Designing a predictive framework on logistical failures for the global oil and gas supply chain

Разработка прогностической модели логистических сбоев в глобальной цепочке поставок нефти и газа
Sanchari Ghosh, Sandeep Mondal
2025-11-20

Diagnostic Bayesian NetworkFuzzy Inference SystemGraph Neural Networkglobal oil and gas supply chainlogistical failure prediction
• DBN-GNN-FIS based hybrid framework predicts multimodal disruptions for the GSC. • Trained on 2013-2023 disruptions and validated using 2024 benchmarking data. • Delivers predictive precision at 95% CI, CV <3%, SD <0.45%, observed deviation <4%. • Strengthens predictive planning and aligns with SDG 7 (7.3, 7.a, 7.b) targets. The global oil and gas supply chain (GSC) is a complex, multimodal network that depends on seamless coordination across upstream extraction, midstream transport, and downstream distribution. However, existing models often treat these layers independently, overlooking the probabilistic interdependence and propagation of failures. Addressing this limitation, the present study develops a Diagnostic-Predictive-Adaptive learning-based hybrid framework that integrates Diagnostic Bayesian Network (DBN), Graph Neural Network (GNN), and Fuzzy Inference System (FIS) to predict and interpret logistical failures across the GSC. Historical disruption data from 2013-2023 are used to train the framework which combines probabilistic inference (through DBN), relational learning (through GNN), and adaptive weighting (through FIS) to map the interconnectedness among logistical modes (i.e., pipelines, rail-roadways and waterways) for the GSC network. Validation through sensitivity analysis, convergence stability, and benchmarking using 2024 disruption data within a 95% Confidence Interval (CI) demonstrates robust predictive performance, with Co-efficient of Variation (CV) below 3%, Standard Deviations (SD) under 0.45%, and a prediction deviation below 4% relative to observed disruption patterns. It also accurately captures the directional imbalance between inbound-outbound failures, strengthens early-warning systems for capacity and strategic logistical planning and further supports resilient supply chain operations aligned with the United Nations Sustainable Development Goal 7 (7.3, 7.a, and 7.b) targets on energy efficiency, innovation, and reliability.
1
Developed a DBN-GNN-FIS hybrid framework that predicts and interprets interconnected multimodal logistical failures across the global oil and gas supply chain.
2
The framework captures inbound–outbound failure imbalances and supports early warning, capacity planning, and resilient logistics aligned with SDG 7 targets.
3
The framework integrates Bayesian probabilistic inference, graph-based relational learning, and fuzzy adaptive weighting to model failure propagation across pipelines, rail-roadways, and waterways.
4
The model achieved a coefficient of variation below 3%, standard deviation below 0.45%, and prediction deviation below 4% relative to observed disruption patterns.
5
Training on 2013–2023 disruption data and validation against 2024 benchmarks demonstrated predictive performance within a 95% confidence interval.

The global oil and gas supply chain, including its upstream extraction, midstream transport, downstream distribution, and multimodal logistical network

The prediction and interpretation of interconnected logistical failures, disruption propagation, and inbound–outbound failure imbalances across the supply chain

Publication Details
Publication Date
2025-11-20
Journal
Publisher
ISSN
Access Type
Author Information
Authors
Sanchari Ghosh
Sandeep Mondal
Explore further
Open the scid.ai AI chat with a ready-made request: it will find papers on a similar topic and help build a literature review.
Find similar papers in the chat →
Make a presentation
100%