value, the more severe the particle degradation phenomenon. The resampling threshold is set to N. th When N eff <N th If so, resampling will be performed.

[0042] To achieve the above objectives, a second aspect of this application provides a product remaining life prediction device, the device comprising:

[0043] The acquisition module is used to acquire product degradation data and extract degradation curves based on the degradation data.

[0044] The modeling module is used to perform degradation modeling on the degradation curve according to the Wiener process, obtain the degradation model, and determine the offline parameters based on the degradation model;

[0045] The training module is used to divide the degradation curve into a training set, a test set, and a validation set for training a long short-term neural network.

[0046] A module is established to establish state transition equations and measurement equations based on the Wiener process and to sample from the offline parameters;

[0047] The recursive module is used to predict the system state at time n based on the long short-term memory neural network, and recursively perform the state transition equation to time n based on the particle parameters at the current time. The particle weights in the recursive process are combined as the particle weights at the current time.

[0048] The prediction module is used to normalize based on particle weights to obtain the probability density of the product's remaining lifetime at the current moment.

[0049] The update module is used to update parameters based on the real-time state data obtained at each moment.

[0050] To achieve the above objectives, a third aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method described in the first aspect.

[0051] To achieve the above objectives, a fourth aspect of the present application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in the first aspect.

[0052] This application proposes a product remaining life prediction method, apparatus, electronic device, and storage medium. The method involves acquiring product degradation data and extracting degradation curves from the data. Degradation models are then built using the Wiener process to determine offline parameters. The degradation curves are divided into training, testing, and validation sets for training a long short-term memory (LSTM) neural network. State transition equations and measurement equations are established based on the Wiener process, and samples are taken from the offline parameters. The LSM neural network predicts the system state at time n, and the state transition equation is recursively applied to time n based on the particle parameters at the current time. The particle weights accumulated during the recursion are used as the particle weights at the current time. Normalization is performed on the particle weights to obtain the product's remaining life probability density at the current time. Parameters are updated based on the actual state data acquired at each time step. Based on this, this application proposes a product remaining life prediction method that integrates LSM neural networks and particle filtering. This method achieves good prediction accuracy for the remaining life of products. This method not only calculates the remaining lifetime density curve of a product, but also increases the prediction accuracy based on ordinary particle filtering by integrating a long short-term memory neural network, reducing particle degradation caused by noise. This provides a theoretical basis and technical support for the sustainable development of product safety and economy. The embodiments of this application combine the predictive power of neural networks with particle filtering, which not only provides the probability density of failure, but also achieves higher prediction accuracy than ordinary particle filtering. Attached Figure Description

[0053] Figure 1 This is a flowchart of the product remaining life prediction method provided in the embodiments of this application;

[0054] Figure 2 Characteristic curves of NASA's turbofan engine;

[0055] Figure 3 The probability density curve for the remaining lifespan of engine number 258;

[0056] Figure 4 The probability density curve for the remaining lifespan of engine number 244;

[0057] Figure 5 The probability density curve for the remaining lifespan of engine No. 208;

[0058] Figure 6 The probability density curve for the remaining lifespan of engine number 256;

[0059] Figure 7 This is a schematic diagram of the product remaining life prediction device provided in the embodiments of this application;

[0060] Figure 8 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0062] It should be noted that although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first," "second," etc., in the specification, claims, and the aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0063] To address the technical problem of low prediction accuracy in existing technologies due to the inability to provide a device's remaining lifetime density curve, this application provides a product remaining lifetime prediction method, apparatus, electronic device, and storage medium. The method involves acquiring product degradation data and extracting a degradation curve from the data; performing degradation modeling on the degradation curve using the Wiener process to obtain a degradation model; determining offline parameters based on the degradation model; dividing the degradation curve into training, testing, and validation sets for training a long short-term memory neural network; establishing state transition equations and measurement equations based on the Wiener process and sampling from the offline parameters; predicting the system state at time n using the long short-term memory neural network; recursively applying the state transition equation to time n based on the particle parameters at the current time; combining the particle weights during the recursion process as the particle weights at the current time; normalizing the particle weights to obtain the product's remaining lifetime probability density at the current time; and updating the parameters based on the actual state data acquired at each time step. Based on this, this application proposes a product remaining lifetime prediction method that integrates a long short-term memory neural network and particle filtering. This method provides good prediction accuracy for the remaining lifespan of products. This method not only calculates the remaining lifetime density curve of a product, but also increases the prediction accuracy based on ordinary particle filtering by integrating a long short-term memory neural network, reducing particle degradation caused by noise. This provides a theoretical basis and technical support for the sustainable development of product safety and economy. The embodiments of this application combine the predictive power of neural networks with particle filtering, which not only provides the probability density of failure, but also achieves higher prediction accuracy than ordinary particle filtering.

[0064] The product remaining life prediction method, apparatus, electronic device, and storage medium provided in this application are specifically described through the following embodiments. First, the product remaining life prediction method in this application is described.

[0065] Figure 1 This is an optional flowchart of the product remaining life prediction method provided in the embodiments of this application. Figure 1 The method may include, but is not limited to, steps S101 to S107.

[0066] Step S101: Obtain the degradation data of the product and extract the degradation curve based on the degradation data;

[0067] Step S102: Perform degradation modeling on the degradation curve according to the Wiener process to obtain the degradation model, and determine the offline parameters based on the degradation model;

[0068] Step S103: Divide the degradation curve into a training set, a test set, and a validation set, and train the long short-term neural network.

[0069] Step S104: Establish the state transition equation and measurement equation based on the Wiener process, and sample from offline parameters;

[0070] Step S105: Based on the prediction of the system state at time n by the long short-term memory neural network, the state transition equation is recursively applied to time n based on the particle parameters at the current time, and the particle weights in the recursion process are combined as the particle weights at the current time.

[0071] Step S106: Normalize according to particle weights to obtain the probability density of the remaining lifetime of the product at the current moment.

[0072] Step S107: Update the parameters based on the real state data acquired at each moment.

[0073] In some embodiments, degradation data of a product is acquired, wherein the product may include, but is not limited to, bearings, lithium batteries, engines, etc. A trend-based degradation curve is extracted based on the characteristics of the product degradation data. Degradation modeling is then performed on the degradation curve using the Wiener process. It should be noted that this application is based on the Wiener process, and therefore has greater universality than existing technologies.

[0074] In some embodiments, the degradation curve of the product is divided into a training set, a test set, and a validation set. The samples in the training set are used to determine the product's failure threshold and to train the neural network. The test set is used to evaluate the training effect of the neural network, and the validation set is used to evaluate the effectiveness of the proposed method.

[0075] In some embodiments, the value of the fixed parameter θ and the distribution of the random parameters λ and σ are determined using a two-stage method based on samples from the training set. For products with low security monitoring requirements, reliability assessment can be performed based on the definition of first-arrival time. If high security monitoring is required for individual products, online parameter updates are necessary.

[0076] In some embodiments, degradation data of the product is acquired, and degradation curves are extracted from the degradation data. Degradation modeling is performed on the degradation curves according to the Wiener process to obtain a degradation model, and offline parameters are determined based on the degradation model. The degradation curves are divided into training, testing, and validation sets for training a long short-term memory neural network. State transition equations and measurement equations are established according to the Wiener process, and samples are taken from the offline parameters. The system state at time n is predicted using the long short-term memory neural network, and the state transition equation is recursively applied to time n based on the particle parameters at the current time. The particle weights in the recursive process are combined as the particle weights at the current time. Normalization is performed based on the particle weights to obtain the probability density of the product's remaining lifetime at the current time. Parameters are updated based on the real state data acquired at each time step. Based on this, by combining the predictive power of the neural network with particle filtering, not only can the probability density of failure be given, but its prediction accuracy is also higher than that of ordinary particle filtering. This method can derive the probability density curve of the remaining lifetime, providing product users with a wider range of product maintenance strategies.

[0077] In some embodiments, this method is a product remaining life prediction method that integrates a long short-term memory neural network and particle filtering. The method predicts the remaining life of a product with good accuracy. This method not only calculates the product's remaining life density curve, but also increases the prediction accuracy based on ordinary particle filtering by integrating a long short-term memory neural network, reducing particle degradation caused by noise. This provides a theoretical basis and technical support for the sustainable development of product safety and economy.

[0078] The following specific embodiments further illustrate the product remaining lifetime prediction method that integrates long short-term memory neural networks and particle filtering provided in this application, specifically including the following steps:

[0079] 1. Obtain product degradation data and extract trend-based degradation curves based on the characteristics of the product degradation data.

[0080] 2. Based on the Wiener process, a degradation model was constructed for the degradation curve, resulting in the following degradation model:

[0081]

[0082] Where X(t) k ) for t kThe device degradation state is defined by time step, k is the number of observations; λ is the drift parameter, used to characterize the degradation rate; μ(τ; θ) represents the time-varying nonlinear function of parameter θ, used to describe the nonlinearity of the device degradation state; B(t k Let μ(τ) be the standard Brownian motion, and σ be the diffusion parameter, used to describe the time-varying random fluctuations and the degree of fluctuation in the degradation process, respectively. Different functional forms of μ(τ; θ) can describe different forms of the degradation process. For example, when... The time-degradation process becomes a linear degradation process. It can be used to represent the degeneracy process of a power function.

[0083] 3. The product degradation curve is divided into a training set, a test set, and a validation set. The samples in the training set are used to determine the product's failure threshold and to train the neural network. The test set is used to evaluate the training effect of the neural network, and the validation set is used to evaluate the effectiveness of the proposed method.

[0084] 4. Based on the samples in the training set, determine the value of the fixed parameter θ, and the distributions of the random parameters λ and σ, using a two-stage method. For products with low safety monitoring requirements, reliability can be assessed based on the definition of first-arrival time.

[0085]

[0086]

[0087] Where K is the total number of samples drawn from the distribution of random parameters λ and σ, m is the index of the sample, and w is the failure threshold.

[0088] 5. If there are high requirements for the safety monitoring of individual products, online parameter updates are necessary. The situations described above only reflect the general characteristics of the degradation process of similar equipment and are insufficient to meet the remaining life prediction requirements for specific serviceable equipment that are of greater concern in engineering. Considering the gradual acquisition of degradation data, observation sequences can be utilized. The parameter estimation problem can be viewed as a recursive Bayesian filtering problem, i.e., estimating the posterior distribution.

[0089]

[0090] in, For the end t k The online degradation state sequence at any given moment. For t k The model parameters are estimated at each time step; the offline parameters are updated using the observation sequence to better reflect individual degradation differences, and a state-space model is constructed as follows.

[0091]

[0092] in, Due to parameter B(t) k The presence of ) leads to the current high weight of particles being caused by noise; therefore, based on the Long Short-Term Memory neural network, the subsequent t... k+n Prediction of time. The degradation sequence predicted by the neural network is defined as... Keep With the parameters unchanged, it can be recursively derived to t based on equation (5). k+n At each step of the recursion, the weights are calculated. Let t be the time interval. k The weighted particle set of parameters at time t is M is the number of particles.

[0093]

[0094] in, For in t k-1 The predicted particle state at time t k Similarly, the weights assigned to the likelihood between the system's true states at different times are... This can be understood as transforming the system's true state into the state predicted by a long short-term neural network, where N is the number of prediction steps. The initial conditions are the process estimated in step 4, i.e., random sampling from the distributions followed by λ and σ. According to the principle of particle filtering, at t k Available at any time

[0095]

[0096]

[0097] in,

[0098]

[0099] For m = 1, 2, ..., M, normalize according to the weights to obtain...

[0100]

[0101] When executing the particle filtering algorithm, severe particle degradation occurs, meaning only a small number of particles have high importance weights, while most particles have very low importance weights. Therefore, a resampling step is added to duplicate particles with higher weights and discard those with lower weights. To measure the degree of particle degradation, an effective sampling size is introduced, defined as...

[0102]

[0103] Where, N eff The smaller the value, the more severe the particle degradation phenomenon. The resampling threshold is set to N. th When Neff <N th When t < 0, resampling is performed. Therefore, at t < 0, resampling is performed. k The formula for calculating the probability density of the remaining lifespan of a product that is constantly predicting degradation is as follows:

[0104]

[0105] 6. Combining Figures 2 to 6 As shown, a comparative experiment was conducted using the NASA turbofan engine dataset, which was developed by the NASA Army Research Laboratory using the Commercial Modular Aero Propulsion System Simulation (C-MAPSS). Twenty-one features were selected from the system output to characterize the engine degradation process. The FD002 training set was used as the experimental sample. A mean square error (MSE) metric was introduced, which considers both the accuracy and uncertainty of the remaining life prediction. k The definition of time is as follows:

[0106]

[0107] In equation (13): For t k The MSE value represents the actual remaining lifetime of the device at any given time. A smaller MSE value indicates higher accuracy in the predicted remaining lifetime and more precise results. Table 1 below shows the experimental results for the MSE on the validation set:

[0108] MSE Particle filtering without neural networks 1.8686e+03 Particle filtering with long short-term memory neural networks 1.8341e+03

[0109] As shown in Table 1, the MSE value of the particle filter with long short-term memory neural network is smaller than that of the particle filter without neural network, indicating that the product remaining life prediction accuracy of the proposed method of integrating long short-term memory neural network and particle filter is higher and the results are more accurate.

[0110] Please see Figure 7 This application also provides a product remaining life prediction device, which can implement the above-described product remaining life prediction method. The device includes:

[0111] The acquisition module 710 is used to acquire the degradation data of the product and extract the degradation curve based on the degradation data;

[0112] Modeling module 720 is used to perform degradation modeling on degradation curves based on the Wiener process, obtain degradation model, and determine offline parameters based on degradation model;

[0113] Training module 730 is used to divide the degradation curve into training set, test set and validation set for training long short-term neural network;

[0114] Module 740 is established to establish state transition equations and measurement equations based on the Wiener process and to sample from offline parameters;

[0115] The recursive module 750 is used to predict the system state at time n based on the long short-term memory neural network, and recursively perform the state transition equation to time n based on the particle parameters at the current time, and integrates the particle weights in the recursive process as the particle weights at the current time.

[0116] The prediction module 760 is used to normalize based on particle weights to obtain the probability density of the product's remaining lifetime at the current moment.

[0117] The update module 770 is used to update parameters based on the real state data acquired at each moment.

[0118] Based on this, the product remaining life prediction device of this application embodiment includes: an acquisition module 710 acquiring product degradation data and extracting degradation curves based on the degradation data, the degradation curves being used to characterize the probability density of product failure; a modeling module 720 performing degradation modeling on the degradation curves according to the Wiener process to obtain a degradation model, and determining offline parameters based on the degradation model; a training module 730 dividing the degradation curves into training sets, test sets, and validation sets for training a long short-term neural network; an establishment module 740 establishing state transition equations and measurement equations according to the Wiener process, and sampling from the offline parameters; a recursion module 750 predicting the system state at time n based on the long short-term memory neural network, recursively performing state transition equations to time n based on the particle parameters at the current time, and combining the particle weights in the recursion process as the particle weights at the current time; a prediction module 760 normalizing based on the particle weights to obtain the probability density of the product's remaining life at the current time; and an update module 770 updating the parameters based on the real state data acquired at each time. This application proposes a product remaining lifetime prediction method that integrates long short-term memory neural networks and particle filtering. The method acquires product degradation data, extracts degradation curves from the data, models the degradation curves using the Wiener process, and determines offline parameters based on the model. The degradation curves are divided into training, testing, and validation sets for training a long short-term neural network. State transition equations and measurement equations are established based on the Wiener process, and samples are taken from the offline parameters. The system state at time n is predicted using the long short-term memory neural network, and the state transition equation is recursively applied to time n based on the particle parameters at the current time. The particle weights accumulated during the recursion are used as the particle weights at the current time. Normalization is performed based on the particle weights to obtain the product's remaining lifetime probability density at the current time. Parameters are updated based on the actual state data acquired at each time step. Based on this, this application proposes a product remaining lifetime prediction method that integrates long short-term memory neural networks and particle filtering. This method predicts the remaining lifetime of products with good accuracy. This method not only obtains the product's remaining lifetime density curve but also increases the prediction accuracy based on ordinary particle filtering by integrating long short-term memory neural networks, reducing particle degradation caused by noise. This provides a theoretical basis and technical support for the sustainable development of product safety and economy. The embodiments of this application combine the predictive power of neural networks with particle filtering, which not only provides the probability density of failures, but also has higher prediction accuracy than ordinary particle filtering.

[0119] The specific implementation of the product remaining life prediction device is basically the same as the specific implementation of the product remaining life prediction method described above, and will not be repeated here.

[0120] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the aforementioned product remaining life prediction method. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0121] Please see Figure 8 , Figure 8 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes:

[0122] The processor 801 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application.

[0123] The memory 802 can be implemented in the form of read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory 802 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 802 and called by the processor 801 to execute the product remaining life prediction method of the embodiments of this application. Specifically, the method involves acquiring product degradation data, extracting degradation curves based on the degradation data, performing degradation modeling on the degradation curves according to the Wiener process to obtain a degradation model, determining offline parameters based on the degradation model, dividing the degradation curves into training, testing, and validation sets, and training a long short-term neural network (LSN), establishing state transition equations and measurement equations according to the Wiener process, and sampling from the offline parameters, predicting the system state at time n based on the LSN, recursively applying the state transition equations to time n based on the particle parameters at the current time, and combining the particle weights in the recursion process as the particle weights at the current time, normalizing based on the particle weights to obtain the product's remaining life probability density at the current time, and updating the parameters based on the real state data acquired at each time. Based on this, this application proposes a product remaining life prediction method that integrates a long short-term memory neural network and particle filtering. The method predicts the remaining life of the product with good accuracy. This method not only calculates the remaining lifetime density curve of a product, but also increases the prediction accuracy based on ordinary particle filtering by integrating a long short-term memory neural network, reducing particle degradation caused by noise. This provides a theoretical basis and technical support for the sustainable development of product safety and economy. The embodiments of this application combine the predictive power of neural networks with particle filtering, which not only provides the probability density of failure, but also achieves higher prediction accuracy than ordinary particle filtering.

[0124] The input / output interface 803 is used to implement information input and output.

[0125] The communication interface 804 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0126] A bus transmits information between various components of a device (e.g., processor 801, memory 802, input / output interface 803, and communication interface 804).

[0127] The processor 801, memory 802, input / output interface 803, and communication interface 804 communicate with each other within the device via a bus.

[0128] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described product remaining life prediction method.

[0129] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0130] The product remaining useful life prediction method, device, electronic device, and storage medium provided in this application's embodiments acquire product degradation data and extract degradation curves from the degradation data; perform degradation modeling on the degradation curves using the Wiener process to obtain a degradation model, and determine offline parameters based on the degradation model; divide the degradation curves into training, testing, and validation sets for training a long short-term neural network; establish state transition equations and measurement equations based on the Wiener process, and sample from the offline parameters; predict the system state at time n based on the long short-term memory neural network, recursively apply the state transition equations to time n based on the particle parameters at the current time, and combine the particle weights in the recursion process as the particle weights at the current time; normalize based on the particle weights to obtain the product's remaining useful life probability density at the current time; and update the parameters based on the real state data acquired at each time. Based on this, this application proposes a product remaining useful life prediction method that integrates a long short-term memory neural network and particle filtering. The method predicts the remaining useful life of products with good accuracy. This method not only calculates the remaining lifetime density curve of a product, but also increases the prediction accuracy based on ordinary particle filtering by integrating a long short-term memory neural network, reducing particle degradation caused by noise. This provides a theoretical basis and technical support for the sustainable development of product safety and economy. The embodiments of this application combine the predictive power of neural networks with particle filtering, which not only provides the probability density of failure, but also achieves higher prediction accuracy than ordinary particle filtering.

[0131] It will be understood by those skilled in the art that all or some of the steps and systems in the methods disclosed above can be implemented as software, firmware, hardware, and suitable combinations thereof. Some or all of the physical components can be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable programs, data structures, program modules, or other data). Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, as is known to those skilled in the art, communication media typically contain computer-readable programs, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0132] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0133] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0134] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0135] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0136] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0137] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0138] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0139] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0140] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0141] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0142] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A product residual life prediction method characterized by, The method comprises: obtaining degradation data of a product, and extracting a degradation curve according to the degradation data; performing degradation modeling on the degradation curve according to a Wiener process to obtain a degradation model, and determining offline parameters according to the degradation model; dividing the degradation curve into a training set, a test set and a validation set, and performing training of a long short-term neural network; establishing a state transition equation and a measurement equation according to the Wiener process, and sampling from the offline parameters; predicting a system state at an n-th moment according to a long short-term memory neural network, recursively performing the state transition equation to the n-th moment according to particle parameters at a current moment, and integrating particle weights in a recursive process as particle weights at the current moment; normalizing the particle weights to obtain a probability density of a remaining life of the product at the current moment; updating parameters according to real state data obtained at each moment; wherein the degradation model is: where, is the degradation state of the equipment at time is the number of observations; is the drift parameter, which characterizes the degradation rate; represents a time-varying nonlinear function of the parameter to describe the nonlinearity of the degradation state of the equipment; is a standard Brownian motion, is a diffusion parameter, respectively, to describe the time-varying random fluctuations and the degree of fluctuations in the degradation process; The values of the fixed parameters and the distribution of the random parameters and are found by a two-stage method; the establishing of the state transition equation and the measurement equation according to the Wiener process comprises: The observation sequence will be utilized The parameter estimation problem is considered as a recursive Bayesian filtering problem, estimating the posterior distribution wherein, is the cutoff sequence of online degradation states at time instant is the model parameters to be estimated at time instant ; the offline parameters are updated using the observation sequence, and the state space model is constructed as wherein ; the predicting of the system state at the n-th moment according to the long short-term memory neural network, the recursive performance of the state transition equation to the n-th moment according to the particle parameters at the current moment comprises: The degradation sequence predicted according to the neural network is defined as , the parameters are kept unchanged and then the calculation of the weights is performed at each step of the recursion based on the equation for the instant . the integrating of the particle weights in the recursive process as the particle weights at the current moment comprises: Recall the weighted particle set of the parameter at the time instant is , is the number of particles, wherein, is the likelihood of the predicted state of the system at time and the true state of the system at time is the weight assigned to the likelihood of the predicted state of the system at time is the predicted state of the system at time is the number of steps predicted by the neural network. the normalizing of the particle weights to obtain the probability density of the remaining life of the product at the current moment comprises: According to the principle of particle filter, at time t, the wherein δ() is a Dirac function, and wherein , For , according to the weight, normalization is performed to obtain Therefore, in The formula for predicting the residual life probability density of the degraded product at time t is: 。 2. The method of claim 1, wherein, after the normalizing of the particle weights, comprising: copying particles with larger weight values and discarding particles with smaller weight values, introducing an effective sampling size, defined as: wherein, The smaller, means the particle degradation phenomenon is more serious, set the resampling threshold for When resampling is performed.​ 3. A product residual life prediction device characterized by comprising: The device is used to implement the product remaining life prediction method in any one of claims 1 to 2, and the device comprises: an acquisition module configured to obtain degradation data of a product, and extract a degradation curve according to the degradation data; a modeling module configured to perform degradation modeling on the degradation curve according to a Wiener process to obtain a degradation model, and determine offline parameters according to the degradation model; a training module configured to divide the degradation curve into a training set, a test set and a validation set, and perform training of a long short-term neural network; an establishing module configured to establish a state transition equation and a measurement equation according to the Wiener process, and sample from the offline parameters; a recursive module configured to predict a system state at an n-th moment according to a long short-term memory neural network, recursively perform the state transition equation to the n-th moment according to particle parameters at a current moment, and integrate particle weights in a recursive process as particle weights at the current moment; a prediction module configured to normalize the particle weights to obtain a probability density of a remaining life of the product at the current moment; an updating module configured to update parameters according to real state data obtained at each moment.

4. An electronic device, comprising: The electronic device comprises a memory and a processor, the memory stores a computer program, and the processor implements the product remaining life prediction method in any one of claims 1 to 2 when executing the computer program.

5. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 4. The computer program is executed by the processor to implement the product remaining life prediction method in any one of claims 1 to 2.

Images