Method and system for predicting residual service life of equipment

Abstract

The invention discloses an equipment remaining service life prediction method and system, and belongs to the technical field of equipment state monitoring, and the method comprises the steps: obtaining the whole life cycle degradation data of equipment as a data set; performing data dimension reduction by using an automatic encoder to obtain an equipment health index HI; inputting the HI into a fusion type RUL prediction network model, and predicting a probability density curve of the RUL; wherein the model comprises a probability prediction network, a Wiener process model and a fusion module; the probability prediction network is used for obtaining a probability density curve in a numerical form; the Wiener process model is used for obtaining a probability density curve in an analysis form; and the fusion module is used for dynamically weighting the two curves to realize RUL prediction. According to the method, the RUL probability density curve prediction is realized, the problem that an existing prediction model based on machine learning can only obtain a point prediction result is solved, and the interpretability of the model is improved in combination with a random process.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment status monitoring, and in particular to a method and system for predicting the remaining service life of equipment. Background Art

[0002] With the continuous advancement of science and technology, equipment in various fields is evolving towards greater intelligence and sophistication. However, over long periods of operation, equipment performance deteriorates due to the coupling of multiple physical fields, such as changes in the working environment, operational errors, or material aging, leading to performance degradation. If fault conditions are not addressed, these degradation trends will continue to accumulate and eventually damage the equipment, preventing it from operating properly, causing certain pollution, economic losses, and even unimaginable disasters. Therefore, to avoid the safety risks and huge economic losses caused by equipment failure, fault diagnosis and life expectancy prediction (RUL) of equipment are of great significance.

[0003] In view of this situation, the fault prediction and health management technology system achieves dynamic perception and trend prediction of equipment degradation status by integrating multi-source heterogeneous monitoring data. As equipment complexity increases, monitoring data exhibits characteristics such as strong temporal sequence, strong coupling, nonlinearity, and high dimensionality. Therefore, predictive maintenance technology has received widespread attention. Existing RUL prediction methods can be divided into physical model-based methods, data-driven methods, and methods driven by the fusion of physical models and data. Among them, data-driven methods have been widely used in RUL prediction of complex equipment by establishing a nonlinear mapping relationship between the feature space of monitoring data and the health status of the equipment.

[0004] Among data-driven approaches, recurrent neural network models are widely used for RUL prediction. They iteratively predict time series data to estimate the RUL value of a device. However, while traditional recurrent neural networks are a highly efficient technology capable of achieving accurate RUL predictions, they also face several challenges. Firstly, their predictions are mostly point-based, making it difficult to describe the inherent uncertainty of the degradation process. Secondly, the "black box" nature of purely data-driven models leads to a lack of correlation between the prediction results and the physical failure mechanism, resulting in insufficient interpretability of the prediction results. Summary of the Invention

[0005] The present invention provides a method and system for predicting the remaining useful life of equipment to solve the problem that the existing data-driven RUL prediction method has mostly point prediction results, which makes it difficult to describe the uncertainty of the degradation process itself, and the "black box" characteristics of the pure data-driven model lead to a lack of correlation between the prediction results and the physical failure mechanism, thereby resulting in insufficient interpretability of the prediction results.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In one aspect, the present invention provides a method for predicting the remaining useful life of equipment, comprising:

[0008] Obtain equipment degradation data throughout its life cycle as a data set;

[0009] Use the autoencoder to reduce the dimension of the data in the dataset and obtain the health index HI of the equipment;

[0010] The HI is input into the fused RUL prediction network model to predict the probability density curve of the equipment's remaining useful life (RUL) to achieve RUL prediction; wherein the model includes a probability prediction network, a Wiener process model and a fusion module; the probability prediction network is used to obtain a probability density curve in numerical form based on the HI; the Wiener process model is used to obtain a probability density curve in analytical form based on the HI; the fusion module is used to dynamically weight the probability density curves in numerical form and analytical form to obtain a fused curve to achieve RUL prediction.

[0011] Furthermore, the loss function L of the autoencoder is HI Expressed as:

[0012] L HI =mL R +nL D

[0013] Among them, L R is the reconstruction loss function; m is the reconstruction loss function L R The corresponding weight; L D is the degradation trend loss function; n is the degradation trend loss function L D The corresponding weight;

[0014] Reconstruction loss function L R The expression is:

[0015]

[0016] Among them, y i is the original input data; is the reconstructed data; N is the total number of data points;

[0017] Degenerate trend loss function L D The calculation method is:

[0018] Construct differential loss function L diff :

[0019] L diff =δσ 2 -γμ

[0020]

[0021] Among them, σ 2 is the difference variance of HI; δ is the weight of the difference variance; μ is the difference mean of HI; γ is the weight of the difference mean; l i is the number of HI values of the i-th device; is the HI value of the i-th device at time t; is the HI value of the i-th device at time t+1;

[0022] Construct the starting and ending point deviation loss function L var :

[0023] L var =αs 2 +βξ 2

[0024]

[0025] Among them, s 2 is the mean square error between the threshold of the starting point of the equipment degradation process and the starting point of the life span; α is s 2 The corresponding weight; ξ 2 is the mean square error between the thresholds of the equipment degradation process end point and the life end point; β is ξ 2 The corresponding weight; is the degradation starting point of the i-th device; is the life end point of the i-th device; y1 is the life start threshold; y0 is the life end threshold;

[0026] To L diff and L var The summation gives the degradation trend loss function L D :

[0027] L D =L var +L diff =αs 2 +βξ 2 -γμ+δσ 2

[0028] Furthermore, the probability prediction network includes: an LSTM network layer, a self-attention mechanism layer, a Gaussian layer and a scaled dot product attention mechanism module; wherein, the LSTM network layer is used to process the input HI, and the processing result is input into the self-attention mechanism layer, and the output result of the self-attention mechanism layer is input into the Gaussian layer, and the Gaussian layer is used to output the mean μ and variance σ of HI; the scaled dot product attention mechanism module is used to optimize the mean μ output by the Gaussian layer, and the process is as follows: all degraded window data of the training set and the corresponding true value of HI at the next moment are set as the key vector K and value vector V of the scaled dot product attention mechanism respectively; during testing, the test set is subjected to a scaled dot product attention mechanism calculation with the key vector K and the value vector V to obtain a similarity vector ω, and ω is used to correct the mean μ output by the Gaussian layer to obtain a new mean μ′, thereby outputting a Gaussian distribution containing similarity information.

[0029] Furthermore, the log-likelihood loss function of the probability prediction network is Expressed as:

[0030]

[0031] Among them, z represents the true value of HI at the prediction time; σ represents the variance at the prediction time.

[0032] Furthermore, the process of obtaining a probability density curve in numerical form based on HI using the probability prediction network includes:

[0033] Based on the Gaussian distribution generated by the probability prediction network, multiple sets of degradation trajectories are generated through moment-by-moment sampling, degradation sequence recording, and recursive iteration, dynamically recording all degradation trajectories that reach the failure threshold. The known HI degradation sequence is used as input, and the μ′ and σ generated by the probability prediction network are used to construct the Gaussian distribution at the first prediction moment. M samples are sampled from it, and each sample is concatenated with the historical degradation sequence to form a new degradation sequence. The new degradation sequence is input into the probability prediction network to generate the corresponding Gaussian distribution. One sample is sampled from each of the M Gaussian distributions to form a new sample set. At subsequent prediction moments, random sampling is repeated M times from the sample set of the previous moment to form a new degradation sequence for recursive prediction. During the recursive prediction process, the degradation sequence of each sample needs to be recorded to ensure the temporal order between samples. When the degradation state of a sample reaches the end-of-life threshold for the first time, the recursive process of the degradation sequence to which the sample belongs is stopped until all samples degrade to the end-of-life threshold. Wherein, M is a preset positive integer value.

[0034] The degradation state range of the sampling points at each moment is calculated by quantile statistics, and the distribution of HI is mapped to the probability density function of RUL in combination with the threshold, so as to obtain the probability density curve in numerical form.

[0035] Furthermore, the process of obtaining the analytical probability density curve based on HI using the Wiener process model includes:

[0036] Based on the equipment monitoring data, the profile likelihood function optimization is used in combination with the two-dimensional grid search method to determine the parameters of the Wiener process model of the power function drift coefficient, complete the degradation modeling, and obtain the Wiener process model.

[0037] By utilizing the characteristics of the constructed Wiener process model and the first arrival time theory of random processes, the probability density function expression of RUL is output, thereby obtaining the probability density curve in analytical form.

[0038] Furthermore, the process of dynamically weighting the numerical and analytical probability density curves using the fusion module includes:

[0039] Calculate weight coefficient

[0040]

[0041] in, Represents the peak point of the probability density curve f1(r) in numerical form; Represents the peak point of the probability density curve f2(r) in analytical form; represents the RUL point estimate predicted by the fusion module;

[0042] when When the fusion curve The calculation formula is:

[0043] when When only f2(r) is used as

[0044] when When only f1(r) is used as

[0045] On the other hand, the present invention also provides a system for predicting the remaining useful life of equipment, comprising:

[0046] An information acquisition module is used to obtain the degradation data of the equipment throughout its life cycle as a data set;

[0047] A feature extraction module is used to reduce the dimension of the data in the data set acquired by the information acquisition module using an automatic encoder to obtain a health index HI of the device;

[0048] The RUL prediction module is used to input the HI extracted by the feature extraction module into the fused RUL prediction network model to predict the probability density curve of the remaining service life RUL of the equipment to achieve RUL prediction; wherein, the model includes a probability prediction network, a Wiener process model and a fusion module; the probability prediction network is used to obtain a probability density curve in numerical form based on HI; the Wiener process model is used to obtain a probability density curve in analytical form based on HI; the fusion module is used to dynamically weight the probability density curves in numerical form and analytical form to obtain a fused curve to achieve RUL prediction.

[0049] On the other hand, the present invention further provides an electronic device, comprising a processor and a memory; wherein the memory stores at least one instruction, and the instruction is loaded and executed by the processor to implement the above method.

[0050] In yet another aspect, the present invention further provides a computer-readable storage medium, wherein the storage medium stores at least one instruction, and the instruction is loaded and executed by a processor to implement the above method.

[0051] The beneficial effects brought about by the technical solution provided by the present invention include at least:

[0052] 1. This invention uses a multi-factor loss function to guide an autoencoder (AE) to extract features from sensor monitoring data and construct a health index (HI). This HI is then input into a constructed probability prediction network. Combined with a recursive sampling prediction strategy, this method predicts the probability density curve of the RUL. This overcomes the problem that existing RUL prediction methods often produce point predictions, which struggle to describe the inherent uncertainty of the degradation process.

[0053] 2. To address the problem of insufficient interpretability of prediction results due to the "black box" characteristics of pure data-driven models, the present invention uses the Wiener process to predict the RUL probability density function expression, constructs a fusion module to achieve dynamic weighted fusion of the probability prediction network and the probability density curve of the Wiener process, and improves the interpretability of the prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0055] Figure 1 is a flow chart of a method for predicting the remaining useful life of equipment provided by an embodiment of the present invention;

[0056] Figure 2 1 is a schematic diagram of a network model of an HI extraction network AE provided in an embodiment of the present invention;

[0057] Figure 3 Schematic diagram of the network architecture of the probability prediction network AG-LSTM provided by an embodiment of the present invention;

[0058] Figure 4 is a schematic diagram of a recursive sampling prediction strategy provided by an embodiment of the present invention;

[0059] Figure 5 is a schematic diagram of the structure of the fusion module provided in an embodiment of the present invention;

[0060] Figure 6 This is a structural block diagram of a system for predicting the remaining useful life of equipment provided by an embodiment of the present invention;

[0061] Figure 7 It is a structural block diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0062] To make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0063] First, it should be noted that in the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as an "example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "exemplarily" is intended to present concepts in a concrete manner. In addition, in the embodiments of the present invention, the meaning of "and / or" can be both or either of the two.

[0064] First embodiment

[0065] This embodiment provides a method for predicting the remaining useful life of a device. The method can be implemented by an electronic device, which can be a terminal or a server. The execution process of the method is as follows: Figure 1 As shown, the following steps are included:

[0066] S1, obtain the equipment life cycle degradation data as a dataset;

[0067] It should be noted that this method is applicable to the prediction of the remaining useful life of equipment such as aircraft engines and rolling bearings. In actual application, the degradation data of the equipment throughout its life cycle is determined according to demand. For example, for aircraft engines, the degradation data of the equipment throughout its life cycle mainly include fan blade temperature, engine pressure ratio, high-pressure shaft converted speed, etc.

[0068] S2, uses the autoencoder to reduce the dimension of the data in the dataset and obtain the health index HI of the equipment;

[0069] It should be noted that the HI extraction network AE structure is as follows Figure 2 As shown; In order to use the autoencoder to reduce the dimension of the sensor data in the dataset, this embodiment designs a loss function that comprehensively considers factors such as the fidelity of the original degraded data, the linearity and monotonicity of the degradation, and the consistency of the failure threshold to guide the construction of the health index HI; Specifically, the loss function of the HI extraction network AE is shown in the following formula (1):

[0070] L HI =mL R +nL D (1)

[0071] Among them, L R is the reconstruction loss function; m is the reconstruction loss function L R The corresponding weight; L D is the degradation trend loss function; n is the degradation trend loss function L D Corresponding weights; Specifically, in this embodiment, m=0.5, n=0.5 is set, that is, the reconstruction loss function L R and the degradation trend loss function L D Assign equal weights.

[0072] Reconstruction loss function L R The mean square error (MSE) loss is used, and its expression is shown in the following formula (2):

[0073]

[0074] Among them, y i is the original input data; is the data reconstructed by AE; N is the total number of data points;

[0075] Degenerate trend loss function L D The calculation method is:

[0076] In order to ensure the linearity and monotonicity of HI degradation, the differential loss function L is constructed diff :

[0077] L diff =δσ 2 -γμ (3)

[0078]

[0079] Among them, σ 2 is the difference variance of HI; δ is the weight of the difference variance; μ is the difference mean of HI; γ is the weight of the difference mean; li The current device HI value; is the HI value of the i-th device at time t; is the HI value of the i-th device at time t+1;

[0080] In order to ensure the consistency of HI failure threshold, the starting and ending point deviation loss function L is constructed var :

[0081] L var =αs 2 +βξ 2 (6)

[0082]

[0083] Among them, s 2 is the mean square error between the threshold of the starting point of the equipment degradation process and the starting point of the life span; α is s 2 The corresponding weight; ξ 2 is the mean square error between the thresholds of the equipment degradation process end point and the life end point; β is ξ 2 The corresponding weight; is the degradation starting point of the i-th device; is the life end point of the i-th device; y1 is the life start threshold, set to 1; y0 is the life end threshold, set to 0;

[0084] L diff and L var The summation gives the degradation trend loss function L D :

[0085] L D =L var +L diff =αs 2 +βξ 2 -γμ+δσ 2 (9)

[0086] The weight values set in this embodiment are: α=0.3, β=0.3, γ=0.2, δ=0.2.

[0087] S3, inputting the HI into a fusion RUL prediction network model to predict the probability density curve of the equipment's remaining useful life RUL, thereby achieving RUL prediction; wherein the model includes a probability prediction network, a Wiener process model, and a fusion module; the probability prediction network is used to obtain a probability density curve in numerical form based on the HI; the Wiener process model is used to obtain a probability density curve in analytical form based on the HI; the fusion module is used to dynamically weight the probability density curves in numerical form and analytical form to obtain a fused curve, thereby achieving RUL prediction;

[0088] It should be noted that after obtaining the HI, this embodiment is based on the constructed fusion RUL prediction network model, combined with an innovative recursive sampling prediction strategy, to predict the probability density curve of the RUL.

[0089] The probability prediction network AG-LSTM is based on the LSTM long short-term memory recurrent neural network, combined with the Gaussian layer whose output is the mean and variance, the self-attention mechanism and the scaled dot product attention mechanism. Its architecture is as follows Figure 3 As shown in the figure, it includes a G-LSTM probability prediction module and a scaled dot product attention mechanism. The G-LSTM probability prediction module includes an LSTM network, a self-attention mechanism layer, and a Gaussian layer. By introducing the scaled dot product attention mechanism into the G-LSTM probability prediction module, similarity analysis of degraded trajectories is implemented, and a dual-branch AG-LSTM network is constructed. The data processing process of the AG-LSTM network is as follows:

[0090] The input HI is processed by the LSTM network layer, and the processing result is input into the self-attention mechanism layer. The output result of the self-attention mechanism layer is input into the Gaussian layer, and the Gaussian layer is used to output the mean μ and variance σ of HI; the scaled dot product attention mechanism module is used to optimize the mean output of the Gaussian layer. The process is as follows: all degradation window data of the training set and the corresponding true value of HI at the next moment are set as the key vector K and value vector V of the scaled dot product attention mechanism respectively; during testing, the test set is subjected to the scaled dot product attention mechanism calculation with the key vector K and the value vector V to obtain the similarity vector ω, and the vector ω and μ are spliced. The spliced result is input into the fully connected layer, and the fully connected layer outputs a new mean μ′ containing the degradation trajectory similarity information to correct μ, thereby constructing a Gaussian distribution.

[0091] Furthermore, the loss function of the probability prediction network is Expressed as:

[0092]

[0093] Where z represents the true value of HI at the prediction moment; μ′ represents the mean of the prediction output by AG-LSTM, and σ represents the variance of the prediction output by AG-LSTM.

[0094] Based on the above, this embodiment is based on the constructed probability prediction network, combined with Gaussian distribution sampling and recursive prediction strategy, and designs a recursive sampling prediction strategy to realize the probability density curve prediction of RUL; wherein, the implementation process of the recursive sampling prediction strategy is as follows Figure 4 Shown, including:

[0095] Step 1: Based on the Gaussian distribution generated by the AG-LSTM, multiple degradation trajectories are generated through moment-by-moment sampling, degradation sequence recording, and recursive iteration. All degradation trajectories that reach the failure threshold are dynamically recorded. The known HI degradation sequence is used as input. The μ′ and σ generated by the AG-LSTM are used to construct a Gaussian distribution at the first prediction moment. 5,000 samples are sampled from this Gaussian distribution. Each sample is concatenated with the historical degradation sequence to form a new degradation sequence. This new degradation sequence is then input into the AG-LSTM to generate the corresponding Gaussian distribution. One sample is sampled from each of the 5,000 Gaussian distributions to form a new sample set. At subsequent prediction moments, random sampling is repeated 5,000 times from the sample set at the previous moment to form a new degradation sequence for recursive prediction. During the recursive prediction process, the degradation sequence of each sample is recorded to ensure temporal consistency between samples. When the degradation state of a sample reaches the end-of-life threshold for the first time, the recursive process for that sample's degradation sequence is terminated. This process continues until all samples have degraded to the end-of-life threshold.

[0096] Step 2: Use quantile statistics to calculate the degradation state range of the sampling points at each moment, and combine the threshold to map the distribution of HI to the probability density function of RUL. In order to show the degradation probability range, quantiles are used to give the confidence interval of the HI degradation process. The quantile formulas are shown in the following equations (11) to (14):

[0097] L p =p·(N+1) (11)

[0098]

[0099] d=L p -k (13)

[0100] Q p =(1-d)·X k +d·X k+1 (14)

[0101] Among them, L p is the position of the pth quantile after sorting, p is the proportion corresponding to the quantile, N is the total number of samples, and k is L p The integer part of the position is rounded down, d is the decimal part of the position, Q p is the pth quantile, X k and X k+1 The kth and k+1th HI values are sorted during the degradation process. The lower limit d1 of the 90% confidence interval for engine RUL degradation is set to 0.05 and the upper limit d2 is set to 0.95, and the median of the predicted RUL is set to 0.5.

[0102] Step 3: For different monitoring time points of the equipment, the recursive sampling prediction strategy is independently executed to realize the prediction of the RUL probability density curve at any time.

[0103] Furthermore, the RUL probability density curve is obtained by using the Wiener process model: combining the profile likelihood function parameter estimation method and the two-dimensional search method to construct a RUL prediction method based on the Wiener process model to obtain the RUL probability density curve; the specific implementation process is as follows:

[0104] Step 1: Based on the equipment monitoring data, the profile likelihood function optimization combined with the two-dimensional grid search method is used to determine the parameters of the Wiener process model of the power function drift coefficient and complete the degradation modeling;

[0105] Step 2: Using the characteristics of the constructed Wiener process model and the first arrival time theory of random processes, the probability density function expression of RUL is output.

[0106] Furthermore, the fusion module Figure 5 As shown in the figure, it consists of a fully connected layer network, with the loss function using MSE. The input is the peak points of the probability density curves f1(r) and f2(r), and the output is the point estimate of RUL. The fusion module consists of three parts: input data, degradation model, and probability density curve fusion model. It aims to fuse the AG-LSTM and Wiener process to generate the probability density curve of RUL, so that the fused probability density curve has a certain degree of interpretability. Specifically, the process of dynamically weighted fusion of probability density curves using this fusion module is as follows:

[0107] Step 1: Calculate the weight coefficient

[0108]

[0109] in, Represents the peak point of the probability density curve f1(r) in numerical form; Represents the peak point of the probability density curve f2(r) in analytical form; represents the RUL point estimate predicted by the fusion module;

[0110] Step 2: Use the obtained weight coefficient The fusion probability density curve is obtained by weighting f1(r) and f2(r); the fusion methods are divided into the following cases:

[0111] when When f1(r) and f2(r) are given different weights to obtain The formula is as follows:

[0112]

[0113] when When only f2(r) is used as

[0114] when When only f1(r) is used as

[0115] In summary, this embodiment provides a method for predicting the remaining useful life of equipment, which uses a multi-factor loss function to guide the autoencoder AE to extract the features of sensor monitoring data to construct a health index HI, inputs HI into the constructed probability prediction network, and combines the recursive sampling prediction strategy to predict the probability density curve of RUL. It overcomes the problem that the prediction results of existing remaining useful life prediction methods are mostly in the form of point predictions, which makes it difficult to describe the uncertainty of the degradation process itself. In order to address the problem that the "black box" characteristics of pure data-driven models lead to insufficient interpretability of prediction results, the Wiener process is used to predict the RUL probability density function expression, and the probability density curves of the probability prediction network and the Wiener process are dynamically weighted and fused to improve the interpretability of the prediction results.

[0116] Second embodiment

[0117] This embodiment provides a system for predicting the remaining useful life of an equipment. The structure of the system for predicting the remaining useful life of an equipment is as follows: Figure 6 As shown, it includes the following modules:

[0118] An information acquisition module is used to obtain the degradation data of the equipment throughout its life cycle as a data set;

[0119] A feature extraction module is used to use an automatic encoder to reduce the dimension of the data in the data set obtained by the information acquisition module to obtain health indicator data HI of the device;

[0120] The RUL prediction module is used to input the HI extracted by the feature extraction module into the fused RUL prediction network model to predict the probability density curve of the remaining service life RUL of the equipment to achieve RUL prediction; wherein, the model includes a probability prediction network, a Wiener process model and a fusion module; the probability prediction network is used to obtain a probability density curve in numerical form based on HI; the Wiener process model is used to obtain a probability density curve in analytical form based on HI; the fusion module is used to dynamically weight the probability density curves in numerical form and analytical form to obtain a fused curve to achieve RUL prediction.

[0121] It should be noted that the equipment remaining useful life prediction system of this embodiment corresponds to the equipment remaining useful life prediction method of the above-mentioned first embodiment; the functions implemented by each functional module in the equipment remaining useful life prediction system of this embodiment correspond one-to-one to each process step in the equipment remaining useful life prediction method of the above-mentioned first embodiment; therefore, they will not be repeated here.

[0122] Third embodiment

[0123] This embodiment provides an electronic device, such as Figure 7 As shown, the electronic device includes: a processor and a memory; wherein the processor and the memory can be connected via a communication bus; the memory stores at least one instruction, which is loaded and executed by the processor to implement the method of the first embodiment described above. In addition, the electronic device may also include a transceiver; the processor and the transceiver can be connected via a communication bus, and the transceiver is used to communicate with other devices.

[0124] Next, combine Figure 7 A detailed introduction to the various components of the electronic device is given below:

[0125] Among them, the processor is the control center of the electronic device, and the electronic device may include multiple processors, each of which may be a single-core processor (single-CPU) or a multi-core processor (multi-CPU). The processor here can be a processor or a general term for multiple processing elements. For example, the processor is one or more central processing units (CPUs), or other general-purpose processors, application specific integrated circuits (ASICs), or one or more integrated circuits configured to implement an embodiment of the present invention, such as one or more microprocessors (digital signal processors, DSPs), or one or more field programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor can perform various functions of the electronic device by running or executing software programs stored in the memory and calling data stored in the memory.

[0126] In a specific implementation, as an embodiment, the processor may include one or more CPUs, such as Figure 7The CPU0 and CPU1 shown in FIG are, of course, only exemplary.

[0127] The memory is used to store the software program for executing the solution of the present invention, and the execution is controlled by the processor. The specific implementation method can refer to the above method embodiment and will not be repeated here.

[0128] Optionally, the memory may be a read-only memory (ROM) or other type of static storage device that can store static information and instructions, a random access memory (RAM) or other type of dynamic storage device that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compact disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory may be integrated with the processor or exist independently and accessed through the interface circuit ( Figure 7 (not shown) is coupled to the processor, which is not specifically limited in this embodiment of the present invention.

[0129] The transceiver may include a receiver and a transmitter ( Figure 7 The receiver is used to implement the receiving function, and the transmitter is used to implement the sending function. The transceiver can be integrated with the processor or exist independently and communicate with the electronic device through the interface circuit ( Figure 7 (not shown) is coupled to the processor, which is not specifically limited in this embodiment of the present invention.

[0130] In addition, it should be noted that Figure 7 The structure of the electronic device shown in the figure does not constitute a limitation on the device. The actual device may include more or fewer components than shown, or may combine certain components, or arrange the components differently. In addition, the technical effects achieved by the electronic device when executing the method of the first embodiment can refer to the technical effects described in the first embodiment above, and therefore will not be repeated here.

[0131] Fourth embodiment

[0132] This embodiment provides a computer-readable storage medium storing at least one instruction, which is loaded and executed by a processor to implement the method of the first embodiment described above. The computer-readable storage medium may be a ROM, random access memory, CD-ROM, magnetic tape, floppy disk, or optical data storage device. The instructions stored therein can be loaded by a processor in a terminal to execute the method described above.

[0133] Furthermore, it should be noted that the present invention may be provided as a method, apparatus, or computer program product. Therefore, embodiments of the present invention may take the form of a fully or partially hardware embodiment, a fully or partially software embodiment, or an embodiment combining software and hardware aspects. Furthermore, when implemented using software, embodiments of the present invention may take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The computer program product comprises one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, the processes or functions described in accordance with the embodiments of the present invention are fully or partially generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired connection (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium may be any available medium accessible by a computer or a data storage device such as a server or data center that contains a collection of one or more available media. The available medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a DVD), or a semiconductor medium. The semiconductor medium may be a solid state drive.

[0134] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the process in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0135] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device, so that a series of operation steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for implementing the process in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0136] It should also be noted that, in this document, relational terms such as first and second are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any actual relationship or order between these entities or operations. The terms "include," "comprises," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or terminal device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or terminal device. Without further limitation, an element defined by the phrase "comprising a..." does not preclude the presence of other identical elements in the process, method, article, or terminal device comprising the element. In addition, the term "and / or" is merely a description of an associative relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent: the presence of A alone, the presence of A and B simultaneously, or the presence of B alone, where A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding. "At least one" means one or more, and "more" means two or more. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can be represented by: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or multiple.

[0137] In addition, it can be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0138] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.

[0139] In the several embodiments provided herein, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of functional modules / units is merely a logical functional division. In actual implementation, other division methods may be used, such as multiple units or components being combined or integrated into another device, or some features being ignored or not implemented. Furthermore, the coupling or direct coupling or communication connection shown or discussed between each other may be through some interface, indirect coupling or communication connection between devices or units, and may be electrical, mechanical, or other forms. Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, i.e., they may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of the present embodiment according to actual needs. In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit.

[0140] If the method is implemented in the form of 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 the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0141] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention. It should be noted that, although preferred embodiments of the present invention have been described, those skilled in the art, once understanding the basic inventive concepts of the present invention, may make various improvements and modifications without departing from the principles of the present invention. Such improvements and modifications should also be considered within the scope of protection of the present invention. Therefore, the appended claims are intended to be interpreted as covering the preferred embodiments and all variations and modifications that fall within the scope of the embodiments of the present invention.

Claims

1. A method for predicting the remaining useful life of equipment, characterized in that: include: Obtain equipment degradation data throughout its life cycle as a data set; Use the autoencoder to reduce the dimension of the data in the dataset and obtain the health index HI of the equipment; The HI is input into the fused RUL prediction network model to predict the probability density curve of the equipment's remaining service life (RUL) to achieve RUL prediction; wherein the model includes a probability prediction network, a Wiener process model and a fusion module; the probability prediction network is used to obtain a probability density curve in numerical form based on the HI; the Wiener process model is used to obtain a probability density curve in analytical form based on the HI; the fusion module is used to dynamically weight the probability density curve in numerical form and the probability density curve in analytical form to obtain a fused curve to achieve RUL prediction.

2. The method for predicting the remaining useful life of equipment according to claim 1, wherein: The loss function L of the autoencoder HI Expressed as: L HI =mL R +nL D Among them, L R is the reconstruction loss function; m is the reconstruction loss function L R The corresponding weight; L D is the degradation trend loss function; n is the degradation trend loss function L D The corresponding weight; Reconstruction loss function L R The expression is: Among them, y i is the original input data; is the reconstructed data; N is the total number of data points; Degenerate trend loss function L D The calculation method is: Construct differential loss function L diff : L diff =ds 2 -gm Among them, σ 2 is the difference variance of HI; δ is the weight of the difference variance; μ is the difference mean of HI; γ is the weight of the difference mean; l i is the number of HI values of the i-th device; is the HI value of the i-th device at time t; is the HI value of the i-th device at time t+1; Construct the starting and ending point deviation loss function L var : L var =αs 2 +vx 2 Among them, s 2 is the mean square error between the threshold of the starting point of the equipment degradation process and the starting point of the life span; α is s 2 The corresponding weight; ξ 2 is the mean square error between the thresholds of the equipment degradation process end point and the life end point; β is ξ 2 The corresponding weight; is the degradation starting point of the i-th device; is the life end point of the i-th device; y1 is the life start threshold; y0 is the life end threshold; To L diff and L var The summation gives the degradation trend loss function L D .

3. The method for predicting the remaining useful life of equipment according to claim 1, wherein: The probability prediction network includes: an LSTM network layer, a self-attention mechanism layer, a Gaussian layer and a scaled dot product attention mechanism module; wherein the LSTM network layer is used to process the input HI, and the processing result is input into the self-attention mechanism layer, and the output result of the self-attention mechanism layer is input into the Gaussian layer, and the Gaussian layer is used to output the mean μ and variance σ of HI; the scaled dot product attention mechanism module is used to optimize the mean μ output by the Gaussian layer, and the process is as follows: all degraded window data of the training set and the corresponding true value of HI at the next moment are set as the key vector K and value vector V of the scaled dot product attention mechanism respectively; during testing, the test set is subjected to a scaled dot product attention mechanism calculation with the key vector K and the value vector V to obtain a similarity vector ω, and ω is used to correct the mean μ output by the Gaussian layer to obtain a new mean μ′, thereby outputting a Gaussian distribution containing similarity information.

4. The method for predicting the remaining useful life of equipment according to claim 3, wherein: The log-likelihood loss function of the probability prediction network Expressed as: Among them, z represents the true value of HI at the prediction time; σ represents the variance at the prediction time.

5. The method for predicting the remaining useful life of equipment according to claim 4, wherein: The process of obtaining a probability density curve in numerical form based on HI using the probability prediction network includes: Based on the Gaussian distribution generated by the probability prediction network, multiple sets of degradation trajectories are generated through moment-by-moment sampling, degradation sequence recording, and recursive iteration, dynamically recording all degradation trajectories that reach the failure threshold. The known HI degradation sequence is used as input, and the μ′ and σ generated by the probability prediction network are used to construct the Gaussian distribution at the first prediction moment. M samples are sampled from it, and each sample is concatenated with the historical degradation sequence to form a new degradation sequence. The new degradation sequence is input into the probability prediction network to generate the corresponding Gaussian distribution. One sample is sampled from each of the M Gaussian distributions to form a new sample set. At subsequent prediction moments, random sampling is repeated M times from the sample set of the previous moment to form a new degradation sequence for recursive prediction. During the recursive prediction process, the degradation sequence of each sample needs to be recorded to ensure the temporal order between samples. When the degradation state of a sample reaches the end-of-life threshold for the first time, the recursive process of the degradation sequence to which the sample belongs is stopped until all samples degrade to the end-of-life threshold. Wherein, M is a preset positive integer value. The degradation state range of the sampling points at each moment is calculated by quantile statistics, and the distribution of HI is mapped to the probability density function of RUL in combination with the threshold, so as to obtain the probability density curve in numerical form.

6. The method for predicting the remaining useful life of equipment according to claim 1, wherein: The process of obtaining the analytical probability density curve based on HI using the Wiener process model includes: Based on the equipment monitoring data, the profile likelihood function optimization is used in combination with the two-dimensional grid search method to determine the parameters of the Wiener process model of the power function drift coefficient, complete the degradation modeling, and obtain the Wiener process model. By utilizing the characteristics of the constructed Wiener process model and the first arrival time theory of random processes, the probability density function expression of RUL is output, thereby obtaining the probability density curve in analytical form.

7. The method for predicting the remaining useful life of equipment according to claim 1, wherein: The process of dynamically weighting the numerical probability density curve and the analytical probability density curve using the fusion module includes: Calculate weight coefficient in, Represents the peak point of the probability density curve f1(r) in numerical form; Represents the peak point of the probability density curve f2(r) in analytical form; represents the RUL point estimate predicted by the fusion module; when When the fusion curve The calculation formula is: when When only f2(r) is used as when When only f1(r) is used as 8. A system for predicting the remaining useful life of equipment, characterized in that: include: An information acquisition module is used to obtain the degradation data of the equipment throughout its life cycle as a data set; A feature extraction module is used to reduce the dimension of the data in the data set acquired by the information acquisition module using an automatic encoder to obtain a health index HI of the device; A RUL prediction module is used to input the HI extracted by the feature extraction module into a fusion RUL prediction network model to predict the probability density curve of the remaining service life RUL of the equipment to achieve RUL prediction; Among them, the model includes a probability prediction network, a Wiener process model and a fusion module; the probability prediction network is used to obtain a probability density curve in numerical form based on HI; the Wiener process model is used to obtain a probability density curve in analytical form based on HI; the fusion module is used to dynamically weight the probability density curve in numerical form and the probability density curve in analytical form to obtain a fusion curve to achieve RUL prediction.

Images