|
|
|
|
|
 |
Search published articles |
 |
|
Showing 1 results for M. Pour Ahari
Dr Seyed Morteza Mousavi, Dr Javad Gerami, Dr Mohammadreza Mozaffari, Dr Roya M. Pour Ahari, Dr Mohammadreza Feylizadeh,
Volume 17, Issue 2 (9-2026)
Abstract
Evaluating the eco-efficiency of multi-plant petrochemical clusters operating under stringent regional environmental caps requires robust analytical frameworks capable of handling complex operational complexities and parameter ambiguity. Traditional Data Envelopment Analysis (DEA) models often fail to capture systemic fixed-sum undesirable output reallocation constraints and assume deterministic input-output parameters, leading to biased and overly optimistic performance benchmarks. To overcome these limitations, this study develops a novel hybrid predictive-optimization framework integrating Extreme Gradient Boosting (XGBoost) machine learning with a Robust Fuzzy Fixed-Sum Output Data Envelopment Analysis (R-FSODEA) model. The proposed approach utilizes an XGBoost regressor trained on multi-period operational data across 75 major petrochemical complexes to project baseline expectations and establish data-driven symmetric uncertainty bounds (xij,yrj,ykju ) based on out-of-sample prediction residuals. To address both epistemic fuzzy uncertainty and aleatory operational noise, fuzzy parameters are discretized via α -cuts (α∈(0,1] ), while cardinality-constrained uncertainty budgets (Γ ) govern the level of decision conservatism. Multiple interdependent undesirable environmental emissions (CO2 , NOx , SO2 , PM10 , and COD ) are simultaneously constrained under strict aggregate regional fixed-sum emission ceilings (ykju=Cku ). The empirical results reveal a dynamic trade-off between operational conservatism and systemic performance: as the uncertainty budget increases from nominal (Γ=0 ) to fully robust configurations (Γ=4 ), mean industry eco-efficiency monotonically decreases from 0.8514 to 0.6524 , representing the analytical efficiency penalty required to guarantee feasibility under bad-case data perturbations. Furthermore, statistical correlation analyses (Spearman's ρ>0.908 , Kendall's τ>0.907 ) confirm that while individual complexes undergo localized rank shifts due to operational variance sensitivity, the global efficiency hierarchy remains highly stable. The proposed framework equips industrial executives and environmental regulators with an intelligent, policy-relevant decision-support tool for dynamic emission quota allocation, targeted technical retrofits, and eco-efficiency benchmarking under operational parameter uncertainty.
|
|
|
|
|
|
|
|
|
