Method for calculating carbon emission during assembly construction of cable-stayed buckling cantilever of large-span arch bridge

Abstract

The invention discloses a large-span arch bridge cable-stayed buckling cantilever assembly construction carbon emission calculation method, and belongs to the technical field of bridge construction carbon emission evaluation. Comprising the steps that the construction process is divided into six sub-projects including an arch support, a cable hoisting system, a cable-stayed buckling system, an arch rib, a spandrel structure and a bridge deck system; carbon emission sources of building material production, transportation and construction links in the construction stage are determined, and the carbon emission of each link is calculated by adopting a carbon emission factor method and accumulated; extracting characteristic factors influencing carbon emission to construct a database; three models of random forest, limit gradient lifting and lightweight gradient lifting are used for training, and the model with the highest decision coefficient is used as an optimal prediction model; and introducing an SHAP interpretation model to carry out interpretability analysis, and determining the carbon emission contribution degree of the characteristic factors. According to the method, carbon emission structured modeling and key driving factor identification in the whole construction process are realized, and a theoretical basis is provided for optimizing structural design and regulating and controlling high-carbon material use and construction machinery configuration.

Description

Technical Field

[0001] This invention belongs to the field of bridge carbon emission technology, specifically a method for calculating carbon emissions during the construction of a long-span arch bridge with cable-stayed cantilever assembly. Background Technology

[0002] Extreme weather events and sea-level rise caused by global climate change pose serious threats to human society and the ecological environment. Greenhouse gas emissions, especially carbon dioxide emissions, are considered the main contributor to global warming. According to a statistical report released by the Intergovernmental Panel on Climate Change (IPCC), infrastructure construction activities are a significant source of global carbon emissions, accounting for 40% of global energy consumption and 36% of global greenhouse gas emissions. The carbon emissions associated with transportation infrastructure construction, due to its material-intensive, energy-intensive, and long construction periods, have attracted widespread attention. As a core component of transportation infrastructure, the carbon emissions of bridge engineering have gradually become an important direction in research on carbon management throughout the building life cycle. Driven by the "dual carbon" goals (carbon reduction and emission reduction), some studies have focused on the carbon emissions throughout the bridge life cycle; however, there is still a significant gap in the modeling of systemic carbon emissions under specific construction methods, which remains an urgent problem to be solved.

[0003] Carbon emission accounting provides reliable data support and a solid foundation for promoting the green and low-carbon transformation of the economy and society. Currently, the most widely used carbon emission accounting methods worldwide include the carbon emission factor method, life cycle assessment, and carbon footprint. The carbon emission factor method originated from the 1996 "Guidelines for National Greenhouse Gas Emission Inventories," which clarified the calculation formulas and emission factors for various greenhouse gas emissions. Based on the product relationship between activity level and unit emission intensity, it is suitable for quickly assessing the carbon emissions of broad categories of activities or materials at specific stages and has been widely adopted in national and industry-level emission statistics.

[0004] Steel-concrete composite arch bridges are widely used in the construction of highways, railways and urban transportation infrastructure due to their excellent structural performance and construction advantages. The cable-stayed cantilever assembly construction method is the main method for erecting the arch ribs of long-span arch bridges. This construction method has significant advantages such as good structural self-balancing, small construction interference and strong spanning capacity.

[0005] In recent years, low-carbon research in bridge engineering has expanded from carbon emission accounting to the application evaluation of green construction techniques and decision support for structural upgrading and optimization. Although existing research covers the carbon emission characteristics of bridges throughout their entire life cycle, there are still systematic shortcomings. On the one hand, for long-span bridges, especially heavy structural systems represented by steel-concrete composite arch bridges, there is a lack of systematic carbon emission modeling methods for the construction phase oriented towards their core processes. On the other hand, existing methods mostly focus on the statistical summarization of emission data, lacking quantitative analytical paths based on interpretable models for the mechanisms of carbon emission impact, failing to deeply reveal the key driving factors of carbon emission generation, and limiting the accurate design and implementation of carbon reduction optimization strategies. To address this, a framework for carbon emission accounting and attribution analysis of driving factors applicable to the cable-stayed cantilever assembly construction process of long-span arch bridges has been constructed. This framework aims to fill the research gaps at both the modeling method and engineering application levels, and promote the quantitative and institutionalized development of carbon emission research during the construction period of long-span arch bridges. Summary of the Invention

[0006] To address the problems existing in the prior art, this invention provides a method for calculating carbon emissions during the construction of long-span arch bridges using cable-stayed cantilever assembly. The purpose is to reveal the key drivers of carbon emissions during the construction phase by using machine learning and SHAP (SHapley Additive exPlanations) interpretive models, providing decision support for the green construction of large bridge projects and the low-carbon transformation of the industry.

[0007] To achieve the above objectives, the specific solution of the present invention is as follows:

[0008] A method for calculating carbon emissions during the cantilever assembly construction of a long-span arch bridge with cable-stayed connection includes the following steps:

[0009] Step 1: Based on the construction organization and sequence of procedures, the cantilever assembly construction process of the cable-stayed bridge is divided into six independently accounted sub-projects: arch seat, cable hoisting system, cable-stayed system, arch rib, superstructure and bridge deck system.

[0010] Step 2: Determine the carbon emission sources of each sub-project in the construction phase during the building material production, building material transportation, and construction stages as described in Step 1. The construction phase is from construction preparation to the completion of the main bridge structure. Based on relevant national and local standards for building carbon emission calculation and publicly published research results, determine the carbon emission factor. Use the carbon emission factor method to calculate the carbon emission of each sub-project in the building material production, building material transportation, and construction stages during the construction phase, and sum them up to obtain the total carbon emission of each sub-project during the construction phase.

[0011] Step 3: Organize the carbon emission data of each sub-project calculated in Step 2, and extract the characteristic factors and their types that affect carbon emissions as characteristic parameters to build a database for carbon emission analysis of the construction stage of steel-concrete composite arch bridges for machine learning.

[0012] Step 4: Based on the feature parameters and database from Step 3, train three models—random forest, extreme gradient boosting, and lightweight gradient boosting—to obtain the carbon emission prediction model.

[0013] Step 5: Use mean absolute error, root mean square error and coefficient of determination as evaluation parameters for the prediction accuracy of the carbon emission prediction model, quantitatively evaluate the training results and prediction performance of the three models described in Step 4, and select the model with the highest coefficient of determination R² as the optimal prediction model.

[0014] Step 6: Introduce the SHAP interpretation model and perform interpretability analysis on the optimal prediction model in Step 5 to determine the carbon emission contribution of the characteristic factors mentioned in Step 3 in the overall or individual sub-projects.

[0015] Further, the arch abutment mentioned in step 1 includes foundation pit excavation, foundation pit backfilling, foundation concrete pouring, and installation of the stiffening frame; the cable hoisting system includes ground anchor installation, tower installation, main cable installation, and guy cable installation; the cable-stayed system includes tower installation, ground anchor installation, cable installation, and anchor tensioning; the arch rib includes arch rib processing, arch rib transportation, arch rib installation, core concrete pumping, and outer concrete pouring; the superstructure includes prefabrication of T-beams, T-beam transportation and hoisting, and pouring of columns and cap beams; the bridge deck system includes bridge deck paving.

[0016] Furthermore, the carbon emission factor mentioned in step 2 is determined based on the "Building Carbon Emission Calculation Standard" GB / T51366-2019, the local standard "Building Carbon Emission Accounting Standard" DB3502Z 5053-2019, the "2021 Electricity Carbon Dioxide Emission Factor" standard, and relevant research results.

[0017] Furthermore, the formula for the carbon emission factor method described in step 2 is as follows:

[0018] (1),

[0019] In the formula, CE represents carbon emissions; L A It indicates the activity level, including energy consumption, material use, and workload of machinery or equipment; CEF is the carbon emission factor, which indicates the amount of carbon dioxide emitted per unit of energy or material.

[0020] The formulas for calculating the carbon emissions of each sub-item of the building material production, transportation, and construction phases during the construction phase are as follows:

[0021] (5),

[0022] In the formula, CE i Carbon emissions (kgCO2) for each sub-project during the construction phase; CE pi Carbon emissions of each sub-project in the building materials production process during the construction phase; CE ti Carbon emissions for each sub-project in the transportation of building materials during the construction phase; CE ci Carbon emissions of each sub-project in the construction phase during the construction process;

[0023] The formulas for calculating the total carbon emissions of each sub-project during the construction phase are as follows:

[0024] (6),

[0025] In the formula, CE value represents the total carbon emissions (kgCO2) during the bridge construction phase.

[0026] Furthermore, the characteristic parameters mentioned in step 3 include concrete type, concrete quantity, steel type, steel quantity, transportation distance, construction machinery type, construction machinery energy consumption type, and construction machinery energy consumption.

[0027] Furthermore, the optimal prediction model described in step 5 is a lightweight gradient boosting model;

[0028] The calculation formulas for the evaluation parameters are as follows: (7),

[0029] (8),

[0030] (9),

[0031] In the formula, The actual value; These are the test values ​​for the model; for The arithmetic mean of n; n is the number of test target samples.

[0032] Furthermore, the interpretation model of SHAP described in step 6 is defined as follows: (10)

[0033] In the formula, M represents the number of input features; Let be the SHAP value of feature i; z is the feature variable.

[0034] Advantages of the present invention

[0035] (1) The present invention provides a carbon emission calculation method for the cable-stayed cantilever assembly construction of a long-span arch bridge, and establishes a research framework of "sub-item engineering decomposition - theoretical carbon emission calculation - machine learning prediction - SHAP mechanism analysis", which provides a clear research idea for the carbon emission calculation of the cable-stayed construction method of long-span arch bridges.

[0036] (2) This invention constructs a theoretical model of carbon emissions during construction based on the carbon emission factor method. The steel-concrete composite rigid frame arch bridge constructed using the cable-stayed method is divided into six sub-projects, and the carbon emission structure and contribution ratio of each stage are systematically evaluated.

[0037] (3) This invention combines machine learning algorithms and SHAP analysis to construct a carbon emission prediction model for the construction process, revealing the nonlinear influence mechanism of material consumption, structural characteristics and construction machinery on carbon emissions.

[0038] (4) This invention reveals the carbon emission characteristics of cable-stayed construction of long-span arch bridges, realizes the structured modeling of carbon emissions throughout the construction process and the identification of key driving factors, and provides a theoretical basis for optimizing structural design, regulating the use of high-carbon materials, configuring construction machinery and selecting construction methods. It is of great significance for achieving systematic emission reduction in the construction stage of long-span arch bridges. Attached Figure Description

[0039] Figure 1 A flowchart illustrating the carbon emission calculation method for cable-stayed cantilever assembly construction of long-span arch bridges.

[0040] Figure 2 A graph showing the carbon emission contribution of each sub-project.

[0041] Figure 3 A graph showing the carbon emission contributions of building materials production, transportation, and construction.

[0042] Figure 4 The mean absolute error model (MAE), root mean square error model (RMSE), and coefficient of determination model (R²) are used for calculation. 2 The performance distribution of MAE is shown in the figure; where (a) is the training set of MAE; (b) is the test set of MAE; (c) is the training set of RMSE; (d) is the test set of RMSE; and (e) is the R... 2 The training set; (f) is R 2 The test set.

[0043] Figure 5 A graph showing the results of the feature importance analysis;

[0044] Figure 6 The result of the global impact analysis of the features is shown in the figure.

[0045] Figure 7The following are single-factor feature dependency plots, where (a) is the feature dependency plot for concrete usage; (b) is the feature dependency plot for steel usage; (c) is the feature dependency plot for concrete type; (d) is the feature dependency plot for machinery type; (e) is the feature dependency plot for steel type; (f) is the feature dependency plot for machinery energy consumption; (g) is the feature dependency plot for transportation distance; and (h) is the feature dependency plot for machinery energy consumption type. Detailed Implementation

[0046] The present invention will be further explained and described below with reference to the accompanying drawings and specific embodiments. It should be noted that the specific embodiments are not intended to limit the scope of the present invention.

[0047] like Figures 1 to 7 As shown, this invention provides a method for calculating carbon emissions during the cantilever assembly construction of a long-span arch bridge with cable-stayed joints, comprising the following steps:

[0048] Step 1: Based on the construction organization and sequence of procedures, the cantilever assembly construction process of the long-span arch bridge in this embodiment is divided into six independently accounted sub-projects: arch abutment (AS), cable hoisting system (CH), cable-stayed system (CC), arch rib (AR), superstructure (SS), and bridge deck system (DS). The specific work content of each sub-project is as follows:

[0049] The construction process of the arch abutment (AS) mainly includes key stages such as foundation pit excavation, foundation pit backfilling, foundation concrete pouring, and installation of the rigid frame. The excavation and leveling of the foundation pit were completed by excavators, and the excavated soil and rock were transported by flatbed trucks to a designated quarry for centralized storage. This stage took 100 days. The arch abutment structure mainly uses C30, C40, and C50 grade concrete, with a total volume of 50661.4 m³. 3 The foundation pits consist of 6473.8 m³ and 4543 m³, reinforced with 979.068 tons of HRB400 steel bars as the main material. After the foundation pit preparation was completed, the construction team carried out the 189-day arch abutment pouring. Commonly used equipment during the construction process included concrete mixing plants, concrete mixer trucks, truck-mounted concrete pumps, cranes, and immersion vibrators.

[0050] The construction process of the Cable Carrier (CH) system mainly includes ground anchor installation, tower installation, main cable installation, and guy cable installation. The example uses a "separate lifting and locking" tower, with the CH tower section consuming 916.2 tons of steel. The ground anchors employ an integral gravity anchor structure, using 4709.6 m³ of C30 concrete and 35.901 tons of reinforcing steel. The main cable consumed 162.39 tons of steel wire rope, and the guy cables used φ15.2 diameter steel strands, totaling 1.392 tons. The system installation phase lasted 15 days and involved the use of various construction machinery, primarily winches, floating platforms, tower cranes, and concrete mixing and conveying equipment, all powered by fuel or electricity.

[0051] The construction process of the cable-stayed (CC) system mainly includes the installation of the cable-stayed tower and ground anchors, the installation of the cable-stayed cables, and the tensioning of the anchor cables. The cable-stayed tower is constructed using 1063.89 tons of steel, with longitudinal guy cables installed at the top, consuming 44.4288 tons of steel strand. The cable-stayed cables are used to fasten the arch rib segments, consuming 818.259 tons of low-relaxation prestressed steel strand. The cable-stayed cable ground anchors are constructed using 155.682 tons of steel and 365.1 m³ of cement mortar. The installation process lasted 30 days. In addition to using the winches, tower cranes, and concrete equipment from the existing cable hoisting system, specialized tensioning equipment such as hydraulic jacks and oil pumps were also used during the cable tensioning process.

[0052] The construction process of the arch rib (AR) mainly includes arch rib fabrication, arch rib transportation, arch rib installation, core concrete pumping, and outer concrete pouring. The fabrication and installation of the steel pipe arch took 100 days, consuming 1952.1 tons of steel. The main construction machinery included hydraulic flatbed trucks, plasma cutting machines, submerged arc welding machines, and gantry cranes. Core concrete pumping took 10 days, consuming 3256 cubic meters of C80 concrete. 3 The main construction machinery included concrete pumps, concrete trucks, water pumps, concrete mixing plants, and vacuum equipment; the outsourced concrete pouring took 30 days and consumed 27,944 cubic meters of C60 concrete. 3 The project involves 3,700 tons of reinforcing steel bars and uses main construction machinery including concrete pump trucks, concrete trucks, truck cranes, concrete mixing plants, immersion vibrators, and tower cranes.

[0053] The construction process of the superstructure (SS) included the prefabrication of the superstructure T-beams, transportation and hoisting of the T-beams, and the casting of the superstructure columns and cap beams. The construction took 127 days and consumed a total of 5449.18 cubic meters of C40 concrete. 3 8697.3m³ of C50 concrete 3 The project involved 2858.79 tons of reinforcing bars and 235.074 tons of steel strands. Related construction machinery included concrete pumps, vibrating equipment, cranes, beam transport vehicles, and hoisting systems.

[0054] The construction process of the bridge deck system (DS) includes bridge deck paving. Concrete and modified asphalt concrete are used as the main paving materials. The production of these materials is a major source of carbon emissions. In addition, the asphalt concrete paving and compaction processes rely on heavy construction machinery, and the emissions from diesel consumption are also significant.

[0055] Step 2: Determine the carbon emission sources for each sub-item of the construction phase during the building material production, transportation, and construction stages, as outlined in Step 1. The construction phase encompasses the entire process from construction preparation to the completion of the bridge's main structure, including both direct and indirect carbon emissions generated throughout the process. Based on relevant national and local standards for building carbon emission calculations and publicly available research findings, determine the carbon emission factor. Use the carbon emission factor method to calculate the carbon emissions for each sub-item of the building material production, transportation, and construction stages during the construction phase, and then sum them to obtain the total carbon emissions for all sub-items during the construction phase. The details are as follows:

[0056] 1. Carbon emissions in the building materials production process include the entire production process of concrete and steel. From raw material extraction to the process-related carbon emissions generated by the chemical reactions of raw materials during production, as well as fuel-related carbon emissions from energy consumption, all are included in the accounting scope. Carbon emissions in the building materials transportation process originate from the energy consumption of transportation vehicles.

[0057] Carbon emissions during the construction phase include excavation, pouring, erection, and paving according to bridge design requirements to form the main bridge structure. The carbon emission sources at this stage are the energy consumption of construction machinery and the processing of steel components during construction.

[0058] 2. The basis for calculating carbon emissions in bridge engineering using the carbon emission factor method lies primarily in the matching of carbon emission factors, including material carbon emission factors, energy carbon emission factors, and transportation carbon emission factors. By consulting standards such as the "Standard for Calculating Carbon Emissions in Buildings" GB / T51366-2019, the "Standard for Accounting for Carbon Emissions in Buildings" DB3502Z 5053-2019, and the "2021 Carbon Dioxide Emission Factor for Electricity," as well as relevant research results, the calculation formula for the carbon emission factor method in this embodiment can be obtained as follows:

[0059] (1),

[0060] In the formula, CE represents carbon emissions; L A The activity level is indicated by energy consumption, material use, and workload of machinery or equipment; CEF is the carbon emission factor, which indicates the amount of carbon dioxide emitted per unit of energy or material.

[0061] The carbon emission factors of the main building materials are shown in Table 1.

[0062] Table 1 Carbon emission factors of major building materials

[0063]

[0064] Note: The values ​​in parentheses represent the carbon emission intensity considering scrap steel recycling. The carbon emission intensity considering scrap steel recycling is approximately 0.53 times that without considering scrap steel recycling. Therefore, the recycling coefficient is 0.53. This references the research by Li Qingwei et al., "Research on Carbon Emissions and Carbon Reduction in the Whole Process of Steel Structure Manufacturing".

[0065] Bridge construction projects involve extensive construction machinery operation during the construction phase, resulting in significant energy consumption. The main energy sources are fossil fuels and purchased electricity. Fossil fuels, primarily petroleum products (such as diesel and gasoline), are the main source of fuel for construction machinery. Referring to relevant standards, the carbon emission factor for diesel is taken as 3.10 kgCO2 / kg. The carbon emission factor for purchased electricity is taken as 0.5154 kgCO2 / kWh. In this example, the daily working time for each type of machinery is calculated as 8 hours per machine (or vessel). The unit shift energy consumption of major construction machinery is shown in Table 2.

[0066] Table 2 Energy consumption per unit shift for major construction machinery

[0067]

[0068] The carbon emissions from the transportation of building materials originate from the energy consumption of the transportation vehicles. This embodiment assumes that all building materials are transported by land using medium-sized gasoline trucks with a load capacity of 8 tons, from the building material manufacturer to the construction site. The transported building materials include concrete, steel, and earthwork. Referring to the research results of Quan Zhaoxi et al., "Analysis of Carbon Reduction Effect of Tunnel Construction Spoil Utilization Based on Life Cycle Assessment Method," when building materials are transported by road, since the transport vehicles will return empty from the unloading point, and the environmental conditions during empty-load operation are 0.67 times that during full-load operation, the empty-load return coefficient ε is considered to be 1.67. The transportation carbon emission factor is referenced from the value given in GB / T 51366–2019, which is 0.115 kgCO2e / (t∙km). The transportation distances for cement and steel are 165 km and 300 km, respectively. The transportation distance for other raw materials for concrete (including aggregates and admixtures) adopts the default value of 40 km given in the standard.

[0069] 3. The calculation formulas for the carbon emissions of each sub-item of the building material production, building material transportation, and building construction during the construction phase are as follows:

[0070] (1) Building material production process

[0071] Bridge construction primarily involves two building materials: concrete and steel. Therefore, concrete (including cement, sand, gravel, and fly ash) and steel (including steel pipes, reinforcing bars, and steel strands) are selected as the analysis materials. The carbon emissions of each sub-item in the building material production process during the construction phase are analyzed. The calculation formula is:

[0072] (2),

[0073] In the formula, and These are the carbon emissions generated from the production of concrete and steel in sub-project i, respectively, in kgCO2. It is the consumption of material j in sub-project i, that is, the amount of concrete corresponding to m. 3 The corresponding t for steel; This is the carbon emission factor corresponding to material j, i.e., kgCO2 / m³ for concrete. 3 Steel corresponds to kgCO2 / t.

[0074] (2) Building material transportation

[0075] Carbon emissions from the transportation of building materials primarily originate from the energy consumption of the transportation vehicles. The carbon emissions from each sub-project within the building materials transportation process during the construction phase are as follows: The calculation formula is:

[0076] (3),

[0077] In the formula, i represents the type of building material; M i The volume of building materials transported is expressed in meters (m). 3 or t; D i The transport distance for building materials, in km; CEF ti The carbon emission factor per unit weight of transport distance for building materials under transportation mode i is expressed in kgCO2 / (t·km). This is the empty return coefficient, which is 1.67 when the transportation mode is road transport and 1 for other modes.

[0078] (3) Construction phase

[0079] In the construction phase of the steel-concrete composite arch bridge, the main sources of carbon emissions are the energy consumption of construction machinery used in the arch rib engineering, superstructure, and bridge deck paving. While temporary buildings such as workers' living quarters also involve energy consumption, their proportion is relatively small and statistically difficult to calculate; therefore, they are not considered major emission sources in the calculation. The carbon emissions of each sub-item of the construction phase are as follows: The calculation formula is:

[0080] (4),

[0081] In the formula, Energy carbon emission factor, unit is kgCO2e / kWh or kgCO2e / kg; P j The energy consumption per unit shift of construction machinery j, expressed in kWh or kg; T j N represents the number of operating days of construction machinery j, expressed in days (d). j The quantity of construction machinery j is expressed in units.

[0082] (4) Calculation of carbon emissions for each sub-project during the construction phase

[0083] The carbon emissions of each sub-project during the bridge construction phase are calculated using the following formula:

[0084] (5),

[0085] In the formula, CE i The carbon emissions of each sub-project during the construction phase are expressed in kgCO2.

[0086] (5) Total carbon emissions

[0087] The total carbon emissions during the entire bridge construction phase are obtained by summing up the carbon emissions of each sub-project. The calculation formula is as follows:

[0088] (6),

[0089] In the formula, CE represents the total carbon emissions during the bridge construction phase, expressed in kgCO2.

[0090] Based on the material consumption and construction machinery usage of each of the above-mentioned sub-projects, and combined with known carbon emission factors and existing formulas, a carbon emission calculation model for the construction phase of the cable-stayed cantilever assembly construction technology for large-span arch bridges in this embodiment is constructed to calculate the theoretical carbon emissions during the construction phase.

[0091] 1. Carbon emissions of different sub-projects

[0092] Figure 2 shows the carbon emissions of different sub-projects in the embodiment. The arch abutment (AS), as the main load-bearing foundation of the arch bridge, requires a large amount of concrete and steel reinforcement, resulting in the highest carbon emissions, accounting for 41.70% of the total. The arch rib (AR), as the main load-bearing component, uses a steel-concrete composite structure; the combined use of steel and concrete contributes 35.29% to its carbon emissions. Although the superstructure (SS) is relatively small in scale, its large number of components contributes 13.28% to the overall carbon emissions. The cable-stayed system (CH) and cable-stayed suspension system (CC), being temporary facilities, have long construction periods, but limited material usage, contributing 5.01% and 3.08% to carbon emissions, respectively. The bridge deck system (DS), mainly composed of paving materials, has relatively low material usage and carbon emission factors, resulting in the smallest carbon emission contribution of only 1.63%.

[0093] 2. Carbon emissions at different stages

[0094] Figure 3 shows the carbon emissions at different stages of the implementation plan. The building materials production stage contributed the most carbon emissions, accounting for as much as 88.60%, highlighting the key impact of the building materials production stage; followed by the building materials transportation stage, accounting for 6.37%; the construction stage had the fewest carbon emissions, accounting for only 5.03%.

[0095] Step 3: Organize the carbon emission data of each sub-project calculated in Step 2, and extract the characteristic factors and their types that affect carbon emissions as characteristic parameters. These characteristic parameters include concrete type, concrete usage, steel type, steel usage, transportation distance, construction machinery type, construction machinery energy consumption type, and construction machinery energy consumption. Construct a database for carbon emission analysis during the construction phase of steel-concrete composite arch bridges using machine learning.

[0096] The database uses construction time as the basic recording unit, collecting information on the main materials involved in the construction of each sub-project, transportation parameters, and energy consumption of construction machinery. This data is organized into a dataset containing n samples. , where x i Let y be the feature parameter of sample i (containing information other than carbon emissions). i This represents the carbon emissions corresponding to sample i. Because there are significant differences in equipment energy efficiency and energy utilization efficiency during construction, equipment type is included as an independent feature in the model.

[0097] The carbon emission data of each sub-project will be compiled, and the characteristic factors and their types that affect carbon emissions in Table 3 will be extracted as characteristic parameters to construct a machine learning carbon emission analysis database. The characteristic factors of the bridge in the example are shown in Table 3.

[0098] Table 3. Characteristic variables of the bridge carbon emission model in the example.

[0099]

[0100] Step 4: Based on the feature parameters and database from Step 3, the dataset is trained using three algorithms: Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Lightweight Gradient Boosting (LightGBM) to obtain the carbon emission prediction model.

[0101] Figure 4 shows the error histograms and standard deviations of the Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Lightweight Gradient Boosting (LightGBM) models, illustrating the predictive performance of different models. Figure 4 As shown in Figures (a) to (f), the performance of the RF model is relatively lower than that of XGBoost and LightGBM. Figure 4 As shown in (a) and (b), the average MAE error of the XGBoost and LightGBM models on the training set is less than 0.133, and on the test set it is less than 0.151, with LightGBM having the lowest error at 0.047. Figure 4 As shown in (c) and (d), for RMSE, XGBoost and LightGBM still exhibit excellent performance in the test set, with an average RMSE of less than 0.477 and a test set RMSE of less than 0.46. LightGBM still has the lowest error at 0.228. Figure 4 From (e) and (f), we can see that for R 2 On the standard training set, both XGBoost and LightGBM achieved an average score of 0.99, while only the LightGBM model achieved 0.938 on the test set, with both its mean and standard deviation being relatively low. Given the superior performance of the LightGBM model, this implementation example uses the LightGBM model for interpretability analysis.

[0102] Step 5: The mean absolute error, root mean square error, and coefficient of determination are used as evaluation parameters for the prediction accuracy of the carbon emission prediction model. The training results and prediction performance of the three algorithms described in Step 4 are quantitatively evaluated, and the model with the highest coefficient of determination R² is selected as the optimal prediction model. The optimal prediction model is a lightweight gradient boosting model.

[0103] The calculation formulas for the evaluation parameters are as follows: (7),

[0104] (8),

[0105] (9),

[0106] In the formula, The actual value; These are the test values ​​for the model; for The arithmetic mean of n; n is the number of test target samples.

[0107] Step 6: Introduce the SHAP interpretation model and perform interpretability analysis on the optimal prediction model in Step 5 to determine the carbon emission contribution of the characteristic factors mentioned in Step 3 in the overall or individual sub-projects.

[0108] The SHAP model interpretation method treats the SHAP value as an additive feature attribution method, decomposing the model's predicted value into the sum of the contributions of each input feature to the prediction result. That is, for each predicted sample, the SHAP value represents the contribution of each feature in that sample to the prediction result. The SHAP interpretation model is defined as follows:

[0109] (10)

[0110] In the formula, M represents the number of input features; Let be the SHAP value of feature i; z is the feature variable.

[0111] In machine learning, the SHAP value is used to quantify the impact of different input features on the output target. It measures the contribution of each feature to the prediction result, as these features work together to complete the prediction. The formula for determining the SHAP value of feature i is:

[0112] (11),

[0113] In the formula, F is the non-zero input set in z; S is a subset of F, in which the i-th feature is excluded from F, and is a unified measure of additive feature attribution, called the SHAP value.

[0114] Analyzing the feature importance of each variable in a machine learning model using SHAP values ​​can visualize the contribution of individual variables to the predicted value, helping to explore the degree of contribution of various influencing factors to carbon emissions.

[0115] By introducing the SHAP general interpretation model, the interpretability analysis of the prediction results of the LightGBM model in this embodiment is performed, and the results are as follows:

[0116] 1. Feature Sensitivity Analysis

[0117] The contribution of each feature variable was quantitatively evaluated, and the results of the feature importance analysis are as follows: Figure 5As shown. Overall, the impact of building material properties on carbon emission prediction results is much higher than that of construction equipment and transportation conditions. It can be observed that the SHAP value of the concrete quantity (CQ) variable is the highest, with a contribution rate exceeding 30%, indicating its dominant position in carbon emission prediction and being the core factor affecting carbon emissions. Followed by the steel quantity (SQ) and concrete type (CT), the SHAP values reach 0.204 and 0.199 respectively, corresponding to contribution rates of 28.01% and 27.33% respectively, indicating that these two variables also have an important influence on the carbon emission level. Further analysis reveals that although the carbon emission factor of high-strength concrete is significantly higher than that of ordinary concrete, in actual engineering applications, the usage of high-strength concrete is relatively limited. In this embodiment, the proportion of concrete with a strength grade of C60 and above is approximately 36.85%. Therefore, as a direct dimension, the cumulative effect of CQ, that is, the material quantity × unit carbon emission factor, plays a leading role in the total emission contribution. Therefore, CT has a relatively smaller impact on the prediction results than CQ. In contrast, the construction machinery type (MT) and steel type (ST) are secondary features, with SHAP values of 0.0444 and 0.0299 respectively, and contribution rates of 6.10% and 4.11% respectively. Although the factors related to construction equipment have a less significant promoting effect on the overall carbon emissions than building materials, they still have a certain impact under specific working conditions. The SHAP values of variables such as the energy consumption type of construction machinery (MET), transportation distance (DT), and energy consumption of construction machinery (MEQ) are generally low, indicating that the proportion of transportation links and the energy consumption of some construction machinery in the carbon emission composition is limited and the impact is relatively minor.

[0118] 2. Global Impact Analysis of Features

[0119] The global impact analysis of features shows the overall impact of each feature on the model output in the form of a density scatter plot, as Figure 6 shown. Among them, the SHAP value distributions of CQ and SQ are relatively wide, indicating their significant influence in carbon emission prediction and showing obvious fluctuations in the output results in different value ranges. The SHAP values of ST are mainly distributed between -1 and 1.5. As the concrete strength grade increases, its positive contribution to carbon emissions gradually increases, and this trend is consistent with the relationship between material quantity and carbon emission intensity. The change trend of the SHAP value of MT also shows a similar pattern, indicating that the selection of different types of construction machinery affects the carbon emission level to a certain extent. In contrast, the SHAP value distributions of ST and MEQ are relatively concentrated, and the overall contribution is relatively small, but still shows a positive impact as the feature value increases. In addition, although the SHAP values of MET and DT are generally low, they still have a considerable impact on carbon emissions in specific intervals.

[0120] 3. Single-Factor Feature Analysis

[0121] The SHAP model's single-factor feature dependency plot analysis is used to examine the impact of individual feature values ​​or categorical variables on predicted values, providing an in-depth analysis of the contribution of each carbon emission influencing factor to the total carbon emissions of this project. Specifically, for example... Figure 7 As shown in (a) to (h). Figure 7 The results show that the impact of concrete usage on carbon emissions decreases, with significant fluctuations in the low-volume range of 0-2000, while stabilizing or even showing a negative contribution at high volumes. The impact of steel usage on carbon emissions exhibits a non-linear relationship, fluctuating significantly in low-volume usage and stabilizing as usage increases. With increasing concrete grade, the SHAP value generally rises, with higher-grade concrete contributing more significantly to carbon emissions. Most machinery types have a moderate impact on carbon emissions, but certain equipment types show a high positive contribution. Regarding steel types, steel and reinforcing bars (types 1 and 2) have significant fluctuations in their impact on carbon emissions, while steel strand (type 3) has a smaller impact. Overall, the energy consumption of machinery has a limited impact on carbon emissions, but high-energy-consuming machinery (e.g., with energy consumption approaching 1400) significantly increases carbon emissions. Longer transport distances generally result in higher carbon emissions, especially in the 150-300 km range, and the mode of transport and loading conditions also significantly affect carbon emission levels. Analysis of energy consumption types shows that, under certain conditions, the positive impact of electric-driven equipment on carbon emissions may be greater than that of diesel equipment, highlighting the importance of optimizing power sources and equipment energy efficiency at construction sites.

Claims

1. A method for calculating carbon emissions during the cantilever assembly construction of a long-span arch bridge with cable-stayed joints, characterized in that... Includes the following steps: Step 1: Based on the construction organization and sequence of procedures, the cantilever assembly construction process of the cable-stayed bridge is divided into six independently accounted sub-projects: arch seat, cable hoisting system, cable-stayed system, arch rib, superstructure and bridge deck system. Step 2: Determine the carbon emission sources of each sub-project in the construction phase during the building material production, building material transportation, and construction stages as described in Step 1. The construction phase is from construction preparation to the completion of the main bridge structure. Based on relevant national and local standards for building carbon emission calculation and publicly published research results, determine the carbon emission factor. Use the carbon emission factor method to calculate the carbon emission of each sub-project in the building material production, building material transportation, and construction stages during the construction phase, and sum them up to obtain the total carbon emission of each sub-project during the construction phase. Step 3: Organize the carbon emission data of each sub-project calculated in Step 2, and extract the characteristic factors and their types that affect carbon emissions as characteristic parameters to build a database for carbon emission analysis of the construction stage of steel-concrete composite arch bridges for machine learning. Step 4: Based on the feature parameters and database from Step 3, train three models—random forest, extreme gradient boosting, and lightweight gradient boosting—to obtain the carbon emission prediction model. Step 5: Use mean absolute error, root mean square error and coefficient of determination as evaluation parameters for the prediction accuracy of the carbon emission prediction model, quantitatively evaluate the training results and prediction performance of the three models described in Step 4, and select the model with the highest coefficient of determination R² as the optimal prediction model. Step 6: Introduce the SHAP interpretation model and perform interpretability analysis on the optimal prediction model in Step 5 to determine the carbon emission contribution of the characteristic factors mentioned in Step 3 in the overall or individual sub-projects.

2. The method according to claim 1, characterized in that, Step 1 describes the arch abutment, which includes foundation pit excavation, foundation pit backfilling, foundation concrete pouring, and installation of the stiffening frame; the cable hoisting system includes ground anchor installation, pylon installation, main cable installation, and guy cable installation; the cable-stayed system includes pylon installation, ground anchor installation, cable installation, and anchor tensioning; the arch rib includes arch rib processing, arch rib transportation, arch rib installation, core concrete pumping, and outer concrete pouring; the superstructure includes prefabrication of T-beams, T-beam transportation and hoisting, and pouring of columns and cap beams; and the bridge deck system includes bridge deck paving.

3. The method according to claim 1, characterized in that, The carbon emission factor mentioned in step 2 is determined based on the "Building Carbon Emission Calculation Standard" GB / T51366-2019, the local standard "Building Carbon Emission Accounting Standard" DB3502Z 5053-2019, the "2021 Electricity Carbon Dioxide Emission Factor" standard, and relevant research results.

4. The method according to claim 1, characterized in that, The formula for the carbon emission factor method described in step 2 is as follows: （1）， In the formula, CE represents carbon emissions; L A It indicates the activity level, including energy consumption, material use, and workload of machinery or equipment; CEF is the carbon emission factor, which indicates the amount of carbon dioxide emitted per unit of energy or material. The formulas for calculating the carbon emissions of each sub-item of the building material production, transportation, and construction phases during the construction phase are as follows: （5）， In the formula, CE i Carbon emissions (kgCO2) for each sub-project during the construction phase; CE pi Carbon emissions of each sub-project in the building materials production process during the construction phase; CE ti Carbon emissions for each sub-project in the transportation of building materials during the construction phase; CE ci Carbon emissions of each sub-project in the construction phase during the construction process; The formulas for calculating the total carbon emissions of each sub-project during the construction phase are as follows: （6）， In the formula, CE value represents the total carbon emissions (kgCO2) during the bridge construction phase.

5. The method according to claim 1, characterized in that, The characteristic parameters mentioned in step 3 include concrete type, concrete quantity, steel type, steel quantity, transportation distance, construction machinery type, construction machinery energy consumption type, and construction machinery energy consumption.

6. The method according to claim 1, characterized in that, The optimal prediction model mentioned in step 5 is a lightweight gradient boosting model; The calculation formulas for the evaluation parameters are as follows: (7), （8）， （9）， In the formula, The actual value; These are the test values ​​for the model; for The arithmetic mean of n; n is the number of test target samples.

7. The method according to claim 1, characterized in that, The interpretation model of SHAP described in step 6 is defined as follows: (10) In the formula, M represents the number of input features; Let be the SHAP value of feature i; z is the feature variable.

Images