Conventional prediction methods faced challenges such as multi-source heterogeneity and strong parameter uncertainty, so equipment, process, and resource-related factors were integrated to identify and define manufacturing scenarios, enabling the unified representation and description of carbon emission influences. The ensemble mechanism of random forest decision trees was combined with Bayesian adaptive hyperparameter optimization to establish a three-stage prediction framework “feature selection, model training, parameter tuning” for the high-efficiency prediction of carbon emissions. A Monte Carlo-Bayesian optimized random forest approach for uncertainty analysis was developed, where sensitive carbon emission parameters were identified and their impacts were quantified to enhance reliability through targeted parameter optimization. A case study on wind turbine blade machining demonstrated the effectiveness of the proposed method. The results show excellent agreement between predicted and actual carbon emissions. After uncertainty analysis, the coefficient of variation is reduced by 0.0347, significantly improving the reliability of the prediction results and supporting more robust decision-making.
随机森林(random forest,RF)通过训练数据集进行Bootstrap抽样,对特征空间进行随机子集选择,并行构建多棵结构独立的决策树,再通过结果平均的方式聚合所有决策树的预测输出。RF性能高度依赖超参数配置,不同超参数间存在复杂的非线性耦合关系,其组合形成的搜索空间呈指数级扩张,传统手动调优或网格搜索方法不仅难以触及最优参数区间,还需消耗大量计算资源与时间成本。为解决此问题,本文引入贝叶斯优化算法,其核心优势在于通过高斯过程等概率模型拟合目标函数,并基于采集函数动态指导下一轮参数搜索方向,此方法能够高效识别最优的随机森林超参数组合。因此,本文提出融合贝叶斯优化的随机森林算法(Bayesian-optimized random forest,BORF),形成“特征筛选-模型训练-参数调优”三阶预测体系,实现制造场景驱动的加工过程碳排放高效、精准预测。算法流程如图3所示,具体步骤如下。
4)不确定性分析。常用的不确定性评估指标包括碳排放最大值、最小值、均值、标准差、变异系数(coefficient of variation,CoV)、峰度和偏度。本文采用CoV作为评估相对离散程度的关键指标,其定义为标准差与均值的比值,能够反映结果的相对离散程度。若CoV值低于预设阈值,碳排放预测结果具备较高可信度与稳健性;若CoV值超过阈值,则表明输入参数或模型存在较大不确定性,需进一步优化,以提高碳排放评估结果的可靠性。
YANMeng, LITao, YANGChen, et al. Calculation Method of LCA Parameters for Green Design and Assessment Collaboration of Electromechanical Products[J]. China Mechanical Engineering, 2023, 34(12): 1453-1464.
[3]
BRILLINGERM, WUWERM, SMAJICB, et al. Novel Method to Predict the Energy Consumption of Machined Parts in the Design Phase to Attain Sustainability Goals[J]. Journal of Manufacturing Processes, 2023, 101: 1046-1054.
[4]
PARKJ H, LIM S, YEO C, et al. DFGAT for Recognizing Design Features from a B-rep Model for Mechanical Parts[J]. Robotics and Computer-Integrated Manufacturing, 2025, 93: 102938.
[5]
ZHOUTianyu, LIHengxu, LIXinming, et al. Feature-based Modeling for Variable Fractal Geometry Design Integrated into CAD System[J]. Advanced Engineering Informatics, 2023, 57: 102006.
JIAKang, TANGJing, RENDongxu, et al. Construction of Assembly Information Model Based on Assembly Feature Automatic Extraction for STEP Models[J]. China Mechanical Engineering, 2025, 36(1): 123-132.
[8]
KONGLin, WANGLiming, LIFangyi, et al. Multi-layer Integration Framework for Low Carbon Design Based on Design Features[J]. Journal of Manufacturing Systems, 2021, 61: 223-238.
[9]
KONGLin, WANGLiming, LIFangyi, et al. A Life-cycle Integrated Model for Product Eco-design in the Conceptual Design Phase[J]. Journal of Cleaner Production, 2022, 363: 132516.
[10]
UMEDAY, FUKUSHIGES, TONOIKEK. Evaluation of Scenario-based Modularization for Lifecycle Design[J]. CIRP Annals, 2009, 58(1): 1-4.
[11]
AKASAPUU, HEHENBERGERP. A Design Process Model for Battery Systems Based on Existing Life Cycle Assessment Results[J]. Journal of Cleaner Production, 2023, 407: 137149.
[12]
ZHANGXiufen, ZHANGShuyou, HUZhiyong, et al. Identification of Connection Units with High GHG Emissions for Low-carbon Product Structure Design[J]. Journal of Cleaner Production, 2012, 27: 118-125.
GUOHongfei, LUXinyu, RENYaping, et al. Reinforcement Learning-based Swarm Evolutionary Algorithm to Solve Two-sided Multi-objective Synchronous Parallel Disassembly Line Balancing Problem[J]. Journal of Mechanical Engineering, 2023, 59(7): 355-366.
[15]
KHALOIEH, DOLÁNYIM, TOUBEAUJ F, et al. Review of Machine Learning Techniques for Optimal Power Flow[J]. Applied Energy, 2025, 388: 125637.
[16]
RAIANS, ALIS M, SARKERM R, et al. Assessing Sustainability Risks in the Supply Chain of the Textile Industry under Uncertainty[J]. Resources, Conservation and Recycling, 2022, 177: 105975.
[17]
AHN H, CLERMONTM, LANGNERJ. Comparative Performance Analysis of Frontier-based Efficiency Measurement Methods—a Monte Carlo Simulation[J]. European Journal of Operational Research, 2023, 307(1): 294-312.
[18]
MRÓWCZYŃSKAM, SKIBAM, LEŚNIAKA, et al. A New Fuzzy Model of Multi-criteria Decision Support Based on Bayesian Networks for the Urban Areas' Decarbonization Planning[J]. Energy Conversion and Management, 2022, 268: 116035.
[19]
BAUDRYG, MACHARISC, VALLÉET. Range-based Multi-actor Multi-criteria Analysis: a Combined Method of Multi-actor Multi-criteria Analysis and Monte Carlo Simulation to Support Participatory Decision Making under Uncertainty[J]. European Journal of Operational Research, 2018, 264(1): 257-269.
[20]
PENGShitong, LITao, ZHAOJiali, et al. Petri Net-based Scheduling Strategy and Energy Modeling for the Cylinder Block Remanufacturing under Uncertainty[J]. Robotics and Computer-Integrated Manufacturing, 2019, 58: 208-219.
[21]
MENGQiang, LIFangyi, ZHOULirong, et al. A Rapid Life Cycle Assessment Method Based on Green Features in Supporting Conceptual Design[J]. International Journal of Precision Engineering and Manufacturing-Green Technology, 2015, 2(2): 189-196.
[22]
MIRZAEID, BEHBAHANINIAA, ABDALISOUSANA, et al. A Novel Approach to Repair Time Prediction and Availability Assessment of the Equipment in Power Generation Systems Using Fuzzy Logic and Monte Carlo Simulation[J]. Energy, 2023, 282: 128842.
[23]
KONGLin, WANGLiming, LIFangyi, et al. Life Cycle-oriented Low-carbon Product Design Based on the Constraint Satisfaction Problem[J]. Energy Conversion and Management, 2023, 286: 117069.