Soufiane Douimia
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Soufiane Douimia has a total of 10 co-authors. Here are the top co-authors by number of shared publications:
- Abdelghani Bekrar
Université Polytechnique Hauts-de-France
3 shared publications - Abdessamad Ait El Cadi
LAMIH, UMR CNRS 8201, Université Polytechnique Hauts-de-France, Le Mont Houy, 59313 Valenciennes, France
2 shared publications Abdelghani Bekrar
1 shared publicationAbdessamad Ait El Cadi
1 shared publicationDavid Fillon
1 shared publication- Hamza Ouarnoughi
Higher National School of Computer Science -ESI ex INI Université Polytechnique Hauts-de-France LAMIH CNRS Smail Niar Université Polytechnique Hauts-de-France LAMIH/CNRS INSA Hauts-de-France
1 shared publication
Latest Publications
- Open Access
- 8 July 2026
- Published by Elsevier in SSRN Electronic Journal
Expert-guided progressive learning for label-efficient multi-fault diagnosis of industrial robots
Industrial robots in high-throughput manufacturing exhibit rare and mechanically coupled faults, making cycle-level diagnosis difficult under limited labels. This paper addresses the downstream artificial-intelligence problem of converting a fixed cycle-level feature interface (anomaly flags, Excess-Ratio (ER) localization profiles and encoder features from low-rate motor currents) into label-efficient multi-fault diagnosis with selective human handover. The engineering application is predictive maintenance of production-grade Kawasaki BX100S welding robots in an automotive manufacturing environment. The main artificial-intelligence contribution is a confidence- and distribution-aware selective predictor that abstains when classifier margin, ER ambiguity or out-of-distribution energy indicate unsafe autonomous diagnosis. Supporting components include cycle-phase-aware contrastive adaptation, active learning used as a label-efficiency mechanism, confidence-weighted supervised contrastive fine-tuning with normal replay, and an annotation-memory layer for auditability and case retrieval. On a controlled multi-fault campaign, the framework reaches 55.3 percent balanced accuracy over five coupled fault classes using 202 annotations (6.4 percent of the cycle pool), and 94.5 percent F1-score on a kinematically distinct brake fault; the coupled-class performance is reported as evidence that full autonomous diagnosis is unreliable under severe label scarcity, motivating the selective handover. In a cross-time deployment five months later, the handover policy reduces wrong autonomous decisions by 37 percent and selective risk by 18 percent. Historical computerized maintenance management system (CMMS) records confirm that the targeted fault axes match dominant maintenance drivers. Expert annotations are emulated by an oracle using controlled fault-injection labels; silent baseline drift remains unsolved and requires temporal monitoring.
- 8 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessExpert-guided progressive learning for label-efficient multi-fault diagnosis of industrial robots
Industrial robots in high-throughput manufacturing exhibit rare and mechanically coupled faults, making cycle-level diagnosis difficult under limited labels. This paper addresses the downstream artificial-intelligence problem of converting a fixed cycle-level feature interface (anomaly flags, Excess-Ratio (ER) localization profiles and encoder features from low-rate motor currents) into label-efficient multi-fault diagnosis with selective human handover. The engineering application is predictive maintenance of production-grade Kawasaki BX100S welding robots in an automotive manufacturing environment. The main artificial-intelligence contribution is a confidence- and distribution-aware selective predictor that abstains when classifier margin, ER ambiguity or out-of-distribution energy indicate unsafe autonomous diagnosis. Supporting components include cycle-phase-aware contrastive adaptation, active learning used as a label-efficiency mechanism, confidence-weighted supervised contrastive fine-tuning with normal replay, and an annotation-memory layer for auditability and case retrieval. On a controlled multi-fault campaign, the framework reaches 55.3 percent balanced accuracy over five coupled fault classes using 202 annotations (6.4 percent of the cycle pool), and 94.5 percent F1-score on a kinematically distinct brake fault; the coupled-class performance is reported as evidence that full autonomous diagnosis is unreliable under severe label scarcity, motivating the selective handover. In a cross-time deployment five months later, the handover policy reduces wrong autonomous decisions by 37 percent and selective risk by 18 percent. Historical computerized maintenance management system (CMMS) records confirm that the targeted fault axes match dominant maintenance drivers. Expert annotations are emulated by an oracle using controlled fault-injection labels; silent baseline drift remains unsolved and requires temporal monitoring.
- Open Access
- 23 June 2026
- Published by Elsevier in SSRN Electronic Journal
Hybrid Predictive Models for Demand and Commodity Price Forecasting in Supply Chains: A Machine Learning Perspective
Despite extensive development of individual hybrid forecasting models, no systematic comparison of hybridization strategies exists across standardized commodity datasets. This study evaluates twelve hybrid model configurations spanning three frameworks- Combination, Weighting, and Decomposition on two agricultural commodity time series: Vietnamese coffee demand (monthly,2000- 2020) and global wheat prices (daily, 1975 -2025). Results demonstrate that weight-based hybridization consistently outperforms alternatives, delivering 6.9% RMSE improvement over the best standalone baseline in both domains. An ETS-LSTM weighted model achieves MAPE of 9.14% for coffee demand; an ARIMAX-LSTM weighted model attains 8.30% for wheat price. Counterintuitively, the Combination approach theoretically grounded by Zhang[1]- consistently underperforms standalone models, with degradations up to 35.6% for coffee, contradicting the linear-plus-nonlinear decomposition assumption for agricultural commodity series. Structured interviews with nine Vietnamese SMEs identify financial constraints and technical expertise as primary adoption barriers. These findings establish empirically grounded selection guidelines for hybrid forecasting strategies in supply chain contexts.
- 23 June 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessHybrid Predictive Models for Demand and Commodity Price Forecasting in Supply Chains: A Machine Learning Perspective
Despite extensive development of individual hybrid forecasting models, no systematic comparison of hybridization strategies exists across standardized commodity datasets. This study evaluates twelve hybrid model configurations spanning three frameworks- Combination, Weighting, and Decomposition on two agricultural commodity time series: Vietnamese coffee demand (monthly,2000- 2020) and global wheat prices (daily, 1975 -2025). Results demonstrate that weight-based hybridization consistently outperforms alternatives, delivering 6.9% RMSE improvement over the best standalone baseline in both domains. An ETS-LSTM weighted model achieves MAPE of 9.14% for coffee demand; an ARIMAX-LSTM weighted model attains 8.30% for wheat price. Counterintuitively, the Combination approach theoretically grounded by Zhang[1]- consistently underperforms standalone models, with degradations up to 35.6% for coffee, contradicting the linear-plus-nonlinear decomposition assumption for agricultural commodity series. Structured interviews with nine Vietnamese SMEs identify financial constraints and technical expertise as primary adoption barriers. These findings establish empirically grounded selection guidelines for hybrid forecasting strategies in supply chain contexts.
- Open Access
- 2 July 2025
- Published by MDPI AG in Engineering Proceedings
- Vol. 97 (1) , 46
Enterprise-Wide Data Integration for Smart Maintenance: A Scalable Architecture for Predictive Maintenance Applications at Toyota Manufacturing
Manufacturing enterprises implementing Industry 4.0 technologies face significant challenges in integrating heterogeneous maintenance data sources and deploying AI solutions effectively. While various AI methods exist for predictive maintenance, the fundamental challenge lies in creating a cohesive architecture that enables seamless data flow and AI deployment. This paper presents a standardized architecture framework with initial implementation steps at Toyota Motor Manufacturing France. The proposed architecture introduces a four-layer approach: (1) a unified data acquisition layer integrating IoT sensors, CMMS, and legacy systems through standardized interfaces (OPC UA/MQTT), (2) a data quality and standardization layer ensuring consistent formats and automated validation, (3) a modular AI deployment layer supporting anomaly detection (Wavelet-based analysis and Deep Learning) and remaining useful life prediction (LSTM networks), and (4) a maintenance workflow integration layer with bi-directional feedback. Key innovations include a unified maintenance data model, configurable data quality pipelines, and human-in-the-loop decision support. A conceptual validation suggests this architecture can improve integration efficiency and reduce equipment downtime. This research contributes to smart maintenance by providing a scalable architecture that balances interoperability, data quality, and practical deployment in brownfield environments.
- 2 July 2025
- Published by MDPI AG in Engineering Proceedings
- Vol. 97 (1) , 46
Open AccessEnterprise-Wide Data Integration for Smart Maintenance: A Scalable Architecture for Predictive Maintenance Applications at Toyota Manufacturing
Manufacturing enterprises implementing Industry 4.0 technologies face significant challenges in integrating heterogeneous maintenance data sources and deploying AI solutions effectively. While various AI methods exist for predictive maintenance, the fundamental challenge lies in creating a cohesive architecture that enables seamless data flow and AI deployment. This paper presents a standardized architecture framework with initial implementation steps at Toyota Motor Manufacturing France. The proposed architecture introduces a four-layer approach: (1) a unified data acquisition layer integrating IoT sensors, CMMS, and legacy systems through standardized interfaces (OPC UA/MQTT), (2) a data quality and standardization layer ensuring consistent formats and automated validation, (3) a modular AI deployment layer supporting anomaly detection (Wavelet-based analysis and Deep Learning) and remaining useful life prediction (LSTM networks), and (4) a maintenance workflow integration layer with bi-directional feedback. Key innovations include a unified maintenance data model, configurable data quality pipelines, and human-in-the-loop decision support. A conceptual validation suggests this architecture can improve integration efficiency and reduce equipment downtime. This research contributes to smart maintenance by providing a scalable architecture that balances interoperability, data quality, and practical deployment in brownfield environments.