Aug 2026· Advanced Electromagnetics· Vol 15, pp. 3514-3527· 0 citations
Abstract
Accurate assessment of accident severity and regulatory deficiencies is essential for improving safety management and intelligent decision-making in complex industrial systems. This study proposes an interpretable machine learning framework to evaluate the influence of regulatory variables on coal mine accident severity by integrating data-driven prediction with SHAP-based feature attribution. Based on grounded-theory analysis of 382 accident investigation reports, more than 800 textual statements were condensed into 62 initial concepts, 13 secondary indicators, and five categories of regulatory factors, including design schemes, management systems, technical documents, organizational measures, and process methods. These variables were encoded and incorporated into Logistic Regression, C4.5, CART, CHAID, and Random Forest models for comparative analysis. Experimental results demonstrate that the Random Forest model achieves the best predictive performance in terms of accuracy and AUC, while SHAP analysis provides quantitative interpretation of the contribution and interaction of regulatory variables. The findings indicate that organizational measures, technical documentation, process methods, and management systems are the dominant determinants of accident severity, enabling transparent risk assessment and targeted intervention strategies. The proposed framework establishes an effective methodology for interpretable predictive analytics, intelligent safety monitoring, and data-driven decision support in complex engineering environments, offering valuable references for distributed sensing systems, industrial information fusion, and intelligent monitoring architectures related to Electromagnetic Waves, Antennas and Propagation engineering applications.
This study analyzes China's national coal mine accident data (2000-2025) to evaluate safety performance. Using one-way analysis of variance and random forest regression model, the study quantitatively analyzes temporal trends, spatial clustering, accident causes and types. A time-series model was developed to predict the million-ton mortality rate, and the optimal ETS(M,Ad,N) realizes a dynamic 5-year mortality rate forecast. Results show fatal accidents peak in March, June-August and November-December. Geographically, central and southwestern regions are high-risk, with Shanxi Province alone accounting for 15.6% of accidents. Analysis of accident causes revealed that 47.2% were attributable to human factors. By type, mechanical accidents were most common, and fire accidents were the least. Based on historical statistical trends, the ETS(M,Ad,N) model projects that the mortality rate per million tons will continue to decline from 2026 to 2030. These quantitative findings provide a data-driven reference for prioritizing safety interventions and targeted risk management strategies.
Lin-Juan Liu, Wenlin Li, Jia-Long Yang· International Journal of Occ...· 0 citations
The mining industry faces increasing challenges in enhancing productivity within complex operational environments characterized by the interaction of technical, organizational, human, and external factors. Nevertheless, existing studies frequently examine these determinants independently, thereby limiting a comprehensive understanding of their relative importance and combined influence on operational performance. This study develops a decision framework to identify and prioritize the factors affecting productivity in medium-scale mining in the Coquimbo Region, Chile, through the application of the Analytic Hierarchy Process (AHP) based on expert judgment obtained from active mining operations. The AHP model was structured using two criteria, seven subcriteria, and twenty-three decision factors, whose consistency and reliability were assessed through the consistency ratio (CR) and Cronbach’s alpha coefficient. The results revealed a marked predominance of Internal factors (75.0%) over External factors (25.0%), indicating that productivity is influenced primarily by variables that can be managed at the organizational level. Among the evaluated subcriteria, Work planning achieved the highest priority (32.0%), whereas Scheduling and control (10.2%), Human factors (6.4%), and Working conditions (5.2%) emerged as the most influential decision factors. Furthermore, sensitivity analysis confirmed the robustness and stability of the model. Beyond establishing a prioritization of productivity determinants, this study provides a decision-support framework that can assist mining companies in strengthening productivity management and improving operational performance in medium-scale mining.
The operational dynamics of the steel industry constitute one of the work systems with the highest severity and accident rates. To address this, this research multidimensionally evaluates and stochastically predicts the preventive capability of the safety system in a steel plant. Using a quantitative, evaluative, and longitudinal three-phase design, the retrospective evaluation of nine preventive variables was employed using Six Sigma metrics (DPMO, Z, Y), along with the evaluation of overall performance through the Geometric Capability Indicator (GCI) and the Arithmetic Capability Indicator (ACI), and the stochastic modeling of the process using Markov chains. It was demonstrated that evaluating processes in isolation hides structural inefficiencies, as four variables showed an Excellent individual performance (Z≈6.0), but the comprehensive multivariate evaluation revealed a Deficient systemic state (GCI of 0.471 and ACI of 0.493). Furthermore, Markov modeling on the compliance of the process management index predicted a 100% probability of long-term stagnation in a deficient absorbing state (x1=1). It is concluded that the proposed method functions as a rational anticipation system that provides potential managerial benefits by offering early warning indicators of operational degradation, supporting corrective decision-making on unstable preventive indicators.
Tomás José Fontalvo Herrera, Enrique J. Delahoz-Domínguez, Neiser Rodelo Barrios· Engineer· 0 citations
Grey relational analysis serves as an effective tool for processing small-sample and poor-information systems. Its application in oilfield development has grown substantially. This study systematically reviews the methodological evolution from traditional grey relational analysis to improved forms and weighted improved forms. It clarifies differences in theoretical assumptions, distinction coefficient determination, and weight integration mechanisms across these methods. The work aims to provide methodological support for multi-source information fusion and quantitative decision-making in complex reservoir settings. On this basis, the study systematically synthesizes four major application domains in oilfield development: preferential channel identification and prediction, reservoir evaluation and classification, productivity prediction with key factor analysis, and development optimization with risk warning. Using specific case examples, the analysis examines method applicability and critical considerations across these scenarios. Grey relational analysis has formed a complete technical chain from theoretical foundations to engineering applications in oilfield development. The methodological evolution exhibits clear trends toward adaptive parameter adjustment and information-based weighting. Engineering applications span the entire development lifecycle, establishing quantitative linkages between static geological characteristics and dynamic production responses. Current methods show limitations in nonlinear relationship characterization, multi-physical field coupling, high-dimensional dynamic sequence processing, and weight interpretability. Future directions include theoretical model innovation, spatiotemporal relational expansion, and deep integration with deep learning architectures.
Han Zhang, Chenji Wei, Guangya Zhu et al.· Journal of Physics, Conferen...· 0 citations
This study explores whether modern machine learning (ML) techniques can effectively use structured open-access cross-country indicators to predict the World Bank’s Logistic Performance Index scores. The work examines different regression algorithms through a comprehensive evaluation of 139 countries employing three-based methods. Additional models were also tested as benchmarks. The analysis includes a rigorous treatment of missing data through iterative imputation, a robust model assessment using cross-validation, and interpretability via SHAP values. The results show that the Extra Tree model (test R2 = 0.889, RMSE = 0.195) shows the highest predictive precision and generalization when including geographic and income groups as features. As a final step, the index was predicted for 13 countries not included in the official 2023 LPI release. These provisional reference estimates can serve as complementary benchmarks for countries where the availability and quality of big data necessary to build the KPIs do not meet the requirements of the official methodology, supporting diagnostic screening, trend assessment, and preliminary policy analysis in data-scarce contexts.
Eric Torres Ramirez, Percy J. Rosas, N. Gamarra-Vargas et al.· Journal of Shipping and Trad...· 0 citations
Addressing the challenges of processing multi-source heterogeneous data and the lagging identification of cost dispute risks in construction project settlement auditing, this paper constructs a machine learning-based numerical-semantic dual-path cost dispute precise identification model. This study extracts key features from three dimensions—numerical sensitivity, semantic conflict, and project environmental background—to construct high-dimensional feature vectors. The core architecture of the model consists of parallel dual paths: Path A utilizes the XGBoost algorithm to capture explicit numerical risks within engineering quantity and price deviations; Path B leverages an Attention-BiLSTM model to mine implicit fingerprints of rights and responsibilities conflicts within contract and change order texts. By introducing a gated fusion unit, the system achieves non-linear mapping and trade-off between numerical probabilities and semantic indices, and performs global parameter tuning in conjunction with a Bayesian optimization strategy. Case study results confirm that the model performs excellently on a dataset of 150 real settlement nodes, achieving an F1-score of 0.901 and an accuracy of 92.5%, with identification performance significantly surpassing traditional single-path identification models. Practical evaluation data indicates that the model can shorten the time consumed for individual audits by 65% and provide early warnings on average 14 d ahead. This study provides high-precision technical means for the intelligent auditing of construction project costs, offering theoretical support and practical reference for achieving the transformation from traditional experience-based auditing to data-driven risk prevention and control.
Ziyuan Zhang, Miao-Miao Yu, Yu Lei et al.· Engineering Research Express· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
