A device residual life prediction method based on broadcast attention mechanism

Abstract

The application discloses a device residual life prediction method based on a broadcast attention mechanism, and belongs to the technical field of device residual life prediction, and comprises the following steps: collecting device operation state data; pre-processing the collected device operation state data; training a preset device residual life prediction model by using the pre-processed device operation state data; wherein the device residual life prediction model comprises a broadcast attention mechanism; and realizing device residual life prediction by using the trained device residual life prediction model. The application can improve the accuracy, calculation efficiency and robustness of device residual life prediction, thereby better supporting intelligent operation and maintenance and predictive maintenance.

Description

Technical Field

[0001] This invention relates to the field of equipment remaining life prediction technology, and in particular to a method for predicting equipment remaining life based on a broadcast attention mechanism. Background Technology

[0002] Throughout its lifespan, equipment typically undergoes a progression from a normal state to latent degradation, then to noticeable degradation, and finally, a gradual escalation to unusable or failed conditions. By categorizing and sorting these different states, a system performance degradation state sequence can be created. Different response plans can then be implemented based on each degradation state, reducing costs while improving stability. Effectively assessing the specific state of the equipment allows for tailored responses; for example, completely normal equipment requires no maintenance, latent degradation necessitates frequent checks, and noticeable degradation requires overhauling. Since these states occur before failure, degradation assessment can effectively prevent unexpected equipment malfunctions, and the absence of serious failures reduces maintenance costs such as major overhauls.

[0003] Data-driven remaining lifetime prediction methods require analyzing the underlying model of system operational data. Based on this model, the remaining lifetime can be accurately predicted. The reliability of this method is inseparable from a large amount of historical operational data; historical data must be acquired and used as a specific sample to repeatedly adjust the accuracy of the prediction model. Following a data-driven approach, this lifetime prediction method can be divided into two common types: parametric and non-parametric methods. Parametric methods require assuming a certain distribution, then fitting historical data to obtain the parameters of the distribution, and using these parameters to predict future lifetime. Non-parametric methods, on the other hand, directly analyze historical data, without needing to assume a distribution beforehand. Through statistical analysis of historical data, a prediction model is built to predict future lifetime.

[0004] To predict the remaining lifespan of servo system equipment, it is typically necessary to utilize numerous sensors to collect signal parameters such as temperature, vibration, current, and voltage. These signal quantities are time-series data with time dependencies. Currently, Long Short-Term Memory (LSTM) networks are widely used to capture long-term dependencies in time series data, achieving great success in time series data analysis. For example, integrating 3D convolutions into LSTMs and using CNN-based modules to decouple physical dynamics. However, despite the widespread use of convolutional methods, their inherent limitations in capturing spatiotemporal dependencies due to their locality reduce their efficiency. Furthermore, issues such as imbalanced feature extraction, feature loss, and low learning rates persist.

[0005] In summary, existing methods for predicting the remaining life of equipment have the following main drawbacks:

[0006] Traditional methods based on statistical and physical models typically assume that data follows a specific probability distribution (such as a normal distribution or a Weibull distribution) and rely on these distributions to build predictive models. However, in practical applications, the degradation process of equipment is often affected by a variety of complex factors, including environmental conditions, load variations, and random disturbances. These factors can cause dynamic changes in the data distribution, making it difficult for fixed distribution assumptions to accurately fit the actual lifespan degradation process, thus affecting prediction accuracy.

[0007] Furthermore, existing methods still have certain shortcomings in feature extraction. For example, a single CNN model excels at extracting local features, but due to its limited receptive field, it performs poorly in capturing long-term spatiotemporal dependencies and may overlook some key global features. In addition, traditional feature extraction methods are prone to feature imbalance, where some key information may be weakened or lost, thus affecting the overall performance of the prediction model, resulting in an incomplete representation of device health status and reducing the accuracy of remaining life prediction. Summary of the Invention

[0008] This invention provides a device remaining lifetime prediction method based on a broadcast attention mechanism to solve the technical problem of low prediction accuracy in existing device remaining lifetime prediction methods.

[0009] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0010] On one hand, the present invention provides a device remaining lifetime prediction method based on a broadcast attention mechanism, the device remaining lifetime prediction method based on a broadcast attention mechanism includes:

[0011] Collect equipment operating status data;

[0012] The collected equipment operating status data is preprocessed; wherein, the preprocessing includes: data cleaning and data denoising, data normalization and data augmentation;

[0013] The pre-processed equipment operating status data is used to train a preset equipment remaining life prediction model; wherein, the equipment remaining life prediction model includes a broadcast attention mechanism.

[0014] The remaining life of equipment is predicted using a trained equipment remaining life prediction model.

[0015] Furthermore, the equipment operating status data includes temperature data, vibration data, current data, and voltage data during equipment operation; the data format adopts a multi-channel time series.

[0016] Furthermore, the data cleaning and data denoising include:

[0017] Use the sliding window mean to fill in missing data;

[0018] An outlier data point was removed using a method based on One-Class SVM.

[0019] Wavelet transform filtering is used for data denoising.

[0020] Furthermore, the data enhancement includes:

[0021] Training samples are generated using a fixed-length sliding window; long-term sequences are randomly pruned.

[0022] Furthermore, the device remaining lifetime prediction model includes: a convolutional scalar long short-term memory neural network module, a broadcast attention module, a convolutional difference decoder module, a feature fusion module, and a fully connected layer;

[0023] The preprocessed equipment operating status data is input into a convolutional scalar long short-term memory neural network module to deeply mine the features between sequences and extract the first feature information of the data; the first feature information is input into a broadcast attention module, and the output of the broadcast attention module is input into a convolutional difference decoder module to extract the second feature information of the data; the feature fusion module fuses the first feature information and the second feature information to obtain fused feature information; the fully connected layer realizes the prediction of the remaining life of the equipment based on the fused feature information.

[0024] Furthermore, the data processing procedure for the convolutional scalar long short-term memory neural network module includes:

[0025] First, the preprocessed device operating status data is processed by convolutional operations to generate feature maps; then, one-dimensional max pooling is used to convert each kernel size of the feature maps into a single output with the maximum number of kernels; subsequently, the output of one-dimensional max pooling is used to capture deeper spatiotemporal feature information of the data through an sLSTM network.

[0026] Furthermore, the broadcast attention module consists of a 3×3 convolutional layer and a broadcast attention encoder, with ReLU activation function and Batch Normalization used for training optimization.

[0027] Furthermore, the data processing procedure of the broadcast attention encoder includes:

[0028] First, three broadcast factors are initially set as m=3, n=5, and k=7, corresponding to three strip convolution kernels of different scales. Then, a preset broadcast factor adaptive correction algorithm is used to dynamically adjust the numerical range of m, n, and k, so as to flexibly adjust the multi-scale receptive field when facing different input structures and scene distributions, thereby improving the adaptability and modeling ability of feature extraction.

[0029] The input to the broadcast attention encoder first passes through an n×n convolutional kernel in the backbone branch to capture feature information. The captured feature information is then input into the multimodal broadcast branch, which includes three branches of different scales. Each of the three branches of different scales employs a separable convolutional operation. These branches are concatenated with convolutional kernels of different sizes, and the scales corresponding to each branch are 1×m→m×1, 1×n→n×1, and 1×k→k×1, respectively.

[0030] Features from branches at different scales are summed element-wise at the convergence node to fuse multi-scale information. Then, they are passed through a 1×1 convolutional layer to reduce or transform the channel dimension.

[0031] Skip connections are used for gradient propagation and information reuse. The output of the main branch is directly connected to the final 1×1 convolutional layer. Addition is used to add the output of the main branch to the output of the multimodal broadcast branch.

[0032] Multiplicative connections are used for feature weighting and selection, allowing the input to be multiplied element-wise with the final output, thus implementing a gating mechanism to control the transmission of feature information, enhance important features and suppress irrelevant features.

[0033] Furthermore, the process by which the broadcast factor adaptive correction algorithm dynamically adjusts the numerical ranges of m, n, and k includes:

[0034] Within each training epoch, the average activation intensity and gradient contribution of the branch output features at each scale in the multimodal broadcast branch are calculated.

[0035] The branches at different scales are sorted according to their gradient contributions. The kernel size of the convolution is reduced for branches with responses below a preset threshold, and the kernel size of the branches with responses above a preset threshold is increased, thus forming an adaptive scale evolution.

[0036] Stable convergence is achieved through an exponential smoothing update formula, ensuring continuity and differentiability during the training process.

[0037] The corrected m, n, and k can be used as a reference for the design of the next stage of convolutional kernels, or for the search and optimization stage.

[0038] Furthermore, the convolutional interpolation decoder module includes four convolutional interpolation decoders; each convolutional interpolation decoder includes a bilinear interpolation module and a 3×3 convolution module.

[0039] The data processing steps of the convolutional difference decoder module include:

[0040] The output of the broadcast attention encoder is upsampled by bilinear interpolation using the first convolutional difference decoder, and the upsampled result is fused with the input of the broadcast attention encoder using a 3×3 convolution as the output of the first convolutional difference decoder. After the first convolutional difference decoder, each convolutional difference decoder upsamples the output of the broadcast attention encoder by bilinear interpolation, and the output of the broadcast attention encoder and the output of the previous convolutional difference decoder are fused using a 3×3 convolution. Finally, the features from the last three convolutional difference decoders are fused using pointwise convolution.

[0041] In another aspect, the present invention also provides an electronic device comprising a processor and a memory; wherein the memory stores at least one instruction, which is loaded and executed by the processor to implement the above-described method.

[0042] In another aspect, the present invention also provides a computer-readable storage medium storing at least one instruction, which is loaded and executed by a processor to implement the above method.

[0043] The beneficial effects of the technical solution provided by this invention include at least the following:

[0044] First, this invention employs a convolutional scalar long short-term memory network combined with 1D-CNN for local feature extraction, and improves the long-term dependency modeling capability and prediction accuracy of device lifetime prediction through sLSTM deep modeling of spatiotemporal dependencies. Next, this invention introduces a broadcast factor adaptive correction algorithm, which dynamically adjusts the parameters of convolutional kernels at different scales (e.g., m, n, k) according to feature responses, achieving adaptive matching of the receptive field. This mechanism enhances the flexibility and robustness of the model in multi-scale modeling, making the network more stable and efficient when processing input features of different sizes, structures, and complexities. Second, this invention designs a broadcast attention encoder, utilizing the broadcast factor adaptive correction algorithm, depthwise convolution, multi-scale strip convolution, and 1×1 convolution to construct an efficient attention mechanism. This improves global feature extraction capabilities while achieving linear complexity computation, balancing information integrity and computational efficiency. Combined with a convolutional interpolation decoder for feature reconstruction optimization, this effectively improves feature representation capabilities, ensures prediction stability and robustness, and facilitates practical deployment.

[0045] Through the above improvements, the present invention excels in enhancing the accuracy, computational efficiency, and application adaptability of equipment remaining life prediction, providing efficient and reliable technical support for intelligent operation and maintenance and predictive maintenance. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is a flowchart of the device remaining lifetime prediction method based on broadcast attention mechanism provided in an embodiment of the present invention;

[0048] Figure 2 This is a schematic diagram of a convolutional scalar long short-term memory neural network module provided in an embodiment of the present invention;

[0049] Figure 3 This is a schematic diagram of a broadcast attention encoder provided in an embodiment of the present invention;

[0050] Figure 4 This is a schematic diagram of the broadcast attention module provided in an embodiment of the present invention;

[0051] Figure 5 This is a schematic diagram of the convolutional interpolation decoder module provided in an embodiment of the present invention;

[0052] Figure 6 This is a system block diagram of the electronic device provided in the embodiments of the present invention. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0054] First, it should be noted that in the embodiments of the present invention, the words "exemplarily," "for example," etc., are used to indicate that they are examples, illustrations, or descriptions. Any embodiment or design scheme described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the term "exemplarily" is intended to present the concept in a specific manner. Furthermore, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either one or the other.

[0055] First Embodiment

[0056] To address the shortcomings of existing equipment remaining lifetime prediction methods in terms of spatiotemporal feature extraction, feature fusion, and computational complexity, this embodiment provides an equipment remaining lifetime prediction method based on a broadcast attention mechanism. The aim is to construct an efficient, accurate, and deployable equipment remaining lifetime prediction method to improve the accuracy, computational efficiency, and robustness of equipment remaining lifetime prediction, thereby reducing equipment failure risks, lowering maintenance costs, and enhancing the level of intelligence in equipment management, ultimately better supporting intelligent operation and maintenance and predictive maintenance.

[0057] like Figure 1 As shown in the figure, the implementation process of this embodiment is as follows: First, equipment operating status data is collected through various sensors (temperature, vibration, current, voltage, etc.), and the data is cleaned, denoised, normalized, and augmented to improve data quality and model generalization ability. Second, a CNN-sLSTM module is designed, using 1D-CNN to extract local features and combining it with sLSTM for spatiotemporal feature modeling to improve the learning ability that relies on long-term information. Next, a broadcast factor adaptive correction algorithm is designed. This algorithm dynamically adjusts the numerical range of broadcast factors m, n, and k by analyzing the weight distribution of broadcast branches at different scales on feature responses to achieve adaptive matching of the receptive field. Subsequently, a broadcast attention encoder is introduced, which uses depthwise convolution, multi-scale strip convolution, and 1×1 convolution to achieve multi-scale feature extraction, enhance the fusion of local and global information, and reduce computational complexity. Furthermore, a convolutional interpolation decoder is constructed, using bilinear interpolation and 3×3 convolution for upsampling and feature reconstruction to optimize the model's feature expression ability, improve information integration ability, prevent feature loss, and improve prediction stability. During training, the weighted mean square error (WMSE) is used to calculate the RUL error, and the number of training rounds is dynamically adjusted to obtain the optimal model.

[0058] This method can be implemented by an electronic device. Specifically, the execution process includes the following steps:

[0059] S1, collects equipment operating status data;

[0060] It should be noted that equipment health monitoring data is typically collected by multiple sensors, such as temperature sensors to monitor equipment heat generation, vibration sensors to detect mechanical structural abnormalities, and current / voltage sensors to monitor the electrical status of motors or electronic equipment. The data format is a multi-channel time series, i.e., X = [X1, X2, ..., X...]. t ], Among them, X i The n-dimensional sensor features represent the i-th time step.

[0061] S2, preprocessing the collected equipment operating status data; wherein, the preprocessing includes: data cleaning and data denoising, data normalization and data augmentation;

[0062] Specifically, in this embodiment, the data preprocessing process is as follows:

[0063] S21, Data Cleaning and Denoising: Missing data were filled using the sliding window mean; outlier data points were removed using a One-Class SVM-based method; and denoising was performed using wavelet transform filtering.

[0064] S22, Data Normalization and Standardization: Due to the different dimensions of different sensors, the Z-score standard is adopted.

[0065] The formula is expressed as:

[0066] S23, Data Augmentation and Dataset Generation: To increase the amount of training data and improve model robustness, the following methods are used for data augmentation: generating training samples using a fixed-length sliding window (e.g., 50 or 100 steps); and randomly pruning long-term sequences to improve the model's generalization ability.

[0067] S3, using preprocessed equipment operating status data to train a preset equipment remaining life prediction model; wherein, the equipment remaining life prediction model includes a broadcast attention mechanism.

[0068] S4, use the trained equipment remaining life prediction model to predict the equipment remaining life;

[0069] It should be noted that, to improve the accuracy of equipment remaining life prediction, this embodiment proposes a CNN-sLSTM model composed of CNN and sLSTM to deeply mine features between sequences, reduce data noise, and improve prediction accuracy. Specifically, this model extracts local time series features through 1D-CNN and uses sLSTM to deeply model spatiotemporal dependencies, enhancing the learning ability of long-term dependency information. In addition, this embodiment also designs a broadcast factor adaptive correction algorithm, a broadcast attention encoder, and a convolutional interpolation decoder. It uses multi-scale strip convolution and 1×1 convolution to build an efficient attention mechanism, which improves the global feature extraction capability while reducing computational complexity, achieving linear complexity calculation, and improving real-time performance. Furthermore, it combines the convolutional interpolation decoder for feature reconstruction optimization, strengthens the interaction and reconstruction between different features, avoids information loss, balances information integrity and computational efficiency, and improves the stability and robustness of the model.

[0070] Specifically, in this embodiment, the process of constructing the equipment remaining life prediction model is as follows:

[0071] Step 1, Design a Convolutional Scalar Long Short-Term Memory Neural Network Module (CNN-sLSTM): Use 1D-CNN for time series feature extraction, h t =f(W·X) t +b), where X t Given the input time-step data, W and b are the weights and biases of the CNN, and f(·) is the ReLU activation function. CNN models can effectively reduce the number of features to prevent overfitting, while also reducing time and parameter complexity, such as... Figure 2 As shown, feature maps are first generated by convolutional operations, and then pooling layers extract the most important features to compute local sufficient statistics. One-dimensional max pooling transforms each kernel size of the input into a single output with the maximum number of kernels to reduce or downsample the version of the input, which is then captured by an sLSTM network to capture deeper spatiotemporal feature information of the data.

[0072] Step 2: Design the Broadcast Factor Adaptive Tuning (BFAT) algorithm: During network training, to further improve the model's adaptability to different tasks and input features, a broadcast factor adaptive tuning algorithm is introduced. This algorithm dynamically adjusts the numerical ranges of broadcast factors m, n, and k by analyzing the weight distribution of broadcast branches at different scales on feature responses, thereby improving the matching ability of the receptive field. Specifically:

[0073] 1) Within each training epoch, calculate the average activation intensity and gradient contribution of the output features of each scale broadcast branch;

[0074] 2) Based on the contribution ranking, the kernel size is reduced for branches with lower responses and increased for branches with higher responses, forming an adaptive scale evolution;

[0075] 3) This adjustment process achieves stable convergence through an exponential smoothing update formula, ensuring continuity and differentiability during the training process;

[0076] 4) The corrected m, n, and k will serve as a reference for the next stage of convolutional kernel design, or be further used in the search and optimization stage.

[0077] Step 3, Design the Broadcast Attention Encoder (BAEncoder): To extract multi-scale features and fuse their respective feature domains, a broadcast attention encoder was designed, employing a pyramid structure, such as... Figure 3As shown, it consists of four parts: the broadcast factor adaptive correction algorithm designed in step 2, a depthwise convolution for aggregating local information, a depthwise bar convolution for capturing multi-scale broadcast branches, and a 1×1 convolution for modeling the relationship between different channels. Finally, a dual residual connection is used, and the output of the 1×1 convolution is directly used as the attention weight to reweight the input of BA. Specifically:

[0078] 1) Broadcast Factor Correction: Initially, three broadcast factors are set as m=3, n=5, and k=7, corresponding to three strip convolution kernels of different scales. The broadcast factor adaptive correction algorithm in step 2 is used to dynamically adjust the numerical range of m, n, and k, so as to flexibly adjust the multi-scale receptive field when facing different input structures and scene distributions, thereby improving the adaptability and modeling ability of feature extraction.

[0079] 2) Main branch: The input features first pass through an n×n convolution kernel to capture feature information within a larger receptive field.

[0080] 3) Multimodal broadcasting branch:

[0081] a) Separable convolution operations are used through three branches of different scales (1×m→m×1, 1×n→n×1 and 1×k→k×1) to reduce computational cost and improve feature representation capability.

[0082] b) These branches effectively decompose large-scale convolution kernels by concatenating convolution kernels of different sizes, making computation more efficient while still retaining the characteristic of a large receptive field.

[0083] 4) Feature fusion:

[0084] a) Features from different paths are summed element-wise at the convergence node to fuse multi-scale information.

[0085] b) After that, a 1×1 convolutional layer is passed to perform dimensionality reduction or transformation of the channel dimension.

[0086] 5) Residual connection:

[0087] a) Residual connections are used for gradient propagation and information reuse. The output of the backbone branch is directly connected to the final 1×1 convolution and addition is used. This connection allows the backbone output to be added to the result extracted by the multimodal broadcast branch, thereby enhancing gradient flow and preserving input information, and alleviating the gradient vanishing problem during deep network training.

[0088] b) Employing multiplicative connections, feature weighting and selection are performed, allowing the input to be directly multiplied element-wise with the final output. This implements a gating mechanism to control the transmission of feature information, enhance important features, and suppress irrelevant features.

[0089] Step 4, Construct the Broadcast Attention Module: Using the broadcast attention encoder constructed in Step 3, the broadcast attention module consists of a 3×3 convolutional layer and the broadcast attention encoder, as follows: Figure 4 As shown, the input data is processed sequentially through a 3×3 convolutional layer and a broadcast attention encoder. ReLU is used as the activation function, and Batch Normalization (BN) is used for training optimization.

[0090] Step 5, Design the Convolution Interpolation Decoder (CIDecoder): The CI module consists of two parts: bilinear interpolation of the output features from the encoder in Step 3 to achieve upsampling, and fusing information from the encoder and the previous CI module using 3×3 convolutions. Finally, pointwise convolutions are used to fuse features from the last three CI modules, such as... Figure 5 As shown.

[0091] Step 6, Feature Fusion: Input the feature information obtained from the above steps into the feature fusion module for weighted feature fusion.

[0092] Step 7, Model Training and Testing. Given a training sample, data preprocessing is performed by S2. Each sample data is in multi-channel time series format. The preprocessed data is first input into the CNN-sLSTM module built in Step 1 for preliminary local feature extraction and long-term modeling to extract spatiotemporal features. Then, the output features are input into the broadcast attention module built in Step 4 to extract multi-scale features and fuse their respective feature domains. The output is then input into the convolutional feature decoder built in Step 5 for feature reconstruction optimization and integration of multi-scale feature information. Next, the spatiotemporal features obtained from the CNN-sLSTM module and the feature information obtained from the broadcast attention module and the convolutional interpolation decoder are input into the feature fusion module for weighted feature fusion. Finally, the weighted mean squared error (WMSE) is used to calculate the RUL error, as shown in the following formula, and the model is tested on the test set. The training epochs are dynamically adjusted according to the stability of the error.

[0093]

[0094] In summary, this embodiment provides a device remaining lifetime prediction method based on a broadcast attention mechanism. Addressing the problem that traditional time series modeling methods struggle to fully capture long-term and short-term dependencies, resulting in limited prediction accuracy, this embodiment designs a Conv-sLSTM model composed of CNN and sLSTM. It combines 1D-CNN for local feature extraction and sLSTM for deep spatiotemporal feature learning, effectively enhancing the modeling capability of long-term dependency information. Furthermore, addressing the insufficient receptive field adaptability of fixed-scale convolutions when processing diverse inputs, this embodiment proposes a broadcast factor adaptive correction algorithm, which dynamically adjusts the convolution kernel size based on the feature responses and gradient contributions of different scale broadcast branches during training. This strategy enables the model to flexibly match the receptive field range, thereby more effectively modeling multi-scale feature information. To address the issues of information loss and high computational complexity caused by downsampling in traditional feature extraction algorithms, this embodiment proposes a broadcast attention encoder. It utilizes a broadcast factor adaptive correction algorithm, depthwise convolution, multi-scale strip convolution, and 1×1 convolution to construct an attention mechanism, achieving efficient feature extraction while balancing information preservation and computational efficiency. Furthermore, it optimizes information reconstruction through a convolutional interpolation decoder, improving feature representation capabilities and ensuring the accuracy and stability of prediction.

[0095] Based on the above, this method achieves efficient, accurate, and robust prediction of equipment remaining life through multi-level feature extraction, fusion, and optimization mechanisms. It effectively improves the accuracy, computational efficiency, and application robustness of equipment life prediction, providing efficient and reliable technical support for intelligent operation and maintenance and predictive maintenance of industrial equipment.

[0096] Second Embodiment

[0097] This embodiment provides an electronic device, such as... Figure 6 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. Furthermore, 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.

[0098] Below, in conjunction with Figure 6 A detailed introduction to each component of this electronic device is provided below:

[0099] The processor is the control center of the electronic device. The electronic device may include multiple processors, each of which can be a single-core processor (single-CPU) or a multi-core processor (multi-CPU). The term "processor" can refer to a single processor or a collective term for multiple processing elements. For example, a processor can be one or more central processing units (CPUs), other general-purpose processors, application-specific integrated circuits (ASICs), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs), 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 can be a microprocessor or any conventional processor. The processor can perform various functions of the electronic device by running or executing software programs stored in memory and by calling data stored in memory.

[0100] In a specific implementation, as one example, the processor may include one or more CPUs, for example... Figure 6 CPU0 and CPU1 shown are, of course, merely illustrative examples.

[0101] The memory is used to store the software program that executes the solution of the present invention, and the processor controls its execution. For specific implementation methods, please refer to the above method embodiments, which will not be repeated here.

[0102] Optionally, the memory may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory may be integrated with the processor or may exist independently, and may be accessed through the interface circuit of the electronic device (…). Figure 6 (Not shown in the image) is coupled to the processor; however, this embodiment of the invention does not impose specific limitations on this.

[0103] The transceiver may include a receiver and a transmitter. Figure 6 (Not shown separately). The receiver is used to implement the receiving function, and the transmitter is used to implement the transmitting function. The transceiver can be integrated with the processor or exist independently, and is connected through the interface circuit of the electronic device (…). Figure 6 (Not shown in the image) is coupled to the processor, and this embodiment of the invention does not specifically limit this.

[0104] In addition, it should be noted that, Figure 6 The structure of the electronic device shown is not intended to limit the device. Actual devices may include more or fewer components than shown, or combine certain components, or have different component arrangements. Furthermore, the technical effects achieved by this electronic device when performing the method of the first embodiment described above can be referenced to the technical effects described in the first embodiment; therefore, they will not be repeated here.

[0105] Third Embodiment

[0106] 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, etc. The instruction stored therein can be loaded and executed by a processor in a terminal.

[0107] Furthermore, it should be noted that the present invention can be provided as a method, apparatus, or computer program product. Therefore, embodiments of the present invention can take the form of a completely or partially hardware embodiment, a completely or partially software embodiment, or an embodiment combining software and hardware aspects. Moreover, when implemented in software, embodiments of the present invention can 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 includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any usable medium accessible to a computer or a data storage device such as a server or data center containing one or more sets of usable media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive (SSD).

[0108] Embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0109] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing terminal equipment to cause a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0110] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. The terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element. Furthermore, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone, where A and B can be singular or plural. Additionally, the character " / " in this text generally indicates an "or" relationship between the preceding and following objects, but it can also indicate an "AND / OR" relationship. Please refer to the context for specific interpretations. "At least one" refers to one or more items, while "more than" refers to two or more items. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or multiple items. For example, at least one of a, b, or c can be represented as: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0111] Furthermore, it is understood that in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply 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.

[0112] Those skilled in the art will recognize that the units and algorithm steps of the various examples 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 implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0113] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of functional modules / units 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 device, or some features may be ignored or not executed. Furthermore, the shown or discussed mutual couplings or direct couplings or communication connections may be through some interfaces; indirect couplings or communication connections between devices or units 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 can be selected to achieve the purpose of this embodiment according to actual needs. Additionally, the functional units in the various embodiments of this invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0114] If the method 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 the present invention, or the part that contributes to the prior art, or a 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 several 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 described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0115] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention. It should be pointed out that although preferred embodiments of the present invention have been described, those skilled in the art, once they understand the basic inventive concept of the present invention, can make several improvements and modifications without departing from the principles described herein. These 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 including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present invention.

Claims

1. A method for predicting the remaining lifetime of a device based on a broadcast attention mechanism, characterized in that, include: Collect equipment operating status data; The collected equipment operation status data is preprocessed, including data cleaning and denoising, data normalization, and data augmentation. The pre-processed equipment operating status data is used to train a preset equipment remaining life prediction model; the equipment remaining life prediction model includes a broadcast attention mechanism. Use a trained equipment remaining life prediction model to predict equipment remaining life; The device remaining lifetime prediction model includes: a convolutional scalar long short-term memory neural network module, a broadcast attention module, a convolutional difference decoder module, a feature fusion module, and a fully connected layer; The preprocessed equipment operating status data is input into a convolutional scalar long short-term memory neural network module to deeply mine the features between sequences and extract the first feature information of the data; the first feature information is input into a broadcast attention module, and the output of the broadcast attention module is input into a convolutional difference decoder module to extract the second feature information of the data; the feature fusion module fuses the first feature information and the second feature information to obtain fused feature information; the fully connected layer uses the fused feature information to predict the remaining life of the equipment. The data processing steps of the convolutional scalar long short-term memory neural network module include: First, the preprocessed device operating status data is processed by convolutional operations to generate feature maps; then, one-dimensional max pooling is used to convert each kernel size of the feature maps into the maximum number of single outputs; subsequently, for the output of one-dimensional max pooling, the sLSTM network is used to capture deeper spatiotemporal feature information of the data. The broadcast attention module consists of a 3×3 convolutional layer and a broadcast attention encoder. The activation function is ReLU, and Batch Normalization is used for training optimization. The data processing steps of the broadcast attention encoder include: First, three broadcast factors are initially set as m=3, n=5, and k=7, corresponding to three strip convolution kernels of different sizes; the three broadcast factors are adaptively and dynamically adjusted during training. The input to the broadcast attention encoder first passes through an n×n convolutional kernel in the backbone branch to capture feature information. The captured feature information is then input into the multimodal broadcast branch, which includes three branches of different scales. Each of the three branches of different scales employs a separable convolutional operation. These branches are concatenated with convolutional kernels of different sizes, and the scales corresponding to each branch are 1×m→m×1, 1×n→n×1, and 1×k→k×1, respectively. Features from branches at different scales are summed element-wise at the convergence node to fuse multi-scale information. Then, they are passed through a 1×1 convolutional layer to reduce or transform the channel dimension. Skip connections are used for gradient propagation and information reuse. The output of the main branch is directly connected to the final 1×1 convolutional layer. Addition is used to add the output of the main branch to the output of the multimodal broadcast branch. Multiplicative connections are used for feature weighting and selection, which allows the input to be multiplied element-wise with the final output to implement a gating mechanism, thereby controlling the transmission of feature information, enhancing important features and suppressing irrelevant features. The adaptive dynamic adjustment process of the broadcast factor during training includes: Within each training epoch, the average activation intensity and gradient contribution of the branch output features at each scale in the multimodal broadcast branch are statistically analyzed. The branches at different scales are sorted according to their gradient contributions. The kernel size of the convolution is reduced for branches with responses below a preset threshold, and the kernel size of the branches with responses above a preset threshold is increased, thus forming an adaptive scale evolution. Stable convergence is achieved through an exponential smoothing update formula, ensuring continuity and differentiability during the training process. The corrected m, n, and k can be used as a reference for the design of the next stage of convolutional kernels, or for the search and optimization stage. The convolutional interpolation decoder module includes four convolutional interpolation decoders; each convolutional interpolation decoder includes a bilinear interpolation module and a 3×3 convolution module. The data processing steps of the convolutional difference decoder module include: The output of the broadcast attention encoder is upsampled by bilinear interpolation using the first convolutional difference decoder, and the upsampled result is fused with the input of the broadcast attention encoder using a 3×3 convolution as the output of the first convolutional difference decoder. After the first convolutional difference decoder, each convolutional difference decoder upsamples the output of the broadcast attention encoder by bilinear interpolation, and the output of the broadcast attention encoder and the output of the previous convolutional difference decoder are fused using a 3×3 convolution. Finally, the features from the last three convolutional difference decoders are fused using pointwise convolution.

2. The device remaining lifetime prediction method based on broadcast attention mechanism as described in claim 1, characterized in that, The equipment operating status data includes temperature data, vibration data, current data, and voltage data during equipment operation; the data format adopts a multi-channel time series.

3. The device remaining lifetime prediction method based on broadcast attention mechanism as described in claim 1, characterized in that, The data cleaning and data denoising include: Use the sliding window mean to fill in missing data; An outlier data point was removed using a method based on One-Class SVM. Wavelet transform filtering is used for data denoising.

4. The device remaining lifetime prediction method based on broadcast attention mechanism as described in claim 1, characterized in that, The data enhancements include: Training samples are generated using a fixed-length sliding window; time series exceeding a preset length are randomly pruned.

Images