Virtual measurement method and device, equipment, storage medium and product

Abstract

The invention discloses a virtual measurement method and device, equipment, a storage medium and a product, and relates to the technical field of computers, and the virtual measurement method comprises the steps that multi-time-scale abstract processing is conducted on process parameters and production data sequences obtained online, and a multi-scale time sequence group is obtained; based on a preset virtual measurement model, virtual measurement is carried out on the multi-scale time sequence group, a key parameter measurement result is obtained, and the virtual measurement model is obtained by adopting a reinforcement learning mode to carry out virtual measurement training on a preset time sequence data sample. According to the method, the multi-scale time sequence group of the online time sequence characteristics is extracted, virtual measurement is performed on the online time sequence characteristics by using the trained virtual measurement model, the key parameter measurement result of the composite index is obtained, accurate prediction of the composite index is realized, and the accuracy of the prediction result is improved. Therefore, the problem that comprehensive monitoring of the composite indexes is difficult to achieve in precise laser cutting through a traditional method is solved, and the prediction accuracy is improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a virtual measurement method, apparatus, device, storage medium, and product. Background Technology

[0002] In the field of precision laser cutting, ensuring product quality and improving production efficiency are crucial for the sustainable development of enterprises. To achieve this goal, manufacturers need to monitor and analyze multiple key performance indicators in real time, covering aspects such as product size, shape, hardness, and surface quality. However, existing measurement methods rely on direct measurement using physical sensors, which is not only costly and involves complex equipment, but also makes it difficult to obtain key composite indicator data in some cases. Furthermore, direct measurement using physical sensors only focuses on a single dimension or a single time scale, resulting in insufficient understanding of the comprehensiveness and dynamism of composite indicators, thus affecting the accuracy of predictions.

[0003] Currently, research on virtual measurement in the field of precision laser cutting mainly revolves around methods based on image processing, machine learning, and big data analysis. However, these methods still face a series of common challenges in practical applications: key composite indicators are difficult to obtain directly, time-series information is not fully utilized, data preprocessing is complex, and the system's computational load is high, making it difficult to achieve comprehensive, real-time, and accurate monitoring of composite quality indicators.

[0004] To address the aforementioned issues, patent CN120123654A proposes a virtual measurement method and system based on a long-term series prediction architecture. This method integrates a deep learning model and a period calculator module to predict the future long-term series of specific data and provides a new technical path for obtaining composite indicators. Through the fusion analysis of multi-source process data, this system can indirectly derive composite quality indicators that are difficult to measure directly using traditional methods, thus compensating to some extent for the shortcomings of existing virtual measurement technologies in obtaining composite indicators.

[0005] However, this approach has certain limitations in practical applications, primarily its insufficient ability to capture and fuse multi-scale temporal features. The period calculator relies on an autocorrelation function to determine a fixed memory length and prediction window, a mechanism that makes it difficult for the model to effectively capture feature interactions across time scales. Due to the use of a uniform temporal processing granularity, the model cannot simultaneously characterize key physical phenomena at different time scales during laser cutting: for example, microsecond-level high-frequency laser power fluctuations, minute-level equipment thermal deformation effects, and hourly-level tool wear trends. This single-scale analytical framework limits the model's comprehensive understanding of the formation mechanism of composite quality indicators and makes it difficult to reveal the synergistic mechanism between physical processes at different time scales, thus affecting the accuracy and adaptability of the virtual measurement model in complex industrial scenarios.

[0006] Therefore, how to achieve comprehensive monitoring of composite indicators based on full utilization of time-series information is a problem that urgently needs to be solved.

[0007] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0008] The main objective of this application is to provide a virtual measurement method, device, equipment, storage medium, and product, which aims to solve the technical problem of difficulty in fully utilizing time-series information to achieve comprehensive monitoring of composite indicators.

[0009] To achieve the above objectives, this application proposes a virtual measurement method, which includes: The original time-series data is subjected to multi-time-scale abstraction processing to obtain a multi-scale time-series group; the original time-series data includes process parameters and production data sequences acquired online in chronological order during the precision laser cutting process; Based on a preset virtual measurement model, virtual measurements are performed on the multi-scale time series to obtain key parameter measurement results. The virtual measurement model is trained by using reinforcement learning to perform virtual measurements on preset time series data samples.

[0010] This technical solution constructs a closed-loop system with both deep perception and intelligent decision-making capabilities through the deep integration of multi-timescale abstraction and reinforcement learning virtual measurement. Its core advantage lies in its ability to extract multi-scale features that reveal the essence of the process from noisy raw data, laying a solid foundation for accurate modeling. Furthermore, by leveraging a reinforcement learning model with online self-learning capabilities, it achieves real-time, high-precision virtual measurement of key parameters, reducing reliance on physical sensors and continuously addressing production uncertainties such as equipment aging and material fluctuations. This provides a reliable technical path for achieving stable, efficient, and low-cost intelligent production.

[0011] In one embodiment, the step of performing multi-timescale abstraction processing on the original time-series data to obtain a multi-scale time-series group includes: Obtain process parameters and production data during precision laser cutting online; Based on the objective function, the preset multi-scale regularization coefficients and the process parameters and production data are calculated to obtain a multi-scale time series group, wherein the objective function includes: ， in, minObj This indicates that the goal function is to be minimized, and L represents the length of the time series. i , jRepresents sample points, s ij Represents sample points i With sample points j The similarity between them h ij ∈{0,1}, used to indicate sample points i Can it be sampled? j What it represents α represents the coefficients of the multi-scale regularization term, used for abstract scale control of the sequence; || represents the norm.

[0012] The raw data under different device states are acquired, and the multi-scale time series is obtained by solving the objective function and multi-scale regularization terms.

[0013] By using a pre-defined objective function, the process optimization problem is modeled as a problem of solving an objective function with multi-scale regularization constraints. By solving this function, a multi-scale time series of solutions that achieve Pareto optimality for multiple objectives is determined within the decision space composed of process parameters and production data.

[0014] The introduction of multi-scale regularization terms can adaptively extract and enhance the information that contributes most to the target at different time scales according to the preset optimization objective, thereby forming a holographic process view that can simultaneously reflect instantaneous anomalies and long-term trends, providing an unprecedented high-quality data foundation for subsequent analysis.

[0015] In one embodiment, before the step of performing virtual measurements on the multi-scale time series group based on a preset virtual measurement model to obtain the measurement results of key parameters, the method further includes: Obtain raw time-series data samples, which include process parameters and production data sequences corresponding to different equipment states; Based on the preset multi-scale regularization coefficients, objective function and the original time series data sample, the multi-scale time series group is calculated. Feature capture is performed on the multi-scale time series to obtain local features at different time spans; The local features are fused to obtain a multi-scale fused feature library; Based on the multi-scale time series, the multi-scale fusion feature library, and the device status, a virtual measurement model is trained using reinforcement learning.

[0016] By automatically identifying key patterns, waveforms, and statistical characteristics across different time scales, such as seconds and minutes, through local feature capture, no manual pre-setting of feature formulas is required. Fusing local features at different scales means the model can understand the combined effect of high-frequency vibrations and low-frequency drift. Constructing a multi-scale fused feature library through feature capture and fusion of multi-scale time series is crucial for achieving high-precision virtual measurement. This step automates the feature discovery process.

[0017] By taking equipment status as input, the model is no longer a static function of a single operating condition, but a dynamic intelligent agent capable of understanding equipment health. It can learn to distinguish between normal signals in new equipment states and normal signals in old equipment states, thus achieving adaptive measurement for different equipment states. This dynamic cognitive ability completely solves the model drift problem caused by equipment aging or batch differences in traditional models.

[0018] In one embodiment, the step of fusing the local features to obtain a multi-scale fused feature library includes: The local features extracted by different convolutional layers are concatenated to obtain comprehensive features; The LSTM network integrates comprehensive features and includes at least one LSTM layer, a fully connected layer, and an output layer. Based on the integrated features and the preset location codes, a multi-scale fusion feature library is obtained through encoder processing.

[0019] In one embodiment, the step of obtaining a multi-scale fused feature library through encoder processing based on the integrated comprehensive features and preset position codes includes: Based on feature fusion technology, the integrated comprehensive features and the preset position code are concatenated to obtain a comprehensive feature vector; The comprehensive feature vector is processed by an encoder to obtain a multi-scale fused feature library; wherein the encoder includes a self-attention mechanism and a position feedforward network.

[0020] Furthermore, to achieve the above objectives, this application also proposes a virtual measurement device, which includes: The preprocessing module is used to perform multi-timescale abstraction processing on the raw time-series data to obtain a multi-scale time-series group; the raw time-series data includes process parameters and production data sequences acquired online in chronological order during the precision laser cutting process; The measurement module is used to perform virtual measurements on the multi-scale time series group based on a preset virtual measurement model to obtain key parameter measurement results. The virtual measurement model is obtained by training preset time series data samples using reinforcement learning.

[0021] In addition, to achieve the above objectives, this application also proposes a virtual measurement device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the virtual measurement method as described above.

[0022] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and which, when executed by a processor, implements the steps of the virtual measurement method described above.

[0023] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the virtual measurement method described above.

[0024] One or more technical solutions proposed in this application have at least the following technical effects: This invention proposes a virtual measurement method. By extracting online time-series features and applying a trained virtual measurement model, the key parameter measurement results of composite indicators are obtained, enabling accurate prediction of composite indicators. This solves the problem that traditional methods are difficult to achieve comprehensive monitoring of composite indicators in precision laser cutting, thus improving the accuracy of prediction. Attached Figure Description

[0025] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0026] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a flowchart illustrating an embodiment of the virtual measurement method of this application. Figure 2 This is a simplified flowchart illustrating Embodiment 2 of the virtual measurement method of this application. Figure 3 This is a flowchart illustrating the virtual measurement method provided in Embodiment 2 of this application; Figure 4 This is a schematic diagram of the module structure of the virtual measurement device according to an embodiment of this application; Figure 5This is a schematic diagram of the device structure of the hardware operating environment involved in the virtual measurement method in the embodiments of this application; Figure 6 This is a flowchart illustrating Embodiment 3 of the virtual measurement method of this application; Figure 7 This is a flowchart illustrating Embodiment 4 of the virtual measurement method of this application.

[0028] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0030] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0031] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0032] The main solution of this application embodiment is: to perform multi-timescale abstraction processing on the original time series data to obtain a multi-scale time series group; the original time series data is a sequence of process parameters and production data acquired online; Based on a preset virtual measurement model, virtual measurements are performed on the multi-scale time series group to obtain key parameter measurement results. The virtual measurement model is obtained by training preset time series data samples using reinforcement learning.

[0033] In this embodiment, for ease of description, the following description uses a computer as the execution subject.

[0034] Due to the limitations of existing technologies, such as difficulty in obtaining key composite indicator data, insufficient processing of time-series information, high difficulty in data preprocessing, and high computational complexity, it is difficult to achieve comprehensive monitoring of composite indicators.

[0035] This application provides a solution that extracts multi-scale time series features from online time series features, applies a trained virtual measurement model to perform virtual measurement of the online time series features, obtains the measurement results of key parameters of composite indicators, and achieves accurate prediction of composite indicators. This solves the problem that traditional methods are difficult to achieve comprehensive monitoring of composite indicators in precision laser cutting, and achieves comprehensive monitoring of composite indicators, thereby improving the accuracy of prediction.

[0036] Combination Figure 2 In a practical implementation, the virtual measurement model is trained through reinforcement learning by performing time-series multi-scale abstract modeling, time-series segmentation and feature design, feature selection, multi-scale feature library and virtual measurement modeling on time-series data samples. By inputting online time-series data into the trained virtual measurement model, reinforcement learning is applied to the steps of time-series multi-scale abstraction processing, time-series segmentation and feature design, feature selection, and multi-scale feature library, thereby realizing online inference based on the virtual measurement model and obtaining the composite index state parameters of the equipment.

[0037] As can be seen from the above embodiments, this application obtains the key parameter measurement results of composite indicators by extracting multi-scale time series features of online time series and applying a trained virtual measurement model to perform virtual measurement of online time series features, thereby achieving accurate prediction of composite indicators. This solves the problem that traditional methods are difficult to achieve comprehensive monitoring of composite indicators in precision laser cutting, and achieves comprehensive monitoring of composite indicators, thereby improving the accuracy of prediction.

[0038] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device capable of performing the above functions. The following description uses a computer as an example to illustrate this embodiment and the subsequent embodiments.

[0039] Based on this, embodiments of this application provide a virtual measurement method, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the virtual measurement method of this application.

[0040] In this embodiment, the virtual measurement method includes steps S10 to S20: Step S10: Perform multi-time-scale abstraction processing on the original time series data to obtain a multi-scale time series group; It should be noted that the original time-series data involved in this invention specifically refers to the sequence of process parameters and production data generated during the precision laser cutting process. These data sequences completely record the composite indicators of the production data corresponding to each key process parameter during the production process.

[0041] Furthermore, it should be noted that the multi-timescale abstraction method employed in this invention is an advanced processing technique for process parameters and production data sequence data. This method first abstracts the process parameters and production data sequence data according to different time scales to extract data features at different time dimensions. Subsequently, it solves these abstracted data using an objective function, ultimately obtaining a time-series dataset that accurately reflects the dynamic characteristics of the production process. This modeling method not only improves the efficiency and accuracy of data processing but also provides strong data support for subsequent applications such as prediction and optimization.

[0042] Understandably, the multi-scale time series group constructed in this invention represents an innovative processing method for the aforementioned original time series data. Specifically, this process abstracts and reorganizes the original time series data according to different time scales, thereby obtaining a completely new time series dataset. This multi-scale processing method can reveal the inherent patterns and trends of data from different time dimensions, providing richer and more comprehensive information support for subsequent data analysis and modeling.

[0043] In one feasible implementation, refer to Figure 6 Step S10 may include steps A11 to A12: Step A11: Obtain raw time series data online; The system reads raw data from sensors, preprocesses it to obtain real-time status of process parameters and related production data during the manufacturing process, and performs noise reduction on this raw data to eliminate possible interference, noise and outliers, thereby obtaining high-quality raw time-series data.

[0044] It should be noted that the raw time-series data refers to a sequence of process parameters and production data arranged in chronological order. These data sequences completely record the changes in various process parameters, such as temperature, pressure, and humidity, as well as the corresponding changes in production data, such as product size and output, within a specific time period.

[0045] It is understandable that the raw data from the sensors is difficult to use directly and requires analysis, filtering, transformation, and encoding to obtain accurate feature data. By identifying and eliminating data outside the predicted distribution, potential interference, noise, and outliers can be eliminated, resulting in high-quality raw time-series data.

[0046] Step A12: Based on the objective function, the preset multi-scale regularization coefficients and the original time series data are solved to obtain a multi-scale time series group, wherein the objective function includes: ， in, minObj This indicates that the goal function is to be minimized, and L represents the length of the time series. i , j Represents sample points, s ij Represents sample points i With sample points j The similarity between them h ij ∈{0,1}, used to indicate sample points i Can it be sampled? j What it represents αrepresents the coefficients of the multi-scale regularization term, used for abstract scale control of the sequence; || represents the norm.

[0047] Additionally, it should be noted that when the sample points are expanded to a subset of samples, some distribution-based similarity metrics, such as the maximum mean difference, can be sampled to improve the non-linear expressive power of the model.

[0048] The multi-scale time series group is a time series dataset obtained by abstracting and reorganizing the original time series data according to different time scales.

[0049] This application obtains the original time series data and then calculates the preset multi-scale regularization coefficients and the original time series data based on the objective function to obtain a multi-scale time series group, thus solving the technical problems of difficulty in obtaining key composite index data and insufficient time series information processing.

[0050] The above is only one feasible implementation of step S10 provided in this embodiment. This embodiment does not specifically limit the specific implementation of step S10.

[0051] Step S20: Based on a preset virtual measurement model, perform virtual measurement on the multi-scale time series group to obtain key parameter measurement results. The virtual measurement model is obtained by training preset time series data samples using reinforcement learning.

[0052] It is understandable that by performing virtual measurements on multi-scale time series groups based on a pre-defined virtual measurement model, the measurement results of key parameters are obtained. The virtual measurement model can be trained locally or in the cloud. By deploying a trained virtual measurement model in the cloud, the beneficial effects of performance improvement and failure rate reduction can be achieved.

[0053] It should be noted that the virtual measurement model is a reinforcement learning model deployed locally or in the cloud, which is obtained by training virtual measurement on preset time series data samples. The multi-scale time series group is a time series dataset obtained by abstracting and reorganizing the original time series data according to different time scales.

[0054] In one feasible implementation, refer to Figure 7 Step S20 may include steps A21 to A25: Step A21: Based on the preset multi-scale regularization coefficients, objective function, and the time series data samples, the multi-scale time series group is calculated. It should be noted that the time series data sample is sample data of process parameters and production data sequence, the virtual measurement model is a reinforcement learning model obtained by virtual measurement training of preset time series data samples, and the regularization coefficient is used for abstract scale control of the sequence.

[0055] Additionally, it should be noted that the objective function is: ， in, minObj This indicates that the goal function is to be minimized, and L represents the length of the time series. i , j Represents sample points, s ij Represents sample points i With sample points j The similarity between them h ij ∈{0,1}, used to indicate sample points i Can it be sampled? j What it represents α The coefficients of the multi-scale regularization term are used for abstract scale control of the sequence; || represents the norm. When the sample points are expanded to a subset of samples, some distribution-based similarity metrics, such as the maximum mean difference, can be sampled to improve the non-linear expressive power of the model.

[0056] Step A22: Perform feature capture on the multi-scale time series to obtain local features at different time spans; For multi-scale temporal groups, one-dimensional convolutional layers with different kernel sizes are used. The learning ability of the model is enhanced by applying batch normalization and ReLU (Rectified Linear Unit) activation function. Max pooling layers are applied to reduce the feature dimensionality and capture the most significant local features across different time spans.

[0057] Understandably, conventional dimensionality reduction achieves this by removing redundant or irrelevant features. This application utilizes max pooling layers to reduce sampling, thereby decreasing data dimensionality and computational complexity, thus achieving the goal of dimensionality reduction. Understandably, in the process of virtual measurement, if the high-dimensional feature space is not reduced, retaining attributes unrelated to the output will decrease the overall effectiveness of the model.

[0058] It should be noted that the local features refer to features extracted from a local region of the input data through convolution operations; the ReLU activation function is the most popular non-linear activation function in the field of deep learning; the max pooling layer is a special layer widely used in convolutional neural networks to downsample the input feature map to reduce data dimensionality and computational complexity; and batch normalization is a technique in deep learning to solve the internal covariate shift problem when training deep neural networks and accelerate the convergence speed of the network.

[0059] Step A23: Fuse the local features to obtain a multi-scale fused feature library; Local features extracted from different convolutional layers are concatenated to form a comprehensive feature; one or more LSTM (Long Short-Term Memory) layers and fully connected layers are used to capture long-term dependencies in the time series, integrate the concatenated comprehensive features, and design the output layer by selecting an appropriate activation function; fixed or learnable positional encodings are added to the comprehensive features; the comprehensive features with added positional encodings are input into an encoder and processed through one or more encoders to obtain a multi-scale fusion feature library.

[0060] It should be noted that the LSTM layer is a deep learning model specifically designed for processing sequential data, while the fully connected layer is a commonly used neural network layer in deep learning, also known as a densely connected layer. It is used to further integrate and abstract the features learned by the LSTM layer to output more comprehensive time series features.

[0061] Additionally, it should be noted that each encoder consists of a self-attention mechanism and a positional feedforward network, enabling it to capture long-range dependencies in time-series data. Multi-head attention is used to enhance the model's representational power, allowing it to simultaneously attend to information from different time scales.

[0062] Step A24: Based on the multi-scale temporal group and the multi-scale fusion feature library, train the virtual measurement model through reinforcement learning. Based on the multi-scale time series group and the multi-scale fusion feature library, a virtual measurement model is trained through reinforcement learning according to time series features, process parameter features and equipment status features.

[0063] It should be noted that the time series features are extracted using a one-dimensional convolutional network to extract features at different time scales; the process parameter features are obtained by compressing the dimensionality of process parameters through an autoencoder while retaining key information; and the equipment status features are obtained by applying feature selection techniques to identify the equipment status parameters that have the greatest impact on the prediction results.

[0064] Additionally, it should be noted that training a virtual measurement model through reinforcement learning refers to using a multi-head attention mechanism to enhance the model's representational capabilities, enabling the trained model to simultaneously attend to information from different time scales.

[0065] Step A25: Input the multi-scale time series group into the virtual measurement model. The virtual measurement model captures features of the multi-scale time series group to obtain local features at different time spans, and fuses the local features to obtain a multi-scale fusion feature library. Based on the multi-scale fusion feature library, the measurement results of key parameters are obtained.

[0066] It should be noted that the multi-scale time series group is a time series dataset obtained by abstracting and reorganizing the original time series data according to different time scales; the virtual measurement model is a reinforcement learning model obtained by virtual measurement training of preset time series data samples; the local features are features extracted from a local region of the input data through convolution operations; and the key parameter measurement results are the measurement results of the composite index of the virtual modeled item.

[0067] This application, based on an objective function, calculates the preset multi-scale regularization coefficients and the original time-series data to obtain a multi-scale time-series group; it captures features from the multi-scale time-series group to obtain local features at different time spans; it fuses these local features to obtain a multi-scale fused feature library; based on the multi-scale time-series group and the multi-scale fused feature library, it trains a virtual measurement model using reinforcement learning; it inputs the multi-scale time-series group into the virtual measurement model and receives the local features at different time spans output by the virtual measurement model; it inputs the local features into the virtual measurement model and receives the multi-scale fused feature library output by the virtual measurement model; it inputs the multi-scale fused feature library into the virtual measurement model and receives the key parameter measurement results output by the virtual measurement model. This solves the problem of traditional methods struggling to achieve comprehensive monitoring of composite indicators in precision laser cutting, achieving comprehensive monitoring of composite indicators and thus improving prediction accuracy.

[0068] The above is only one feasible implementation of step S20 provided in this embodiment. This embodiment does not specifically limit the specific implementation of step S20.

[0069] This embodiment provides a virtual measurement method. By extracting multi-scale time series groups of online time series features, a trained virtual measurement model is applied to perform virtual measurement on the online time series features to obtain the measurement results of key parameters of composite indicators, thereby achieving accurate prediction of composite indicators. This solves the problem that traditional methods are difficult to achieve comprehensive monitoring of composite indicators in precision laser cutting, and realizes comprehensive monitoring of composite indicators, thereby improving the accuracy of prediction.

[0070] For example, to help understand the implementation flow of the virtual measurement method obtained by combining this embodiment with the above embodiment one, please refer to... Figure 3 , Figure 3 A simplified flowchart of a virtual measurement method is provided, specifically: First, time-series data samples are abstracted using multiple time scales to obtain multi-scale time-series cluster samples. One-dimensional convolutional layers with kernels of different sizes are used on these multi-scale time-series cluster samples. Batch normalization and ReLU (Rectified Linear Unit) activation functions are applied to enhance the model's learning ability. Max pooling layers are applied to reduce feature dimensionality, capturing the most significant local feature samples across different time spans. The local feature samples extracted from different convolutional layers are concatenated to form a comprehensive feature sample. One or more LSTM (Long Short-Term Memory) layers and fully connected layers are used to capture long-term dependencies in the time series. The concatenated comprehensive feature samples are integrated, and an appropriate activation function is selected to design the output layer. Fixed or learnable positional encodings are added to the comprehensive feature samples. The comprehensive feature samples with added positional encodings are input into an encoder and processed through one or more encoder layers to obtain a multi-scale fused feature library sample. Based on the multi-scale time-series cluster samples and the multi-scale fused feature library sample, a virtual measurement model is trained using reinforcement learning. The virtual measurement model is deployed in the cloud. When raw time-series data is acquired, it is sent to the virtual measurement model in the cloud. Through a multi-scale time-series abstraction step, a multi-scale time-series group is obtained. This multi-scale time-series group is then sent to the virtual measurement model in the cloud. Through a feature capture step in the virtual measurement model, local features at different time spans are obtained. Through a feature splicing step in the virtual measurement model, comprehensive features are obtained. Through a feature integration step in the virtual measurement model, integrated comprehensive features are obtained. Through an encoding step in the virtual measurement model, based on the integrated comprehensive features and a preset position code, a multi-scale fusion feature library is obtained through encoder processing. Through a measurement step in the virtual measurement model, key parameter measurement results are obtained based on the multi-scale fusion feature library. By analyzing the key parameter measurement results, the composite index of the current test item is obtained, realizing online inference based on the virtual measurement model and achieving the technical effect of obtaining the composite index status parameters of the equipment.

[0071] As can be seen from the above embodiments, this application obtains the key parameter measurement results of composite indicators by extracting multi-scale time series features of online time series and applying a trained virtual measurement model to perform virtual measurement of online time series features, thereby achieving accurate prediction of composite indicators. This solves the problem that traditional methods are difficult to achieve comprehensive monitoring of composite indicators in precision laser cutting, and achieves comprehensive monitoring of composite indicators, thereby improving the accuracy of prediction.

[0072] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the virtual measurement method of this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0073] This application also provides a virtual measurement device; please refer to... Figure 4 The virtual measurement device includes: The preprocessing module is used to perform multi-time-scale abstraction on the raw time-series data to obtain multi-scale time-series groups; The measurement module is used to perform virtual measurements on the multi-scale time series group based on a preset virtual measurement model to obtain key parameter measurement results. The virtual measurement model is obtained by training preset time series data samples using reinforcement learning.

[0074] And / or, the preprocessing module includes: a solution module for acquiring the original time series data; and a first solution module for solving the preset multi-scale regularization coefficients and the original time series data based on an objective function to obtain a multi-scale time series group, wherein the objective function includes: ， in, minObj This indicates that the goal function is to be minimized, and L represents the length of the time series. i , j Represents sample points, s ij Represents sample points i With sample points j The similarity between them h ij ∈{0,1}, used to indicate sample points i Can it be sampled? j What it represents α represents the coefficients of the multi-scale regularization term, used for abstract scale control of the sequence; || represents the norm.

[0075] And / or, the virtual measurement device further includes: a sample acquisition module for acquiring time-series data samples; a sample calculation module for calculating a multi-scale time series group based on preset multi-scale regularization coefficients, an objective function, and the time-series data samples; a feature capture module for capturing features of the multi-scale time series group to obtain local features at different time spans; a feature splicing module for splicing the local features to obtain a multi-scale fusion feature library; and a model training module for training a virtual measurement model using reinforcement learning based on the multi-scale time series group, the multi-scale fusion feature library, and the device status.

[0076] And / or, the feature concatenation module includes: a first concatenation module, used to concatenate the features of the different convolutional layers based on the LSTM network to obtain comprehensive features; An integration module is used to integrate comprehensive features based on an LSTM network, wherein the LSTM network includes at least one LSTM layer, a fully connected layer, and an output layer; a feature fusion module is used to obtain a multi-scale fused feature library by processing the comprehensive features and a preset position code through an encoder.

[0077] And / or, the feature fusion module includes: a second splicing module, used to splice the integrated comprehensive features and the preset position code based on feature fusion technology to obtain a comprehensive feature vector; and a fusion analysis module, used to process the comprehensive feature vector through an encoder to obtain a multi-scale fusion feature library.

[0078] And / or, the measurement module includes: a local feature module for capturing features of the multi-scale time series to obtain local features at different time spans; a feature library module for fusing the local features to obtain a multi-scale fused feature library; and a key measurement module for obtaining key parameter measurement results based on the multi-scale fused feature library.

[0079] The virtual measurement device provided in this application, employing the virtual measurement method in the above embodiments, can solve the technical problem of the inability to achieve comprehensive monitoring of composite indicators. Compared with the prior art, the beneficial effects of the virtual measurement device provided in this application are the same as those of the virtual measurement method provided in the above embodiments, and other technical features in the virtual measurement device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0080] This application provides a virtual measurement device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform the virtual measurement method in Embodiment 1 above.

[0081] The following is for reference. Figure 5 This document illustrates a structural schematic diagram of a virtual measurement device suitable for implementing embodiments of this application. The virtual measurement device in these embodiments may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 5 The virtual measurement device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0082] like Figure 5 As shown, the virtual measurement device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touchscreen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1003 including, for example, magnetic tape, hard disk, etc.; and communication devices 1009. Communication device 1009 allows the virtual measurement device to communicate wirelessly or wiredly with other devices to exchange data. Although the figures show virtual measurement devices with various systems, it should be understood that implementing or having all of the systems shown is not required. More or fewer systems may be implemented alternatively.

[0083] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0084] The virtual measurement device provided in this application, employing the virtual measurement method described in the above embodiments, can solve the technical problem of the inability to achieve comprehensive monitoring of composite indicators. Compared with the prior art, the beneficial effects of the virtual measurement device provided in this application are the same as those of the virtual measurement method provided in the above embodiments, and other technical features of this virtual measurement device are the same as those disclosed in the method of the previous embodiment, and will not be repeated here.

[0085] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0086] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0087] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the virtual measurement method in the above embodiments.

[0088] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0089] The aforementioned computer-readable storage medium may be included in the virtual measurement device; or it may exist independently and not assembled into the virtual measurement device.

[0090] The aforementioned computer-readable storage medium carries one or more programs. When the aforementioned one or more programs are executed by the virtual measurement device, the virtual measurement device: performs multi-timescale abstraction processing on the original time-series data to obtain a multi-scale time-series group; and performs virtual measurement on the multi-scale time-series group based on a preset virtual measurement model to obtain key parameter measurement results. The virtual measurement model is obtained by using reinforcement learning to train a preset time-series data sample using virtual measurement.

[0091] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0092] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0093] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0094] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described virtual measurement method, thereby solving the technical problem of the inability to achieve comprehensive monitoring of composite indicators. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the virtual measurement method provided in the above embodiments, and will not be repeated here.

[0095] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the virtual measurement method described above.

[0096] The computer program product provided in this application can solve the technical problem of the inability to achieve comprehensive monitoring of composite indicators. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the virtual measurement method provided in the above embodiments, and will not be repeated here.

[0097] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A virtual measurement method, characterized in that, The virtual measurement method includes: The original time-series data is subjected to multi-time-scale abstraction processing to obtain a multi-scale time-series group; the original time-series data includes process parameters and production data sequences acquired online in chronological order during the precision laser cutting process; Based on a preset virtual measurement model, virtual measurements are performed on the multi-scale time series to obtain key parameter measurement results. The virtual measurement model is trained by using reinforcement learning to perform virtual measurements on preset time series data samples.

2. The virtual measurement method as described in claim 1, characterized in that, The step of performing multi-time-scale abstraction processing on the original time-series data to obtain a multi-scale time-series group includes: Obtain process parameters and production data during precision laser cutting online; Based on the objective function, the preset multi-scale regularization coefficients and the process parameters and production data are solved to obtain a multi-scale time series group, wherein the objective function includes: in, minObj This indicates that the goal function is to be minimized, and L represents the length of the time series. i , j s represents the sample points. ij Represents sample points i With sample points j The similarity between them h ij ∈{0,1}, used to indicate sample points i Can it be sampled? j What it represents α represents the coefficients of the multi-scale regularization term, used for abstract scale control of the sequence; || represents the norm.

3. The virtual measurement method as described in claim 1, characterized in that, Before the step of performing virtual measurements on the multi-scale time series group based on a preset virtual measurement model to obtain the measurement results of key parameters, the method further includes: Obtain raw time-series data samples, which include process parameters and production data sequences corresponding to different equipment states; Based on the preset multi-scale regularization coefficients, objective function and the original time series data sample, the multi-scale time series group is calculated. Feature capture is performed on the multi-scale time series to obtain local features at different time spans; The local features are fused to obtain a multi-scale fused feature library; Based on the multi-scale time series, the multi-scale fusion feature library, and the device status, a virtual measurement model is trained using reinforcement learning.

4. The virtual measurement method as described in claim 3, characterized in that, The step of fusing the local features to obtain a multi-scale fused feature library includes: The local features extracted by different convolutional layers are concatenated to obtain comprehensive features; The LSTM network integrates comprehensive features and includes at least one LSTM layer, a fully connected layer, and an output layer. Based on the integrated features and the preset location codes, a multi-scale fusion feature library is obtained through encoder processing.

5. The virtual measurement method as described in claim 4, characterized in that, The step of obtaining a multi-scale fused feature library by an encoder based on the integrated comprehensive features and preset position codes includes: Based on feature fusion technology, the integrated comprehensive features and the preset position code are concatenated to obtain a comprehensive feature vector; The comprehensive feature vector is processed by an encoder to obtain a multi-scale fused feature library; wherein the encoder includes a self-attention mechanism and a position feedforward network.

6. The virtual measurement method as described in claim 1, characterized in that, The steps for performing virtual measurements on the multi-scale time series group based on a preset virtual measurement model to obtain key parameter measurement results include: The multi-scale time series is input into the virtual measurement model, which captures features of the multi-scale time series to obtain local features at different time spans, and fuses the local features to obtain a multi-scale fusion feature library; the key parameter measurement results are obtained based on the multi-scale fusion feature library.

7. A virtual measurement device, characterized in that, The device includes: The preprocessing module is used to perform multi-timescale abstraction processing on the raw time-series data to obtain a multi-scale time-series group; the raw time-series data includes process parameters and production data sequences acquired online in chronological order during the precision laser cutting process; The measurement module is used to perform virtual measurements on the multi-scale time series group based on a preset virtual measurement model to obtain key parameter measurement results. The virtual measurement model is obtained by training preset time series data samples using reinforcement learning.

8. A virtual measurement device, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the virtual measurement method as described in any one of claims 1 to 6.

9. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the virtual measurement method as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the virtual measurement method as described in any one of claims 1 to 6.

Images