Manufacturing scene constraint-based product carbon footprint uncertainty analysis method and system

Abstract

The invention relates to the technical field of carbon footprint uncertainty analysis, and discloses a manufacturing scene constraint-based product carbon footprint uncertainty analysis method and system, and the method comprises the steps: carrying out the noise elimination of an initial sample set, and obtaining a to-be-predicted sample set after the noise elimination; inputting the to-be-predicted sample set after noise elimination into a carbon footprint regression prediction model to obtain a carbon footprint prediction value corresponding to each to-be-predicted sample in the to-be-predicted sample set after noise elimination; calculating a discrete coefficient according to the carbon footprint predicted value; if the discrete coefficient is higher than a set threshold value, parameters of the current to-be-predicted sample need to be modified; after all to-be-predicted samples with parameters needing to be modified are obtained, key parameters are screened out for each to-be-predicted sample, the key parameters are modified, modified to-be-predicted samples are obtained, then a modified to-be-predicted sample set is obtained, uncertainty analysis is carried out based on manufacturing scene information and constraints, and the reasonability and accuracy of results are guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of carbon footprint uncertainty analysis in the processing and manufacturing stage of electromechanical products, and in particular to a product carbon footprint uncertainty analysis method and system based on manufacturing scenario constraints. Background Art

[0002] Carbon emissions from the manufacturing industry have a significant impact on global warming and climate change. The accuracy of carbon emission results in the manufacturing stage plays a decisive role in subsequent optimization decisions. However, the wide range of data sources and complex information associations in the manufacturing stage make the credibility of carbon emission results questionable, which may lead to wrong decisions. Therefore, it is urgent to carry out uncertainty analysis of carbon footprint in the manufacturing stage to improve the credibility of the assessment results. Uncertainty analysis of carbon footprint in the manufacturing stage includes two key contents: (1) comprehensive identification and expression of information related to carbon footprint in the manufacturing stage; (2) reasonable uncertainty analysis methods. However, existing studies have mostly focused on the influencing factors directly related to carbon emissions in the manufacturing stage. Some factors that have a greater impact on carbon emissions have not been considered (such as energy type, geographical location, etc.), and the complex correlation constraints between information lack analysis and expression, which seriously affects the accuracy of carbon emission assessment results. In addition, uncertainty analysis methods focus on analyzing the trends and laws of output results, ignoring the accuracy of input information. Summary of the invention

[0003] In order to address the shortcomings of the prior art, the present invention provides a product carbon footprint uncertainty analysis method and system based on manufacturing scenario constraints; identifies and expresses manufacturing stage information and associated constraints, and conducts uncertainty analysis based on manufacturing scenario information and constraints to ensure the rationality and accuracy of the results.

[0004] On the one hand, a product carbon footprint uncertainty analysis method based on manufacturing scenario constraints is provided, including: (1): Obtain a set of samples to be predicted, wherein the set of samples to be predicted includes: a plurality of samples to be predicted, each sample to be predicted includes: a plurality of attribute parameters to be predicted in a manufacturing scenario; randomly sample the set of samples to be predicted to obtain an initial sample set; remove noise from the initial sample set to obtain a set of samples to be predicted after the noise is removed; (2): Input the sample set to be predicted after removing the noise into the carbon footprint regression prediction model to obtain the carbon footprint prediction value corresponding to each sample to be predicted in the sample set to be predicted after removing the noise; (3): Calculate the dispersion coefficient based on the carbon footprint prediction value; if the dispersion coefficient is higher than the set threshold, it means that the parameters of the current sample to be predicted need to be modified; after obtaining all the samples to be predicted whose parameters need to be modified, enter (4); if the dispersion coefficient is lower than the set threshold, it means that the parameters of the current sample to be predicted do not need to be modified; (4): Filter out key parameters for each sample to be predicted, modify the key parameters to obtain the modified sample to be predicted, and then obtain the modified set of samples to be predicted, and return to (1).

[0005] On the other hand, a product carbon footprint uncertainty analysis system based on manufacturing scenario constraints is provided, including: The acquisition module is configured to: acquire a set of samples to be predicted, the set of samples to be predicted comprising: a plurality of samples to be predicted, each sample to be predicted comprising: a plurality of attribute parameters to be predicted in a manufacturing scenario; perform random sampling on the set of samples to be predicted to obtain an initial sample set; perform noise elimination on the initial sample set to obtain a set of samples to be predicted after the noise is eliminated; The prediction module is configured to: input the set of samples to be predicted after noise removal into the carbon footprint regression prediction model to obtain the carbon footprint prediction value corresponding to each sample to be predicted in the set of samples to be predicted after noise removal; The calculation module is configured to: calculate the dispersion coefficient according to the carbon footprint prediction value; if the dispersion coefficient is higher than the set threshold, it means that the parameters of the current sample to be predicted need to be modified; after obtaining all the samples to be predicted whose parameters need to be modified, enter the modification module; if the dispersion coefficient is lower than the set threshold, it means that the parameters of the current sample to be predicted do not need to be modified; The modification module is configured to: screen out key parameters for each sample to be predicted, modify the key parameters to obtain the modified sample to be predicted, and then obtain the modified set of samples to be predicted, and return to the acquisition module.

[0006] The above technical solution has the following advantages or beneficial effects: The activity characteristics of the manufacturing stage are comprehensively analyzed, the concept of manufacturing scenarios is proposed, and the attribute information of each manufacturing scenario and the associated constraints between the information are identified and analyzed to achieve the standardized expression of the carbon footprint related information and constraints in the manufacturing stage. The random forest algorithm is used to quickly generate a prediction model for carbon footprint. Based on the distribution function and associated constraints of the manufacturing scenario attribute information, sample data is generated through Monte Carlo analysis, and the noise of the sample data is screened based on the comprehensive ranking method of the importance of the manufacturing scenario attribute parameters. The completion time, cost, environment and constraint indicators of the manufacturing process are comprehensively considered to judge the priority of the solution. The constraint satisfaction problem is solved according to the priority order, and the initial sample set that meets the constraints is efficiently screened to ensure the rationality and accuracy of the input sample data. Furthermore, the uncertainty analysis of the carbon footprint results is carried out through the discrete coefficient to identify the key parameters that have a greater impact on the carbon emissions in the manufacturing stage. By improving and enhancing the key parameters, the uncertainty of carbon emissions is reduced and the credibility of the results is improved.

[0007] The present invention proposes the concept of manufacturing scenario to comprehensively express carbon footprint related information, and identifies and defines the manufacturing scenario attribute information and its associated constraints, so as to achieve a comprehensive and accurate description of the manufacturing stage information. The carbon footprint result is predicted by the prediction model of the random forest algorithm. Furthermore, based on the distribution function of the manufacturing scenario attribute information and its associated constraints, the sample data noise screening priority is defined according to the comprehensive ranking of the manufacturing scenario attribute parameters’ importance, the constraint satisfaction problem is solved according to the priority, and the samples that do not meet the requirements are screened and eliminated to ensure the rationality of the uncertain analysis input variables. The uncertainty of the carbon footprint result is analyzed by the discrete coefficient, and the credibility of the result is improved by improving and enhancing the key parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0009] Figure 1 This is a flow chart of the method of embodiment 1. DETAILED DESCRIPTION

[0010] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0011] Processing and manufacturing is the main method for forming products or parts, and its carbon emission assessment results are of great significance to subsequent optimization decisions and carbon reduction and emission reduction. However, the wide range of data sources, uncontrollable collection errors, and simplified assessment models in the manufacturing stage have led to doubts about the credibility of carbon emission results in this stage, which may lead to wrong decisions. Therefore, it is urgent to develop uncertainty analysis methods in the manufacturing stage to identify key parameters that have a greater impact on carbon emissions in the manufacturing stage, improve the credibility of the results through the improvement and enhancement of key parameters, and promote the low-carbon development of the manufacturing industry.

[0012] Uncertainty analysis can identify the impact of information data on product carbon emissions, effectively estimate the change mode and trend of these data, and then trace back and guide design. It is a hot topic in carbon emission assessment research. Through information uncertainty analysis, the credibility and robustness of carbon emission results can be effectively evaluated. Uncertainty analysis of carbon footprint in the manufacturing stage includes two key contents: (1) comprehensive identification and expression of information related to carbon footprint in the manufacturing stage; (2) reasonable uncertainty analysis methods. In terms of information identification and expression, existing research focuses on the influencing factors directly related to carbon emissions in the manufacturing stage, such as processing power and processing time. Some factors that have a greater impact on carbon emissions have not been considered (for example, energy type, geographical location, etc.). In addition, there are complex correlations and constraints between various information, and research on the expression of information correlation constraints has not been carried out, which restricts the tracing of information and seriously affects the accuracy of carbon emission assessment. In the analysis and processing of design information uncertainty, probability theory, fuzzy theory and grey system theory have been widely used, mainly through Monte Carlo analysis, multivariate regression analysis and other methods. Uncertainty results are mainly measured by the changes in input information and output results. Different manufacturing scenarios correspond to different carbon emission results. The rationality and accuracy of input variables have a decisive influence on the results of uncertainty analysis. However, existing uncertainty analysis focuses on the analysis and improvement of output results, lacks research on the rationality and accuracy of input information, and ignores the complex correlation constraints between manufacturing scene attribute information in the manufacturing process, making it difficult to ensure the rationality and accuracy of uncertainty analysis results.

[0013] Embodiment 1, as Figure 1 As shown, this embodiment provides a product carbon footprint uncertainty analysis method based on manufacturing scenario constraints, including: S101: Obtain a set of samples to be predicted, the set of samples to be predicted including: a plurality of samples to be predicted, each sample to be predicted including: a plurality of attribute parameters to be predicted in a manufacturing scenario; randomly sample the set of samples to be predicted to obtain an initial sample set; remove noise from the initial sample set to obtain a set of samples to be predicted after the noise is removed; S102: Inputting the set of samples to be predicted after noise removal into the carbon footprint regression prediction model to obtain the carbon footprint prediction value corresponding to each sample to be predicted in the set of samples to be predicted after noise removal; S103: Calculate the dispersion coefficient according to the carbon footprint prediction value; if the dispersion coefficient is higher than the set threshold, it means that the parameters of the current sample to be predicted need to be modified; after obtaining all the samples to be predicted whose parameters need to be modified, enter S104; if the dispersion coefficient is lower than the set threshold, it means that the parameters of the current sample to be predicted do not need to be modified; S104: Filter out key parameters for each sample to be predicted, modify the key parameters to obtain modified samples to be predicted, and then obtain a modified set of samples to be predicted, and return to S101.

[0014] Furthermore, the S101: randomly sampling the sample set to be predicted to obtain an initial sample set includes: (1-1): Confirm the probability distribution function of the manufacturing scene attribute parameters in the sample set to be predicted; (1-2): According to the probability distribution function of the manufacturing scene attribute parameters, Monte Carlo analysis is used to randomly sample the manufacturing scene attribute parameters to obtain an initial sample data set.

[0015] Furthermore, the (1-1): confirming the probability distribution function of the manufacturing scene attribute parameters in the sample set to be predicted, the probability distribution function includes: a uniform distribution function of discrete variables and a normal distribution function of continuous variable pairs.

[0016] Furthermore, the above (1-2): based on the probability distribution function of the manufacturing scene attribute parameters, Monte Carlo analysis is used to randomly sample the manufacturing scene attribute parameters to obtain an initial sample data set, including: If the attribute parameter conforms to the normal distribution, the random value is obtained according to the normal distribution; If the attribute parameter conforms to the uniform distribution, the value is randomly selected according to the uniform distribution.

[0017] It should be understood that based on the theoretical foundation of probability theory, the characteristics of various attribute information in the manufacturing scenario are deeply analyzed, and the probability distribution functions of these attributes are derived through mathematical statistical fitting of the data.

[0018] These manufacturing scenario attributes can be divided into discrete and continuous types, and the distribution patterns include uniform distribution, normal distribution, etc.

[0019] Among them, the process type, geographical location, energy type, and equipment type in different manufacturing scenario attributes are discrete variables, and their distribution functions conform to the discrete uniform distribution: , ; in, is the input discrete variable attribute parameter that conforms to the uniform distribution, is the minimum value of the parameter, is the maximum value of the parameter.

[0020] The quantitative data of each manufacturing scenario, including processing power, processing time, etc., mainly comes from the error between the design information data and the actual processing data. The distribution of these continuous variables conforms to the normal distribution:

[0021] in The continuous variable attribute parameters that conform to the normal distribution are input.

[0022] Using Monte Carlo analysis, according to the established probability distribution function of manufacturing scenario attribute information, random sampling of input parameters is performed to determine the number of iterations N of Monte Carlo simulation, and to generate a large number of input parameters that meet the specified requirements. .

[0023] Furthermore, the step S101: removing noise from the initial sample set to obtain a sample set to be predicted after the noise is removed, includes: (2-1): For each sample to be predicted, the comprehensive weight of each attribute parameter to be predicted is calculated; (2-2): Sort all attribute parameters to be predicted in descending order according to their comprehensive weights, and determine whether each attribute parameter to be predicted meets the associated constraints in descending order according to their weights; (2-3): If the association constraint is met, continue to judge the next parameter; (2-4): If the association constraint is not met, the current sample to be predicted is deleted from the set of samples to be predicted, and the set of samples to be predicted is obtained after the noise is removed.

[0024] Furthermore, in (2-1): for each sample to be predicted, a comprehensive weight of each attribute parameter to be predicted is calculated, including: Comprehensive weights of manufacturing scenario attribute parameters for: .

[0025] in, Indicates The importance of attribute parameters to completion time, Indicates The importance of each attribute parameter to the cost, Indicates The importance of each attribute parameter to the environment. Indicates The constraint importance of each attribute parameter.

[0026] Among them, the The importance of attribute parameters to completion time , is to The scoring matrix of the completion time of each attribute parameter is input into the hierarchical analysis method; for the time target, the 1-9 scaling method is used to compare the scene attribute parameters in pairs. Among them, 1 means that the two parameters are equally important, 3 means that one parameter is slightly more important than the other parameter, 5 means that one parameter is obviously more important than the other parameter, 7 means that one parameter is strongly more important than the other parameter, and 9 means that one parameter is extremely important than the other parameter; 2, 4, 6, and 8 are the intermediate values ​​of the above adjacent judgments. The judgment matrix is ​​obtained by the above method. , Represents attribute parameters Relative attribute parameters For example, the following judgment matrix P can be obtained:

[0027] No. The importance of attribute parameters to cost , is to The scoring matrix of the cost of each attribute parameter is input into the hierarchical analysis method; the matrix acquisition method is consistent with the time target.

[0028] No. The importance of the attribute parameter to the environment is The scoring matrix of the environment based on the attribute parameters is input into the hierarchical analysis method; the matrix acquisition method is consistent with the time target.

[0029] No. By analyzing the number of constraints on attribute parameters of each manufacturing scenario, the constraints are applied to the The number of constraints on the attribute parameters of a manufacturing scenario is , then The constraint importance of the attribute parameters of the manufacturing scene is .

[0030] Furthermore, the association constraints include: range constraints , follow-up constraints and combination constraints .

[0031] Range Constraints , refers to the range of each attribute parameter limited by the external environment in order to complete the specified processing task, which is generally expressed in the form of an inequality, as shown in the following formula. For example, in order to meet the delivery requirements, the processing time range is restricted.

[0032] Range Constraints , whose expression is: ; In the formula, It is i Scene attribute parameter information; It refers to the minimum threshold of attribute information; Refers to the maximum threshold of attribute information.

[0033] Sequential Constraints , means that there is a certain sequential relationship between the scene attribute parameters. When one parameter is determined, the range of another parameter is also determined. For example, when the device type of the device attribute is determined, the processing power range of the parameter attribute is also determined.

[0034] Sequential Constraints , whose expression is:

[0035] In the formula, , They represent the attribute parameters of the manufacturing scene respectively, and [a, b] is the parameter range corresponding to the sequential constraint.

[0036] Combination Constraints , means that there are two or more manufacturing scenario attribute parameters that jointly determine the value of another attribute parameter. For example, the energy type and emission coefficient of the energy attribute and the geographical location of the description attribute. When both the energy type and geographical location are determined, the corresponding emission coefficient value can be obtained.

[0037] Combination Constraints , whose expression is:

[0038] In the formula, , and They respectively represent the attribute parameters of the manufacturing scene.

[0039] The present invention converts the association constraints between manufacturing scene attributes into a constraint satisfaction problem, and implements the screening of sample data by solving the constraint satisfaction problem.

[0040] The constraint satisfaction problem mainly consists of a set of variables, the optional value range of each variable and a finite set of constraints on the variables, which is mainly represented by a triple model, namely ,in: , The variable representing the problem, is the number of variables. ,in Representation variables Optional value domain. , Represents constraints between variables.

[0041] For each sample, the variable To create scene attribute parameters ; The value range D is the value of each manufacturing scenario attribute parameter; Constraint C is the constructed constraint model , , .

[0042] It should be understood that in order to quickly and efficiently screen out samples that do not meet the constraints, a comprehensive ranking method for the importance of manufacturing scenario attribute parameters is proposed to judge and solve the constraint solving priority. This method is oriented to the attribute parameters of the manufacturing scenario, comprehensively analyzes the constraints between the parameters and the impact of each parameter on the current production, and conducts research based on the importance of the attribute parameters.

[0043] During the product processing and manufacturing process, enterprises comprehensively consider delivery time, benefits and environmental issues, and set three main production goals, including completion time Time, cost Cost and environment Envi. Among them, completion time refers to the total time required for a product or component to reach the predetermined quality standard from the start of processing and manufacturing to the completion of all processes. Cost refers to the total cost required to manufacture one or more parts in the production process, including various expenses such as raw material procurement, labor input, equipment depreciation, energy consumption, management expenses and quality control. The environment mainly refers to the environmental impact caused by the total amount of greenhouse gases such as carbon dioxide released directly or indirectly into the atmosphere due to activities such as energy consumption, raw material use, and waste disposal in the process of producing parts. It reflects the contribution of the production process to climate change. During the processing and manufacturing process, the manufacturing scenario attribute parameters have different degrees of influence on the production goals. Aiming at the manufacturing scenario attribute parameters and production goals, three judgment matrices for completion time, cost and environment are constructed respectively through the hierarchical analysis method. The importance of each attribute parameter to each goal is calculated, that is, the weight belonging to each goal: , , .in ; . Indicates Manufacturing scene attribute parameters; is the number of manufacturing attribute parameters; , , Respectively represent The weights of the manufacturing scenario attribute parameters on the completion time, cost and environment.

[0044] In addition, the constraints between the attribute parameters of each manufacturing scenario also have an important impact on the production target. Based on the constructed constraint model, the constraints of the attribute parameters of each manufacturing scenario are analyzed and applied to the first The number of constraints on the attribute parameters of a manufacturing scenario is Ni, then The constraint weight of the attribute parameters of the manufacturing scene is .

[0045] The weights of all manufacturing scene attribute parameters are calculated and sorted by the above method. The higher the ranking, the higher the weight. The larger the value, the more parameters The more important it is to production, the higher the priority of its solution. Prioritize the nodes with greater comprehensive weights in the input parameters, solve the constraint problems of the parameters in each sample, check whether the constraints meet the defined constraint conditions, and reduce the complexity of the solution. Specific process: prioritize the variables in each sample data, start the constraint check for the variables with greater comprehensive weights, trace back to the associated constraints through positive tests, and determine whether the constraint requirements are met. If a sample does not meet a constraint condition, remove it from the sample. Deletion from a collection can be achieved using the deletion operation of the data structure, such as using the remove method in a list or by rebuilding a new list containing only samples that meet the constraints.

[0046] After screening, a sample set that meets the manufacturing scenario constraints is obtained , Calculate the corresponding carbon footprint value based on the prediction model obtained by random forest training The sample and its carbon footprint value form a new data pair .

[0047] Further, S102: input the set of samples to be predicted after the noise is removed into the carbon footprint regression prediction model to obtain the carbon footprint prediction value corresponding to each sample to be predicted in the set of samples to be predicted after the noise is removed; wherein the carbon footprint regression prediction model, the acquisition process includes: All historical attribute parameters of the manufacturing scenario corresponding to each sample in the training set are used as the input of the random forest model, and the historical carbon footprint corresponding to the current sample is used as the output of the random forest model. The random forest model is trained to obtain the carbon footprint regression prediction model.

[0048] It should be understood that the corresponding carbon footprint value is calculated based on the prediction model obtained by random forest training: .

[0049] The sample and its carbon footprint value form a new data pair .

[0050] It should be understood that the sample set and its carbon footprint data are divided into a training set and a test set. Usually, it can be divided in a ratio of 70% for the training set and 30% for the test set. The data of the training set and the test set are standardized so that the numerical ranges of different variables have similar scales, which is convenient for the random forest algorithm to learn and predict. Determine the parameters of the random forest model, such as the number of decision trees 100 and the maximum depth of each decision tree 10. Use the training set data to train the random forest model to let the model learn the relationship between sample features (manufacturing scenario variables) and carbon footprint values. Use the trained random forest model to predict the carbon footprint of the test set to obtain the predicted carbon footprint value Carbon. Finally, the mean absolute error is used to calculate the average absolute error. Evaluate the performance of the prediction model. A regression prediction model specifically designed to estimate carbon emissions at different manufacturing stages was built. This approach enables efficient and accurate prediction of carbon emissions during the manufacturing process.

[0051]

[0052] In the formula, is the input sample size; is the actual carbon footprint result; Predict carbon footprint results.

[0053] Random forest is a powerful machine learning algorithm that uses a set of different decision trees to make classification or regression predictions. By aggregating and combining the predictions of these trees, random forest can significantly improve the accuracy and robustness of the overall model and reduce the risk of overfitting. Its core principle revolves around a set of decision trees constructed with randomly sampled samples and randomly selected features with replacement.

[0054] However, given the diversity and complex correlation of information in manufacturing scenarios, although the random sampling mechanism in random forests is simple and effective, when the number of samples is large, the required calculation and processing time is long and the solution efficiency is low. In addition, there are also constraints between manufacturing scenario information, such as equipment type. Processing power with parameter attributes There are constraints, and the processing power ranges corresponding to different advanced equipment are different. The random forest sampling method has not yet considered the constraints between information, and it is difficult to ensure the rationality of the training samples, which affects the accuracy of the prediction results. Therefore, the present invention adopts the Monte Carlo analysis method, combines the constraint satisfaction problem with the Monte Carlo sampling, ensures the accuracy of the sample data, and effectively supports the prediction and uncertainty analysis of carbon footprint.

[0055] Furthermore, the training set includes: a plurality of samples, each sample includes: a plurality of historical attribute parameters in a manufacturing scenario and a historical carbon footprint in the manufacturing scenario.

[0056] It should be understood that the carbon footprint refers to the total amount of greenhouse gas emissions directly and indirectly generated by an activity, product or individual throughout its life cycle, mainly carbon dioxide, but also including other greenhouse gases such as methane, nitrous oxide, etc. These gases are usually converted into carbon dioxide equivalents based on their global warming potential values).

[0057] Combined with various data sources collected on-site and integrated with the enterprise ERP / CAM / CAPP system, the attribute data of the manufacturing scene is collected, including description attributes, processing attributes, energy attributes, parameter attributes and equipment attributes, ensuring the accuracy and completeness of the data, and using it as the input variable for uncertainty analysis The output variable is the carbon emissions corresponding to each manufacturing scenario. .

[0058] Furthermore, the several historical attribute parameters in the manufacturing scenario include: Descriptive attributes, processing attributes, energy attributes, parameter attributes and equipment attributes; The description attributes include: geographic location and time; The processing attributes include: processing types such as forging, casting and turning; The energy attributes include: emission coefficient and energy types such as solar power generation, thermal power generation, and hydropower generation; The parameter attributes include: processing time and processing power; The device attributes include: device type and device lifespan.

[0059] It should be understood that product processing and manufacturing is the process of converting raw materials or parts into finished or semi-finished products with specific functions, shapes and qualities through a series of process operations and processing. This process usually includes multiple stages such as selection and preparation of raw materials, processing and molding, assembly and debugging. Carbon emissions are generated at each stage of the processing and manufacturing process, and their size is affected by many factors, including differences in process input and output lists, specifically involving resource and energy consumption and gas emissions. In different scenarios, there are significant differences in the state of the product and its external environment, which has a profound impact on the collected carbon emission data. These data are closely related to specific scenarios, and each scenario will produce different data sets. For example, even under the same heat treatment process, using different processing equipment or operating power may lead to different carbon emission results. Therefore, in order to systematically and accurately depict the green attributes in the manufacturing process, the present invention proposes the concept of "manufacturing scenario". This concept can uniformly express and describe all relevant information that affects the carbon emissions of the manufacturing process, thereby helping to accurately assess carbon emissions.

[0060] Manufacturing scenarios refer to the characterization of various processing and manufacturing activities, including time, location, environment, activities, etc. The process can be the entire technical process or it can be composed of several process segments. The time, location, and environmental background data of the processing and manufacturing stage are closely related to the resource attributes, energy attributes, and environmental emission attributes of the product. In order to improve the accuracy and reliability of these data, the attributes of the manufacturing scenario are divided into descriptive attributes, , Processing properties , Energy properties , parameter attributes and device properties .

[0061] Describing properties , including the time of the manufacturing phase , Geographical location Aspects, expressed as: Time specifies the precise moment when the activity occurs, including year, month, day, or hour and minute, which has an important impact on the storage of process inventory data and the selection of base years for subsequent environmental impact assessments. Geographic location includes origin, destination or specific operation location.

[0062] Processing properties , describes the characteristics of the execution process of manufacturing activities, mainly used to express the processing type of the process , The processing types corresponding to this attribute include casting, forging, welding, heat treatment, machining, painting, assembly, etc.

[0063] Energy properties , used to describe the different types and forms of energy consumed in various stages of processing and manufacturing, including energy types , emission factor , These energy resources can be obtained directly, such as natural energy such as solar energy, wind energy, geothermal energy, etc., or they can be converted into useful energy through processing and conversion, such as coal, oil, natural gas, electricity, etc. Among them, electricity can be further divided into thermal power, hydropower and hybrid power.

[0064] Parameter properties , is a set of data describing various operating conditions and process parameters in the manufacturing process, including processing power , Processing time , expressed as , these parameters have a crucial impact on the quality and performance of the final product.

[0065] Device properties , refers to the type of equipment used to complete the processing task (such as machining centers, CNC machine tools, general machine tools, etc.), equipment life , expressed as: .

[0066] Based on the above description, the manufacturing scenario is defined as follows:

[0067] in, , , , , .

[0068] Furthermore, the S103: calculating the dispersion coefficient according to the carbon footprint prediction value includes: Uncertainty analysis evaluates carbon emission outcomes under various manufacturing scenarios using coefficients of dispersion As the main metric, the uncertainty in these results is quantified.

[0069]

[0070] In the formula, is the sample size; is the carbon footprint result, is the mean of the carbon footprint results.

[0071] If the coefficient of dispersion is lower than the set threshold, it indicates that the result is highly credible and robust. Otherwise, if the coefficient of dispersion is higher than the set threshold, it indicates that the input parameters need to be improved to improve the robustness of the results and ensure that the coefficient of dispersion meets the established standards, thereby improving the reliability of carbon emissions.

[0072] It should be understood that the above technical solution is to conduct uncertainty analysis on the manufacturing scenario in order to gain a deeper understanding of how changes in the input manufacturing scenario affect the carbon emissions of the product manufacturing process. This uncertainty analysis includes two key contents: (1) how to quickly and accurately obtain output results based on input information; (2) how to comprehensively evaluate the impact of input on output. The CV result determines whether the discrete coefficient meets the threshold. This coefficient can generally be set by enterprise designers based on experience or needs. Generally, the threshold of the discrete coefficient is set to 0.15.

[0073] Furthermore, the step S104: screening out key parameters for each sample to be predicted, and modifying the key parameters to obtain a modified sample to be predicted includes: According to the comprehensive weight of manufacturing scenario attribute parameters , according to the comprehensive weight Sort the attribute parameters from large to small, select the top 50% of the attribute parameters as key parameters, and get the key parameter group These parameters have a more significant potential impact on the carbon footprint than other parameters.

[0074] For the key parameter groups, a multivariate linear regression model is constructed:

[0075] in, is the coefficient to be determined, is the error term.

[0076] The multivariate linear regression model is trained by the input sample data, the coefficients to be determined are determined by the least squares method, the linear relationship between key parameters and carbon footprint is analyzed, and the sensitivity of key parameters is preliminarily determined.

[0077] For the multiple linear regression model, the regression coefficients are directly As a sensitivity indicator, The larger the absolute value of The attribute parameters whose absolute values ​​are greater than the set threshold are taken as key parameters.

[0078] Furthermore, the step S104: screening out key parameters for each sample to be predicted and modifying the key parameters includes: Increase or decrease the identified key parameters by 5% to 10%, recalculate the carbon footprint and evaluate the optimization effect, calculate the CV index, until a better optimization effect is achieved to meet the company's requirements.

[0079] Embodiment 2 This embodiment provides a product carbon footprint uncertainty analysis system based on manufacturing scenario constraints, including: The acquisition module is configured to: acquire a set of samples to be predicted, the set of samples to be predicted comprising: a plurality of samples to be predicted, each sample to be predicted comprising: a plurality of attribute parameters to be predicted in a manufacturing scenario; perform random sampling on the set of samples to be predicted to obtain an initial sample set; perform noise elimination on the initial sample set to obtain a set of samples to be predicted after the noise is eliminated; The prediction module is configured to: input the set of samples to be predicted after noise removal into the carbon footprint regression prediction model to obtain the carbon footprint prediction value corresponding to each sample to be predicted in the set of samples to be predicted after noise removal; The calculation module is configured to: calculate the dispersion coefficient according to the carbon footprint prediction value; if the dispersion coefficient is higher than the set threshold, it means that the parameters of the current sample to be predicted need to be modified; after obtaining all the samples to be predicted whose parameters need to be modified, enter the modification module; if the dispersion coefficient is lower than the set threshold, it means that the parameters of the current sample to be predicted do not need to be modified; The modification module is configured to: screen out key parameters for each sample to be predicted, modify the key parameters to obtain the modified sample to be predicted, and then obtain the modified set of samples to be predicted, and return to the acquisition module.

[0080] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. The product carbon footprint uncertainty analysis method based on manufacturing scenario constraints is characterized by: include: Acquire a set of samples to be predicted, the set of samples to be predicted comprising: a plurality of samples to be predicted, each sample to be predicted comprising: a plurality of attribute parameters to be predicted in a manufacturing scenario; perform random sampling on the set of samples to be predicted to obtain an initial sample set; perform noise elimination on the initial sample set to obtain a set of samples to be predicted after the noise is eliminated; The sample set to be predicted after noise removal is input into the carbon footprint regression prediction model to obtain the carbon footprint prediction value corresponding to each sample to be predicted in the sample set to be predicted after noise removal; According to the predicted carbon footprint value, the dispersion coefficient is calculated; if the dispersion coefficient is higher than the set threshold, it means that the parameters of the current sample to be predicted need to be modified; after obtaining all the samples to be predicted whose parameters need to be modified, proceed to the next step; if the dispersion coefficient is lower than the set threshold, it means that the parameters of the current sample to be predicted do not need to be modified; For each sample to be predicted, key parameters are screened out, and the key parameters are modified to obtain the modified sample to be predicted, and then a modified set of samples to be predicted is obtained, and the process returns to the step of obtaining the set of samples to be predicted.

2. The product carbon footprint uncertainty analysis method based on manufacturing scenario constraints as claimed in claim 1 is characterized in that: Randomly sample the sample set to be predicted to obtain the initial sample set, including: (1-1): Determine the probability distribution function of the manufacturing scene attribute parameters in the sample set to be predicted; the probability distribution function includes: a uniform distribution function of discrete variables and a normal distribution function of continuous variable pairs; (1-2): According to the probability distribution function of the manufacturing scene attribute parameters, Monte Carlo analysis is used to randomly sample the manufacturing scene attribute parameters to obtain an initial sample data set; if the attribute parameters conform to the normal distribution, random values ​​are taken according to the normal distribution; if the attribute parameters conform to the uniform distribution, random values ​​are taken according to the uniform distribution.

3. The product carbon footprint uncertainty analysis method based on manufacturing scenario constraints as claimed in claim 1 is characterized in that: The initial sample set is subjected to noise elimination to obtain a sample set to be predicted after noise elimination, including: For each sample to be predicted, the comprehensive weight of each attribute parameter to be predicted is calculated; All the attribute parameters to be predicted are sorted in descending order according to the comprehensive weight, and each attribute parameter to be predicted is judged in descending order according to the weight, whether it meets the association constraint; If the association constraint is met, continue to judge the next parameter; If the association constraint is not met, the current sample to be predicted is deleted from the set of samples to be predicted, so as to obtain the set of samples to be predicted after noise is eliminated.

4. The product carbon footprint uncertainty analysis method based on manufacturing scenario constraints as claimed in claim 3 is characterized in that: For each sample to be predicted, the comprehensive weight of each attribute parameter to be predicted is calculated, including: Comprehensive weights of manufacturing scenario attribute parameters for: ； in, Indicates The importance of attribute parameters to completion time, Indicates The importance of attribute parameters to cost, Indicates The importance of each attribute parameter to the environment. Indicates The constraint importance of each attribute parameter.

5. The product carbon footprint uncertainty analysis method based on manufacturing scenario constraints as claimed in claim 3 is characterized in that: The association constraints include: range constraints , follow-up constraints and combination constraints ; Range Constraints , refers to the range of each attribute parameter limited by the external environment in order to complete the specified processing task; range constraint , whose expression is: ； In the formula, It is Scene attribute parameter information; It refers to the minimum threshold of attribute information; It refers to the maximum threshold of attribute information; Sequential Constraints , means that there is a certain succession relationship between scene attribute parameters; succession constraint , whose expression is: ； In the formula, , They represent the attribute parameters of the manufacturing scene respectively. is the parameter range corresponding to the succession constraint; Combination Constraints , means that there are two or more manufacturing scenario attribute parameters that jointly determine the value of another attribute parameter; combination constraint , whose expression is: ； In the formula, , and They respectively represent the attribute parameters of the manufacturing scene.

6. The product carbon footprint uncertainty analysis method based on manufacturing scenario constraints as claimed in claim 1 is characterized in that: The several attribute parameters to be predicted in the manufacturing scenario include: description attributes, processing attributes, energy attributes, parameter attributes and equipment attributes; the description attributes include: geographic location and time; the processing attributes include: forging, casting and turning; the energy attributes include: solar power generation, thermal power generation and hydropower generation; the parameter attributes include: processing time and processing power; the equipment attributes include: equipment type and equipment life.

7. The product carbon footprint uncertainty analysis method based on manufacturing scenario constraints as claimed in claim 1 is characterized in that: Based on the predicted carbon footprint value, the dispersion coefficient is calculated, including: ； Where M is the number of samples, is the carbon footprint result, is the mean of the carbon footprint results.

8. The product carbon footprint uncertainty analysis method based on manufacturing scenario constraints as claimed in claim 1 is characterized in that: For each sample to be predicted, key parameters are screened out and modified to obtain the modified sample to be predicted, including: According to the comprehensive weight of manufacturing scenario attribute parameters , according to the comprehensive weight Sort the attribute parameters from large to small, select the top 50% of the attribute parameters as key parameters, and get the key parameter group ; For the key parameter groups, a multivariate linear regression model is constructed: ； in, is the coefficient to be determined, is the error term; The multivariate linear regression model is trained by inputting sample data, the coefficients to be determined are determined by the least square method, the linear relationship between key parameters and carbon footprint is analyzed, and the sensitivity of key parameters is preliminarily determined; For the multiple linear regression model, the regression coefficients are directly As a sensitivity indicator, The larger the absolute value of The attribute parameters whose absolute values ​​are greater than the set threshold are taken as key parameters.

9. The product carbon footprint uncertainty analysis method based on manufacturing scenario constraints as claimed in claim 1 is characterized in that: For each sample to be predicted, key parameters are screened out and modified, including: increasing or decreasing the identified key parameters by 5% to 10%.

10. A product carbon footprint uncertainty analysis system based on manufacturing scenario constraints, characterized by: include: The acquisition module is configured to: acquire a set of samples to be predicted, the set of samples to be predicted comprising: a plurality of samples to be predicted, each sample to be predicted comprising: a plurality of attribute parameters to be predicted in a manufacturing scenario; perform random sampling on the set of samples to be predicted to obtain an initial sample set; perform noise elimination on the initial sample set to obtain a set of samples to be predicted after the noise is eliminated; The prediction module is configured to: input the set of samples to be predicted after noise removal into the carbon footprint regression prediction model to obtain the carbon footprint prediction value corresponding to each sample to be predicted in the set of samples to be predicted after noise removal; The calculation module is configured to: calculate the dispersion coefficient according to the carbon footprint prediction value; if the dispersion coefficient is higher than the set threshold, it means that the parameters of the current sample to be predicted need to be modified; after obtaining all the samples to be predicted whose parameters need to be modified, enter the modification module; if the dispersion coefficient is lower than the set threshold, it means that the parameters of the current sample to be predicted do not need to be modified; The modification module is configured to: screen out key parameters for each sample to be predicted, modify the key parameters to obtain the modified sample to be predicted, and then obtain the modified set of samples to be predicted, and return to the acquisition module.

Images