Method and system for predicting and optimizing production energy consumption of printed circuit board in real time

Abstract

The invention discloses a printed circuit board production energy consumption real-time prediction optimization method and system, and relates to the field of printed circuit board production energy consumption, and the method comprises the steps: collecting equipment operation state data and environment parameters in real time through setting an edge data collection node; receiving data acquired by each edge data acquisition node in real time by utilizing an internet of things technology, and establishing a printed circuit board production energy consumption database; constructing a printed circuit board production energy consumption evaluation model, and comprehensively evaluating and predicting the production energy consumption of the printed circuit board; according to an evaluation result of the printed circuit board production energy consumption evaluation model, obtaining a weight value of production energy consumption of each edge data acquisition node; and obtaining the relevance of the production energy consumption of each edge data acquisition node, and dynamically adjusting and optimizing the production energy consumption of the printed circuit board. The method has the advantages that the production energy consumption of the printed circuit board is dynamically evaluated and the waste heat recovery is optimized through the multi-mode sensing data and the hybrid machine learning model, so that the productivity loss of the printed circuit board is remarkably reduced.

Description

Technical Field

[0001] The present invention relates to the field of energy consumption in printed circuit board production, and specifically to a method and system for real-time prediction and optimization of energy consumption in printed circuit board production. Background Art

[0002] As a key component of electronic products, the production process of printed circuit boards (PCBs) involves many complex processes, such as numerically controlled drilling, circuit interconnection via hole formation, application of photosensitive film, exposure, development, etching, stripping, AOI optical inspection, printing of ink, surface treatment, continuity testing, reinforcement / glue application, punching, encapsulation / assembly, functional testing, appearance inspection, etc. The entire production process is long and there are many types of processing equipment, resulting in a huge total energy consumption.

[0003] Currently, with the continuous growth of global energy demand and the increasing prominence of environmental issues, the energy consumption problem in the printed circuit board production industry has received increasing attention. On the one hand, the high energy cost increases the production cost of enterprises and reduces their market competitiveness. On the other hand, a large amount of energy consumption also poses a great pressure on the environment. However, in the existing printed circuit board production process, the management of energy consumption mostly stays in the stage of post-statistical analysis, lacking real-time prediction and optimized control of energy consumption during production, and unable to adjust production parameters and equipment operating states in a timely manner during production to achieve energy consumption reduction. Therefore, it is of great practical significance to develop a method and system that can real-time predict and optimize the energy consumption of printed circuit board production.

[0004] The existing printed circuit board production energy consumption management technology mainly focuses on the model construction of a single link, lacking the global perception data dimension of printed circuit board production, unable to comprehensively process time series dependence, spatial correlation, and dynamic attention mechanism, and lacking the dynamic optimization ability based on the correlation between real-time weights and energy consumption, resulting in a rigid waste heat recovery strategy, a lag in system response, and difficulty in adapting to the dynamic changes of the production environment and process, ultimately causing problems such as low energy consumption prediction accuracy, poor optimization effect, and energy waste. Summary of the Invention

[0005] To solve the above technical problems, a method and system for real-time prediction and optimization of energy consumption in printed circuit board production are provided. This technical solution solves the problems mentioned in the above background art, such as lacking the global perception data dimension of printed circuit board production, being unable to comprehensively process time series dependence, spatial correlation, and dynamic attention mechanism, and lacking the dynamic optimization ability based on the correlation between real-time weights and energy consumption, resulting in a rigid waste heat recovery strategy, a lag in system response, and difficulty in adapting to the dynamic changes of the production environment and process, ultimately causing problems such as low energy consumption prediction accuracy, poor optimization effect, and energy waste.

[0006] To achieve the above object, the technical solution adopted by the present invention is as follows: A real-time prediction and optimization method for the energy consumption in the production of printed circuit boards, comprising: Deploy a multi-modal sensor array in each production process of the printed circuit board, and set edge data acquisition nodes to collect device operation status data and environmental parameters in real time; Utilize the Internet of Things technology to receive the data collected by each edge data acquisition node in real time, and establish a printed circuit board production energy consumption database; Based on the LSTM-Transformer-GNN hybrid machine learning algorithm, construct a printed circuit board production energy consumption evaluation model to comprehensively evaluate and predict the energy consumption in the production of printed circuit boards; According to the evaluation results of the printed circuit board production energy consumption evaluation model, obtain the weight values of the production energy consumption of each edge data acquisition node; Obtain the correlation of the production energy consumption of each edge data acquisition node, and dynamically adjust and optimize the energy consumption in the production of printed circuit boards through waste heat recovery optimization.

[0007] Preferably, the constructing a printed circuit board production energy consumption evaluation model based on the LSTM-Transformer-GNN hybrid machine learning algorithm to comprehensively evaluate and predict the energy consumption in the production of printed circuit boards specifically includes: According to the production process cycle of the printed circuit board, set the window length of the LSTM layer time series, wherein the window length of the LSTM layer time series is within the range of the production process cycle of the printed circuit board; According to the printed circuit board production energy consumption database, use the data collected by each edge data acquisition node as the input quantity; Based on the LSTM neural network, output the device operation status data and environmental parameters with time series, and use them as the input quantity of the Transformer model; Through linear projection, map the data collected by each edge data acquisition node in real time to the same-dimensional space as Key / Value; According to the Transformer model, calculate the correlation between each data in the input quantity and other data, and capture the dependency relationship between any data in the input quantity data; According to the printed circuit board process parameter knowledge graph, based on the mean square error formula, establish a constraint loss function of the LSTM-Transformer model; According to the LSTM-Transformer model, judge and predict the energy consumption value of the printed circuit board production of each edge data acquisition node and the total energy consumption value of the printed circuit board production within the window length; Based on the production processes of each edge data collection node in the production process of printed circuit boards, and based on the CNN neural network, the edge weights are adjusted in real time to accurately reflect the impact of production scheduling changes on energy consumption, and further correct and optimize the energy consumption value of printed circuit board production; The expression for setting the window length of the LSTM layer time series is: ; In the formula, is the window length of the LSTM layer time series, is the production process cycle of the printed circuit board, is the dominant angular frequency in the production data of the printed circuit board, is the floor function symbol to ensure that the window length is an integer time unit; The expression for the constraint loss function of the LSTM-Transformer model is: ; In the formula, is the value of the constraint loss function of the LSTM-Transformer model, is the number of sample groups, is the predicted value of the printed circuit board production energy consumption by the model, is the target value of the printed circuit board production energy consumption, is the weight coefficient with constraint terms in the knowledge graph of printed circuit board process parameters, is the sample set with constraint terms in the knowledge graph of printed circuit board process parameters, is the standard boundary threshold in the knowledge graph of printed circuit board process parameters.

[0008] Preferably, obtaining the weight value of the production energy consumption of each edge data collection node according to the evaluation result of the printed circuit board production energy consumption evaluation model specifically includes: According to the evaluation result of the printed circuit board production energy consumption evaluation model, obtain the printed circuit board production energy consumption values of each edge data collection node within the window length; According to the printed circuit board production energy consumption evaluation model, extract the weights in the LSTM-Transformer-GNN hybrid model respectively; Among them, LSTM is the historical data weight of each edge data collection node, Transformer is the self-attention weight of each edge data collection node, and GNN is the spatial weight of each edge data collection node; According to the weighting formula, obtain the comprehensive weight value of the production energy consumption of each edge data collection node; According to the Softmax normalization formula, normalize the comprehensive weight values of the production energy consumption of each edge data collection node so that the sum of all comprehensive weight values is 1; The expression for the comprehensive weight value of the production energy consumption of each edge data acquisition node is as follows: ; In the formula, is the comprehensive weight value of the production energy consumption of the edge data acquisition node, , , are respectively the historical data weight, the self-attention weight, and the spatial weight of the edge data acquisition node, , , are respectively the bias coefficients of the historical data weight, the self-attention weight, and the spatial weight of the edge data acquisition node, which can be determined by cross-validation.

[0009] Preferably, obtaining the correlation of the production energy consumption of each edge data acquisition node, and dynamically adjusting and optimizing the production energy consumption of the printed circuit board through waste heat recovery specifically includes: According to the printed circuit board production energy consumption database, obtain the device operation status data and environmental parameters of each edge data acquisition node; Based on the device operation status data, environmental parameters, and production energy consumption data of each edge data acquisition node, and based on the Pearson correlation formula, determine the correlation of the production energy consumption of each edge data acquisition node; According to the correlation of the production energy consumption of each edge data acquisition node and the correlation coupling of each node, establish a dynamic adjustment and optimization model for the production energy consumption of the printed circuit board; According to the dynamic adjustment and optimization model of the printed circuit board production energy consumption, dynamically adjust the start / stop and parameters of waste heat recovery.

[0010] Furthermore, this solution proposes a real-time prediction and optimization system for the production energy consumption of printed circuit boards, which is used to implement the real-time prediction and optimization method for the production energy consumption of printed circuit boards as described above, including: A data acquisition module, which is used to deploy a multi-modal sensor array in each production process of the printed circuit board, and by setting edge data acquisition nodes, collect device operation status data and environmental parameters in real time; use the Internet of Things technology to receive the data collected by each edge data acquisition node in real time, and establish a printed circuit board production energy consumption database; A prediction and optimization module, which is used to build an energy consumption evaluation model for printed circuit board production based on the LSTM-Transformer-GNN hybrid machine learning algorithm, comprehensively evaluate and predict the energy consumption of printed circuit board production; according to the evaluation results of the energy consumption evaluation model for printed circuit board production, obtain the weight values of the energy consumption of each edge data acquisition node; obtain the relevance of the energy consumption of each edge data acquisition node, and dynamically adjust and optimize the energy consumption of printed circuit board production through waste heat recovery optimization.

[0011] Preferably, the data acquisition module includes: A data acquisition unit, which is used to deploy a multi-modal sensor array in each production process of the printed circuit board, and set edge data acquisition nodes to collect device operation status data and environmental parameters in real time; A data processing unit, which is used to use the Internet of Things technology to receive the data collected by each edge data acquisition node in real time and establish an energy consumption database for printed circuit board production.

[0012] Preferably, the prediction and optimization module includes: An evaluation and prediction unit, which is used to build an energy consumption evaluation model for printed circuit board production based on the LSTM-Transformer-GNN hybrid machine learning algorithm, and comprehensively evaluate and predict the energy consumption of printed circuit board production; A node weight unit, which is used to obtain the weight values of the energy consumption of each edge data acquisition node according to the evaluation results of the energy consumption evaluation model for printed circuit board production; An energy consumption adjustment unit, which is used to obtain the relevance of the energy consumption of each edge data acquisition node, and dynamically adjust and optimize the energy consumption of printed circuit board production through waste heat recovery optimization.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention provides a real-time prediction and optimization method for the energy consumption in the production of printed circuit boards. By deploying a multi-modal sensor array, the operation status data of equipment and multi-dimensional data of environmental parameters are synchronized, and real-time data acquisition is realized in combination with edge data acquisition nodes. An energy consumption database for the production of printed circuit boards covering all technological links is constructed, providing a richer data basis for subsequent energy consumption analysis. At the same time, based on the LSTM-Transformer-GNN hybrid machine learning algorithm, an energy consumption evaluation model for the production of printed circuit boards is constructed to comprehensively evaluate and predict the energy consumption in the production of printed circuit boards, realizing in-depth mining and accurate prediction of energy consumption data, and significantly improving the reliability of the evaluation results. Secondly, according to the energy consumption evaluation model for the production of printed circuit boards, the weights in the LSTM-Transformer-GNN hybrid model are extracted respectively, and according to the weighted formula, the comprehensive weight value of the energy consumption of each edge data acquisition node is obtained, innovatively integrating the time series modeling ability of LSTM, the self-attention mechanism of Transformer, and the spatial correlation analysis ability of GNN. Finally, through the correlation between the comprehensive weight value and the energy consumption, combined with the thermal mass coupling characteristics of the waste heat recovery system, the real-time dynamic adjustment of the recovery start / stop and parameters is realized, enabling the system to adapt to the production schedule fluctuations, maximizing the waste heat utilization rate and reducing the total energy consumption, so as to achieve the purpose of dynamically evaluating the energy consumption in the production of printed circuit boards and optimizing the waste heat recovery through multi-modal perception data and hybrid machine learning models, and significantly reducing the production capacity loss of printed circuit boards. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 is a flowchart of the real-time prediction and optimization method for the energy consumption in the production of printed circuit boards according to the present invention; Figure 2 is a flowchart of constructing an energy consumption evaluation model for the production of printed circuit boards based on the LSTM-Transformer-GNN hybrid machine learning algorithm according to the present invention to comprehensively evaluate and predict the energy consumption in the production of printed circuit boards; Figure 3 is a flowchart of obtaining the weight value of the energy consumption of each edge data acquisition node according to the evaluation result of the energy consumption evaluation model for the production of printed circuit boards according to the present invention; Figure 4 is a flowchart of obtaining the correlation of the energy consumption of each edge data acquisition node, and dynamically adjusting and optimizing the energy consumption in the production of printed circuit boards through waste heat recovery optimization according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0015] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments in the following description are only examples, and those skilled in the art can think of other obvious variations.

[0016] Refer to Figure 1As shown in the figure, a real-time prediction and optimization method for the energy consumption in the production of printed circuit boards includes: Deploy a multi-modal sensor array in each production process of the printed circuit board. By setting edge data acquisition nodes, collect the device operation status data and environmental parameters in real time; Utilize the Internet of Things technology to receive the data collected by each edge data acquisition node in real time and establish a printed circuit board production energy consumption database; Based on the LSTM-Transformer-GNN hybrid machine learning algorithm, construct a printed circuit board production energy consumption evaluation model to comprehensively evaluate and predict the energy consumption in the production of printed circuit boards; According to the evaluation results of the printed circuit board production energy consumption evaluation model, obtain the weight values of the production energy consumption of each edge data acquisition node; Obtain the correlation of the production energy consumption of each edge data acquisition node, and through waste heat recovery optimization, dynamically adjust and optimize the energy consumption in the production of printed circuit boards.

[0017] It can be explained that in this solution, by deploying a multi-modal sensor array, synchronizing multi-dimensional data such as device operation status data and environmental parameters, and combining edge data acquisition nodes to achieve real-time data collection, a printed circuit board production energy consumption database covering all process links is constructed, providing a richer data basis for subsequent energy consumption analysis. At the same time, based on the LSTM-Transformer-GNN hybrid machine learning algorithm, a printed circuit board production energy consumption evaluation model is constructed to comprehensively evaluate and predict the energy consumption in the production of printed circuit boards, realizing in-depth mining and accurate prediction of energy consumption data, and significantly improving the reliability of the evaluation results. Secondly, according to the printed circuit board production energy consumption evaluation model, extract the weights in the LSTM-Transformer-GNN hybrid model respectively, and according to the weighted formula, obtain the comprehensive weight values of the production energy consumption of each edge data acquisition node, innovatively integrating the time series modeling ability of LSTM, the self-attention mechanism of Transformer and the spatial correlation analysis ability of GNN. Finally, through the comprehensive weight values and energy consumption correlation, combined with the heat-mass coupling characteristics of the waste heat recovery system, the real-time dynamic adjustment of the recovery start / stop and parameters is realized, enabling the system to adapt to the production schedule fluctuations, maximizing the waste heat utilization rate and reducing the total energy consumption, so as to achieve the purpose of dynamically evaluating the energy consumption in the production of printed circuit boards and optimizing waste heat recovery through multi-modal perception data and hybrid machine learning models, and significantly reducing the production capacity loss of printed circuit boards.

[0018] The deployment of a multi-modal sensor array in each production process of the printed circuit board, and the real-time collection of device operation status data and environmental parameters by setting edge data acquisition nodes specifically include: According to the production process of printed circuit boards, multi-modal sensor arrays are deployed in each production process to collect device operation status data and environmental parameters in real time during each production process; In each production process, edge data acquisition nodes are set up to perform filtering and normalization preprocessing on the device operation status data and environmental parameters collected in real time through the edge data acquisition nodes; Adopt the simultaneous acquisition mode to make each edge data acquisition node operate synchronously under the unified time reference, ensuring the time consistency and synchronization of the collected data; For each edge data acquisition node, a standardized communication format specification is uniformly set to ensure a high degree of consistency in the format and structure of the data uploaded by different nodes.

[0019] It can be explained that deploying multi-modal sensor arrays and edge data acquisition nodes in printed circuit board production and constructing a unified data acquisition and communication mechanism are the basis for constructing an efficient energy consumption management system. Its core significance lies in solving the analysis failure problems caused by one-sided data, noise interference, and heterogeneity in traditional solutions through full-dimensional data perception and standardized processing. At the same time, the simultaneous acquisition mode is adopted to ensure that each node operates synchronously under the unified time reference, avoiding data misalignment caused by clock drift or transmission delay, and ensuring time consistency, thus providing an accurate, real-time, and standardized data basis for printed circuit board production energy consumption management and being the key support for improving the energy consumption prediction accuracy and optimization effect.

[0020] The use of Internet of Things technology to receive the data collected by each edge data acquisition node in real time and establish a printed circuit board production energy consumption database specifically includes: Use Internet of Things technology to receive the data collected by each edge data acquisition node in real time; Apply interpolation and data downsampling techniques to process the data of each edge data acquisition node to achieve alignment and position consistency of the data in time or space; According to the process requirements of printed circuit board production, extract the boundary constraints of the data collected by each edge data acquisition node, thereby establishing a printed circuit board process parameter knowledge graph; Based on the historical data of printed circuit board production and the printed circuit board process parameter knowledge graph, establish a printed circuit board production energy consumption database.

[0021] It can be explained that in the energy consumption management of printed circuit board production, receiving the multi-source data of printed circuit board production from edge data acquisition nodes in real time through Internet of Things technology and the printed circuit board production energy consumption database are the core basis for supporting the accurate energy consumption assessment and optimization decision-making of printed circuit board production. Through interpolation and data downsampling techniques, the deviation problem corresponding to the data position can be effectively solved. Through the timestamps and spatial positions of the multi-source data of printed circuit board production, the consistency of data in the time and space dimensions is ensured, providing reliable input for subsequent analysis. On this basis, by extracting data boundary constraints in combination with the process requirements of printed circuit boards and constructing a knowledge graph of printed circuit board process parameters, the boundary constraints between parameters can be explicitly expressed, avoiding energy consumption anomalies caused by parameter overstepping, thereby improving the accuracy and reliability of data.

[0022] Refer to Figure 2 As shown, the energy consumption assessment model of printed circuit board production is constructed based on the LSTM-Transformer-GNN hybrid machine learning algorithm. The comprehensive assessment and prediction of the energy consumption of printed circuit board production specifically include: According to the production process cycle of printed circuit boards, set the window length of the LSTM layer time series, where the window length of the LSTM layer time series is within the range of the production process cycle of printed circuit boards; According to the printed circuit board production energy consumption database, use the data collected by each edge data acquisition node as the input quantity; Based on the LSTM neural network, output the device operation status data and environmental parameters with time series, and use them as the input quantity of the Transformer model; Through linear projection, map the data collected by each real-time edge data acquisition node to the same-dimensional space as Key / Value; According to the Transformer model, calculate the correlation between each data in the input quantity and other data, and capture the dependence relationship between any data in the input quantity data; According to the knowledge graph of printed circuit board process parameters, based on the mean square error formula, establish a constraint loss function of the LSTM-Transformer model; According to the LSTM-Transformer model, judge and predict the printed circuit board production energy consumption value and the total printed circuit board production energy consumption value of each edge data acquisition node within the window length; According to the production processes of each edge data acquisition node in the production process of printed circuit boards, based on the CNN neural network, adjust the edge weights in real time to accurately reflect the impact of production schedule changes on energy consumption, and further correct and optimize the printed circuit board production energy consumption value; The expression for setting the window length of the LSTM layer time series is: ; In the formula, is the window length of the LSTM layer time series, is the production process cycle of the printed circuit board, is the dominant angular frequency in the production data of the printed circuit board, is the floor function symbol to ensure that the window length is an integer time unit; The expression of the constraint loss function of the LSTM-Transformer model is: ; In the formula, is the value of the constraint loss function of the LSTM-Transformer model, is the number of sample groups, is the predicted value of the production energy consumption of the model printed circuit board, is the target value of the production energy consumption of the printed circuit board, is the weight coefficient with constraint terms in the knowledge graph of the printed circuit board process parameters, is the sample set with constraint terms in the knowledge graph of the printed circuit board process parameters, is the standard boundary threshold in the knowledge graph of the printed circuit board process parameters.

[0023] It can be explained that traditional LSTM or neural network models are difficult to balance the influence of time series fluctuations and spatial layout, resulting in a large deviation between the prediction result and the actual situation, problems such as insufficient prediction accuracy, and the disconnection of time series dependence and spatial correlation. Therefore, in this solution, the LSTM layer is used to extract the time series features of the equipment operation status and environmental parameters, which are used as the input of the Transformer model. The self-attention mechanism is used to capture any dependence relationship between data. At the same time, a constraint loss function is constructed in combination with the knowledge graph of the printed circuit board process parameters to ensure that the model output conforms to the process constraints, significantly improving the prediction accuracy. On this basis, the edge weights are dynamically adjusted through the CNN neural network, which can reflect the impact of production scheduling changes on energy consumption in real time, further correct the prediction result, and avoid the error accumulation caused by static model assumptions. Thus, the model can not only output the local energy consumption values of each edge node, but also comprehensively evaluate the total energy consumption of the entire process, providing data support for subsequent dynamic waste heat recovery optimization, and significantly improving the energy efficiency management level of printed circuit board production. Among them, the LSTM-Transformer hybrid model is used to process sensor time series data, predict the energy consumption at the process level in the future window length, realize the short-term prediction of the printed circuit board production energy consumption, and use the GNN neural network to model the energy consumption dependence relationship between processes. Combined with the production scheduling, predict the energy consumption trend of the entire batch, and realize the long-term prediction of the printed circuit board production energy consumption; The expression of the total printed circuit board production energy consumption value is: ; In the formula, is the total energy consumption value of printed circuit board production, is the number of edge data acquisition nodes, is the th energy consumption value of printed circuit board production of the edge data acquisition node.

[0024] Referring to Figure 3 shown, obtaining the weight value of the production energy consumption of each edge data acquisition node according to the evaluation result of the printed circuit board production energy consumption evaluation model specifically includes: According to the evaluation result of the printed circuit board production energy consumption evaluation model, obtain the printed circuit board production energy consumption values of each edge data acquisition node within the window length; According to the printed circuit board production energy consumption evaluation model, extract the weights in the LSTM-Transformer-GNN hybrid model respectively; Among them, LSTM is the historical data weight of each edge data acquisition node, Transformer is the self-attention weight of each edge data acquisition node, and GNN is the spatial weight of each edge data acquisition node; According to the weighted formula, obtain the comprehensive weight value of the production energy consumption of each edge data acquisition node; According to the Softmax normalization formula, normalize the comprehensive weight value of the production energy consumption of each edge data acquisition node so that the sum of all comprehensive weight values is 1; The expression of the comprehensive weight value of the production energy consumption of each edge data acquisition node is: ; In the formula, is the comprehensive weight value of the production energy consumption of the edge data acquisition node, , , are respectively the historical data weight, self-attention weight, and spatial weight of the edge data acquisition node, , , are respectively the emphasis coefficients of the historical data weight, self-attention weight, and spatial weight of the edge data acquisition node, which can be determined by cross-validation.

[0025] It can be explained that the production of printed circuit boards involves the coordination of multiple devices and the coupling of complex processes. The energy consumption contributions of different edge data collection nodes are significantly affected by historical operation rules, real-time correlation relationships, and spatial layouts. Due to the multidimensionality and complexity of the data collected by each edge data collection node, the difficulty of accurately positioning the key energy consumption in printed circuit board production increases. To solve the problems of fuzzy node contribution degrees and scattered optimization objectives in traditional energy consumption analysis, this solution extracts the historical data weights of each edge data collection node through LSTM, extracts the self-attention weights of each edge data collection node through Transformer, and extracts the spatial weights of each edge data collection node through GNN. Thus, through the three types of weight weighting methods of historical data weights, self-attention weights, and spatial weights, the temporal dependence, data correlation, and spatial correlation are fused to obtain a comprehensive weight value. By quantifying the energy consumption priorities of nodes, such as high-weight nodes participating in waste heat recovery first, the accurate quantification of the energy consumption contribution in printed circuit board production is realized, which is the key support for constructing an energy consumption optimization closed-loop and improving the fineness of energy efficiency management, and provides an analysis of the data correlation for subsequent waste heat recovery optimization.

[0026] Referring to Figure 4 As shown, the obtaining of the correlation of the production energy consumption of each edge data collection node and the dynamic adjustment and optimization of the printed circuit board production energy consumption through waste heat recovery specifically include: According to the printed circuit board production energy consumption database, obtain the device operation status data and environmental parameters of each edge data collection node; Based on the device operation status data and environmental parameters of each edge data collection node and the production energy consumption data of each edge data collection node, determine the correlation of the production energy consumption of each edge data collection node based on the Pearson correlation formula; According to the correlation of the production energy consumption of each edge data collection node and the correlation coupling of each node, establish a dynamic adjustment and optimization model for the printed circuit board production energy consumption; According to the dynamic adjustment and optimization model of the printed circuit board production energy consumption, dynamically adjust the start and stop of waste heat recovery and the printed circuit board production energy consumption.

[0027] It can be explained that due to the inability of traditional static waste heat recovery schemes to perceive the energy consumption coupling relationship between nodes, there are often problems such as insufficient recovery or waste of heat sources. In the energy consumption management of PCB production, quantifying the energy consumption correlation of each edge data collection node and dynamically adjusting the waste heat recovery strategy is the key breakthrough to break through the isolation of traditional energy efficiency optimization means and achieve full-process collaborative energy conservation. This solution establishes a dynamic adjustment and optimization model for the printed circuit board production energy consumption through data correlation and energy consumption correlation, and dynamically adjusts the start and stop of waste heat recovery and the printed circuit board production energy consumption through the model. Among them, the conditional formula for the dynamic start and stop of waste heat recovery is: The start condition formula is: ; The stop condition formula is: ; In the formula, is the comprehensive weight value of the production energy consumption of the edge data acquisition node, is the relevance of the production energy consumption of each edge data acquisition node and the correlation coupling relevance of each node, and are respectively the comprehensive weight value of the production energy consumption of the edge data acquisition node and the boundary threshold of the relevance of the production energy consumption of each edge data acquisition node and the correlation coupling relevance of each node, is the predicted recovery efficiency of the waste heat of the printed circuit board production energy consumption, is the minimum value of the predicted recovery efficiency of the waste heat of the printed circuit board production energy consumption, is the real-time recovery efficiency of the waste heat of the printed circuit board production energy consumption; According to the data relevance and energy consumption relevance, obtain the maximum recovery efficiency of the waste heat of the printed circuit board production energy consumption. Among them, the expression of the maximum recovery efficiency of the waste heat of the printed circuit board production energy consumption is: ; In the formula, is the waste heat recovery flow rate from the th edge data acquisition node to the th edge data acquisition node, is the maximum value of the waste heat recovery flow rate of the edge data acquisition node, , are respectively two nodes of the waste heat supply-demand relationship in the printed circuit board production process. The collection of all waste heat recovery paths during the printed circuit board production process is referred to as the recovery pair, is the comprehensive weight value of the production energy consumption of the th edge data acquisition node, is the relevance of the production energy consumption of the th node and the th node of the waste heat supply-demand relationship and the correlation coupling relevance of each node; According to the maximum recovery efficiency of the waste heat of the printed circuit board production energy consumption, establish a dynamic adjustment and optimization model for the printed circuit board production energy consumption. Among them, the expression of the dynamic adjustment and optimization model for the printed circuit board production energy consumption is: ; In the formula, is the minimum energy consumption of the printed circuit board production, is the total printed circuit board production energy consumption value, is the recovery efficiency and is a constant term, is the hot melt ratio, is the outlet temperature of the waste heat recovery of the energy consumption in printed circuit board production, is the required temperature of the waste heat recovery node of the energy consumption in printed circuit board production.

[0028] Furthermore, based on the same inventive concept as the above-mentioned real-time prediction and optimization method for the energy consumption in printed circuit board production, this solution proposes a real-time prediction and optimization system for the energy consumption in printed circuit board production, including: A data acquisition module, which is used to deploy a multi-modal sensor array in each production process of the printed circuit board, and by setting edge data acquisition nodes, to collect the equipment operation status data and environmental parameters in real time; using the Internet of Things technology, to receive the data collected by each edge data acquisition node in real time, and establish a database for the energy consumption in printed circuit board production; A prediction and optimization module, which is used to construct an energy consumption evaluation model for printed circuit board production based on the LSTM-Transformer-GNN hybrid machine learning algorithm, comprehensively evaluate and predict the energy consumption in printed circuit board production; according to the evaluation results of the energy consumption evaluation model for printed circuit board production, obtain the weight values of the energy consumption of each edge data acquisition node; obtain the correlation of the energy consumption of each edge data acquisition node, and dynamically adjust and optimize the energy consumption in printed circuit board production through waste heat recovery optimization; The data acquisition module includes: A data acquisition unit, which is used to deploy a multi-modal sensor array in each production process of the printed circuit board, and by setting edge data acquisition nodes, to collect the equipment operation status data and environmental parameters in real time; A data processing unit, which is used to use the Internet of Things technology to receive the data collected by each edge data acquisition node in real time, and establish a database for the energy consumption in printed circuit board production; The prediction and optimization module includes: An evaluation and prediction unit, which is used to construct an energy consumption evaluation model for printed circuit board production based on the LSTM-Transformer-GNN hybrid machine learning algorithm, comprehensively evaluate and predict the energy consumption in printed circuit board production; A node weight unit, which is used to obtain the weight values of the energy consumption of each edge data acquisition node according to the evaluation results of the energy consumption evaluation model for printed circuit board production; An energy consumption adjustment unit, which is used to obtain the correlation of the energy consumption of each edge data acquisition node, and dynamically adjust and optimize the energy consumption in printed circuit board production through waste heat recovery optimization.

[0029] In summary, the advantages of the present invention are: through multi-modal perception data and a hybrid machine learning model, dynamically evaluate the energy consumption in printed circuit board production and optimize waste heat recovery, significantly reducing the production capacity loss of printed circuit boards.

[0030] The foregoing has shown and described the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments, and what is described in the above embodiments and the specification is only the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements fall within the scope of the present invention claimed. The scope of protection required by the present invention is defined by the appended claims and their equivalents.

Claims

1. A real-time prediction and optimization method for the energy consumption in printed circuit board production, characterized in that, Including: Deploy a multi-modal sensor array in each production process of the printed circuit board. By setting up edge data acquisition nodes, collect the device operation status data and environmental parameters in real time; Utilize the Internet of Things technology to receive the data collected by each edge data acquisition node in real time and establish a printed circuit board production energy consumption database; Based on the LSTM-Transformer-GNN hybrid machine learning algorithm, construct a printed circuit board production energy consumption evaluation model to comprehensively evaluate and predict the printed circuit board production energy consumption; According to the evaluation results of the printed circuit board production energy consumption evaluation model, obtain the weight values of the production energy consumption of each edge data acquisition node; Obtain the correlation of the production energy consumption of each edge data acquisition node, and through waste heat recovery optimization, dynamically adjust and optimize the printed circuit board production energy consumption.

2. The real-time prediction optimization method for the energy consumption in the production of printed circuit boards according to claim 1, wherein The deployment of the multi-modal sensor array in each production process of the printed circuit board, and the real-time collection of the device operation status data and environmental parameters by setting up edge data acquisition nodes specifically include: According to the production process of the printed circuit board, deploy a multi-modal sensor array in each production process to collect the device operation status data and environmental parameters in each production process in real time; In each production process, set up edge data acquisition nodes, and perform filtering and normalization preprocessing on the device operation status data and environmental parameters collected in real time by the edge data acquisition nodes; Adopt a simultaneous time series acquisition mode to make each edge data acquisition node operate synchronously under a unified time reference to ensure the time consistency and synchronization of the collected data; For each edge data acquisition node, uniformly set the standardized communication format specification to ensure a high degree of consistency in the format and structure of the data uploaded by different nodes.

3. A real-time prediction and optimization method for the energy consumption in the production of printed circuit boards according to claim 2, characterized in that, The utilization of the Internet of Things technology to receive the data collected by each edge data acquisition node in real time and establish a printed circuit board production energy consumption database specifically includes: Utilize the Internet of Things technology to receive the data collected by each edge data acquisition node in real time; Apply interpolation and data downsampling technologies to process the data of each edge data acquisition node to achieve alignment and position consistency of the data in time or space; According to the process requirements of printed circuit board production, extract the boundary constraints of the data collected by each edge data acquisition node to establish a printed circuit board process parameter knowledge graph; Based on the historical data of printed circuit board production and the printed circuit board process parameter knowledge graph, establish a printed circuit board production energy consumption database.

4. A real-time prediction and optimization method for the energy consumption in the production of printed circuit boards according to claim 3, wherein, The construction of a printed circuit board production energy consumption evaluation model based on the LSTM-Transformer-GNN hybrid machine learning algorithm to comprehensively evaluate and predict the printed circuit board production energy consumption specifically includes: According to the production process cycle of the printed circuit board, set the window length of the LSTM layer time series, where the window length of the LSTM layer time series is within the range of the printed circuit board production process cycle; According to the printed circuit board production energy consumption database, use the data collected by each edge data acquisition node as the input quantity; Based on the LSTM neural network, output the device operation status data and environmental parameters with time series and use them as the input quantity of the Transformer model; Through linear projection, map the data collected by each real-time edge data acquisition node to the same-dimensional space as Key / Value; According to the Transformer model, calculate the correlation between each data in the input quantity and other data, and capture the dependency relationship between any data in the input quantity data; According to the printed circuit board process parameter knowledge graph, based on the mean square error formula, establish a constraint loss function for the LSTM-Transformer model; According to the LSTM-Transformer model, judge and predict the printed circuit board production energy consumption value and the total printed circuit board production energy consumption value of each edge data acquisition node within the window length; According to the production processes of each edge data acquisition node in the production process of the printed circuit board, based on the CNN neural network, adjust the edge weights in real time to accurately reflect the impact of production scheduling changes on energy consumption, and further correct and optimize the printed circuit board production energy consumption value; The expression of the window length of the LSTM layer timing is as follows: ; In the formula, is the window length of the LSTM layer time series, is the production process cycle of the printed circuit board, is the dominant angular frequency in the production data of the printed circuit board, is the floor function symbol to ensure that the window length is an integer time unit; The expression of the constraint loss function of the LSTM-Transformer model is as follows: ; In the formula, is the value of the constraint loss function of the LSTM-Transformer model, is the number of sample groups, is the predicted value of the energy consumption for printed circuit board production by the model, is the target value of the energy consumption for printed circuit board production, is the weight coefficient with constraint terms in the knowledge graph of printed circuit board process parameters, is the sample set with constraint terms in the knowledge graph of printed circuit board process parameters, is the standard boundary threshold in the knowledge graph of printed circuit board process parameters.

5. A real-time prediction and optimization method for the energy consumption in the production of printed circuit boards according to claim 4, characterized in that, The specific steps of obtaining the weight value of the production energy consumption of each edge data acquisition node according to the evaluation result of the printed circuit board production energy consumption evaluation model include: According to the evaluation result of the printed circuit board production energy consumption evaluation model, obtain the printed circuit board production energy consumption value of each edge data acquisition node within the window length; According to the printed circuit board production energy consumption evaluation model, extract the weights in the LSTM-Transformer-GNN hybrid model respectively; Among them, LSTM is the historical data weight of each edge data acquisition node, Transformer is the self-attention weight of each edge data acquisition node, and GNN is the spatial weight of each edge data acquisition node; According to the weighted formula, obtain the comprehensive weight value of the production energy consumption of each edge data acquisition node; According to the Softmax normalization formula, normalize the comprehensive weight value of the production energy consumption of each edge data acquisition node so that the sum of all comprehensive weight values is 1; The expression of the comprehensive weight value of the production energy consumption of each edge data acquisition node is as follows: ; Wherein, is the comprehensive weight value of the production energy consumption of the edge data acquisition node, , , are respectively the historical data weight of the edge data acquisition node, the self-attention weight of the edge data acquisition node, and the spatial weight of the edge data acquisition node, , , are respectively the bias coefficients of the historical data weight, the self-attention weight, and the spatial weight of the edge data acquisition node, and can be determined by cross-validation.

6. A real-time prediction and optimization method for the energy consumption in printed circuit board production according to claim 5, characterized in that, The specific steps of obtaining the relevance of the production energy consumption of each edge data acquisition node and dynamically adjusting and optimizing the printed circuit board production energy consumption through waste heat recovery optimization include: According to the printed circuit board production energy consumption database, obtain the equipment operation status data and environmental parameters of each edge data acquisition node; According to the equipment operation status data and environmental parameters of each edge data acquisition node and the production energy consumption data of each edge data acquisition node, based on the Pearson correlation formula, determine the relevance of the production energy consumption of each edge data acquisition node; According to the relevance of the production energy consumption of each edge data acquisition node and the correlation coupling relevance of each node, establish a dynamic adjustment and optimization model for the printed circuit board production energy consumption; According to the dynamic adjustment and optimization model of the printed circuit board production energy consumption, dynamically adjust the start / stop and parameters of waste heat recovery.

7. A real-time prediction and optimization system for the energy consumption in printed circuit board production, characterized in that, Used to implement the real-time prediction and optimization method for the printed circuit board production energy consumption as described in any one of claims 1-6, including: A data acquisition module, which is used to deploy a multi-modal sensor array in each production process of the printed circuit board, and by setting edge data acquisition nodes, it can collect device operation status data and environmental parameters in real time; using Internet of Things technology, it can receive the data collected by each edge data acquisition node in real time and establish a printed circuit board production energy consumption database; A prediction and optimization module, which is used to build a printed circuit board production energy consumption evaluation model based on the LSTM-Transformer-GNN hybrid machine learning algorithm, comprehensively evaluate and predict the printed circuit board production energy consumption; according to the evaluation results of the printed circuit board production energy consumption evaluation model, obtain the weight values of the production energy consumption of each edge data acquisition node; obtain the correlation of the production energy consumption of each edge data acquisition node, and dynamically adjust and optimize the printed circuit board production energy consumption through waste heat recovery optimization.

8. A real-time prediction and optimization system for the energy consumption in the production of printed circuit boards according to claim 7, characterized in that, The data acquisition module includes: A data acquisition unit, which is used to deploy a multi-modal sensor array in each production process of the printed circuit board, and by setting edge data acquisition nodes, it can collect device operation status data and environmental parameters in real time; A data processing unit, which is used to use Internet of Things technology to receive the data collected by each edge data acquisition node in real time and establish a printed circuit board production energy consumption database.

9. The real-time prediction and optimization system for the energy consumption in the production of printed circuit boards according to claim 8, characterized in that, The prediction and optimization module includes: An evaluation and prediction unit, which is used to build a printed circuit board production energy consumption evaluation model based on the LSTM-Transformer-GNN hybrid machine learning algorithm, and comprehensively evaluate and predict the printed circuit board production energy consumption; A node weight unit, which is used to obtain the weight values of the production energy consumption of each edge data acquisition node according to the evaluation results of the printed circuit board production energy consumption evaluation model; An energy consumption adjustment unit, which is used to obtain the correlation of the production energy consumption of each edge data acquisition node, and dynamically adjust and optimize the printed circuit board production energy consumption through waste heat recovery optimization.

Images