2025
Fekete, Tamas; Wicaksono, Hendro
Ontology-guided causal discovery and inference for reducing CO2 emissions in transportation Journal Article
In: International Journal of Sustainable Transportation, pp. 1–21, 2025.
Abstract | Links | BibTeX | Tags: artificial intelligence, causal AI, causal inference, machine learning, ontologies, semantic web, sustainability, transportation
@article{fekete2025ontology,
title = {Ontology-guided causal discovery and inference for reducing CO2 emissions in transportation},
author = {Tamas Fekete and Hendro Wicaksono},
url = {https://www.tandfonline.com/eprint/4CUZX6ZZ5J8R8UCGIUEH/full?target=10.1080/15568318.2025.2588608},
doi = {https://doi.org/10.1080/15568318.2025.2588608},
year = {2025},
date = {2025-12-07},
urldate = {2025-12-07},
journal = {International Journal of Sustainable Transportation},
pages = {1–21},
publisher = {Taylor & Francis},
abstract = {This study investigates how ontology-guided causal discovery can be applied to reduce CO2 emissions in transportation. The analysis uses a cross-sectional dataset of 463,568 passenger vehicles inspected in Hungary between January and March 2023, which includes key technical attributes such as engine performance, cylinder capacity, and drive-by noise levels. Using causal discovery algorithms (PC, FCI, and GES) with and without ontology-based constraints, directed acyclic graphs are constructed to identify structural relationships among these variables and CO2 emissions. For causal inference, effect sizes of engine performance and other technical characteristics on emissions are estimated while considering potential confounding factors. The findings show that ontology-informed models improve both the plausibility and interpretability of the discovered causal structures, though limitations remain regarding unobserved variables and nonlinear relationships; accordingly, this validation-focused case study provides a foundation for extensions to behavior-driven contexts (e.g. usage patterns, compliance, market responses) where causal structure and effect magnitudes are more uncertain in advance. The results indicate that cylinder capacity and specific power (engine performance at fixed displacement) are among the strongest contributors to CO2 emissions, with ontology constraints reducing spurious links and increasing robustness across algorithms. Policy implications include the need for regulatory measures that integrate domain knowledge into emissions assessments, as well as the importance of updating technical standards and testing frameworks to reflect causal interactions rather than simple correlations. These insights can support more reliable interventions to lower vehicle-related emissions and contribute to sustainable transportation strategies.},
keywords = {artificial intelligence, causal AI, causal inference, machine learning, ontologies, semantic web, sustainability, transportation},
pubstate = {published},
tppubtype = {article}
}
This study investigates how ontology-guided causal discovery can be applied to reduce CO2 emissions in transportation. The analysis uses a cross-sectional dataset of 463,568 passenger vehicles inspected in Hungary between January and March 2023, which includes key technical attributes such as engine performance, cylinder capacity, and drive-by noise levels. Using causal discovery algorithms (PC, FCI, and GES) with and without ontology-based constraints, directed acyclic graphs are constructed to identify structural relationships among these variables and CO2 emissions. For causal inference, effect sizes of engine performance and other technical characteristics on emissions are estimated while considering potential confounding factors. The findings show that ontology-informed models improve both the plausibility and interpretability of the discovered causal structures, though limitations remain regarding unobserved variables and nonlinear relationships; accordingly, this validation-focused case study provides a foundation for extensions to behavior-driven contexts (e.g. usage patterns, compliance, market responses) where causal structure and effect magnitudes are more uncertain in advance. The results indicate that cylinder capacity and specific power (engine performance at fixed displacement) are among the strongest contributors to CO2 emissions, with ontology constraints reducing spurious links and increasing robustness across algorithms. Policy implications include the need for regulatory measures that integrate domain knowledge into emissions assessments, as well as the importance of updating technical standards and testing frameworks to reflect causal interactions rather than simple correlations. These insights can support more reliable interventions to lower vehicle-related emissions and contribute to sustainable transportation strategies.
Chawalitanont, Akarawint; Bashyal, Atit; Wicaksono, Hendro
In: Journal of Manufacturing Systems, vol. 83, pp. 713–735, 2025.
Abstract | Links | BibTeX | Tags: artificial intelligence, deep learning, industry 4.0, industry 5.0, machine learning, manufacturing, sustainability
@article{chawalitanont2025uncertaintyb,
title = {Uncertainty-aware power consumption prediction in customized stainless-steel manufacturing: A comparative study of hierarchical Bayesian and deep neural models},
author = {Akarawint Chawalitanont and Atit Bashyal and Hendro Wicaksono},
doi = {https://doi.org/10.1016/j.jmsy.2025.10.010},
year = {2025},
date = {2025-12-01},
urldate = {2025-12-01},
journal = {Journal of Manufacturing Systems},
volume = {83},
pages = {713–735},
publisher = {Elsevier},
abstract = {Energy-efficient and data-driven decision-making has become a critical priority in modern manufacturing, particularly in customized or make-to-order (MTO) production where product variability causes large fluctuations in power consumption. Existing prediction models in this domain are often deterministic, lacking the ability to quantify uncertainty and capture hierarchical data dependencies, which limits their reliability for operational use. This study addresses this gap by developing a hierarchical Bayesian learning framework for power consumption prediction in customized stainless-steel manufacturing. The objective is to design models that not only achieve high predictive accuracy but also provide calibrated uncertainty estimates to support risk-aware production decisions. Four models, i.e., Hierarchical Bayesian Linear Regression (HBLR), Hierarchical Bayesian Neural Network (HBNN), Fully Connected Neural Network (FCN), and One-Dimensional Convolutional Neural Network (1D-CNN), were implemented and benchmarked using three inference algorithms: No-U-Turn Sampler (NUTS), Automatic Differentiation Variational Inference (ADVI), and Stein Variational Gradient Descent (SVGD). The innovation lies in systematically quantifying uncertainty using coverage probability, sharpness, and calibration error, and in establishing a unified comparison between probabilistic and deterministic models. Results show that the HBLR–NUTS model achieves the best trade-off between accuracy (RMSE = 11.85) and calibration quality (coverage 0.98), while ADVI offers near-equivalent performance with significantly lower computation time. These uncertainty-aware predictions can be directly integrated into Manufacturing Execution System (MES) and Enterprise Resource Planning (ERP) environments for energy-optimized scheduling and cost-aware planning. The proposed framework provides a scalable, interpretable, and statistically reliable foundation for advancing sustainable, data-driven manufacturing analytics.},
keywords = {artificial intelligence, deep learning, industry 4.0, industry 5.0, machine learning, manufacturing, sustainability},
pubstate = {published},
tppubtype = {article}
}
Energy-efficient and data-driven decision-making has become a critical priority in modern manufacturing, particularly in customized or make-to-order (MTO) production where product variability causes large fluctuations in power consumption. Existing prediction models in this domain are often deterministic, lacking the ability to quantify uncertainty and capture hierarchical data dependencies, which limits their reliability for operational use. This study addresses this gap by developing a hierarchical Bayesian learning framework for power consumption prediction in customized stainless-steel manufacturing. The objective is to design models that not only achieve high predictive accuracy but also provide calibrated uncertainty estimates to support risk-aware production decisions. Four models, i.e., Hierarchical Bayesian Linear Regression (HBLR), Hierarchical Bayesian Neural Network (HBNN), Fully Connected Neural Network (FCN), and One-Dimensional Convolutional Neural Network (1D-CNN), were implemented and benchmarked using three inference algorithms: No-U-Turn Sampler (NUTS), Automatic Differentiation Variational Inference (ADVI), and Stein Variational Gradient Descent (SVGD). The innovation lies in systematically quantifying uncertainty using coverage probability, sharpness, and calibration error, and in establishing a unified comparison between probabilistic and deterministic models. Results show that the HBLR–NUTS model achieves the best trade-off between accuracy (RMSE = 11.85) and calibration quality (coverage 0.98), while ADVI offers near-equivalent performance with significantly lower computation time. These uncertainty-aware predictions can be directly integrated into Manufacturing Execution System (MES) and Enterprise Resource Planning (ERP) environments for energy-optimized scheduling and cost-aware planning. The proposed framework provides a scalable, interpretable, and statistically reliable foundation for advancing sustainable, data-driven manufacturing analytics.
Bashyal, Atit; Bangre, Chidambar Prabhakar; Boroukhian, Tina; Wicaksono, Hendro
Unsupervised domain adaptation framework for photovoltaic power forecasting using variational auto-encoders Journal Article
In: Applied Energy, vol. 400, pp. 126606, 2025.
Abstract | Links | BibTeX | Tags: energy management, green energy, machine learning, sustainability, transfer learning
@article{bashyal2025unsupervised,
title = {Unsupervised domain adaptation framework for photovoltaic power forecasting using variational auto-encoders},
author = {Atit Bashyal and Chidambar Prabhakar Bangre and Tina Boroukhian and Hendro Wicaksono},
doi = {https://doi.org/10.1016/j.apenergy.2025.126606},
year = {2025},
date = {2025-12-01},
urldate = {2025-01-01},
journal = {Applied Energy},
volume = {400},
pages = {126606},
publisher = {Elsevier},
abstract = {The global transition towards renewable energy sources necessitates accurate forecasts of such energy sources for efficient grid management. While deep learning models offer effective solutions for intermittent renewable energy forecasts, they face challenges due to their inherent data intensity. Transfer learning methods have emerged as valuable tools to address such challenges. However, existing transfer learning frameworks used in renewable energy forecasting, require a significant amount of labelled training data for fine-tuning and knowledge transfer, limiting their applicability to scenarios where abundant data are available. This paper introduces a domain adaptation framework that enables seamless knowledge transfer from forecasting models trained with abundant data to models that need to be trained without labelled data. The proposed domain adaptation framework, leverages variational inference techniques to align feature spaces between source and target domains, utilizing a generative variational auto-encoder architecture. Experimental validation across solar parks with varying configurations demonstrates the replicability and adaptability of the proposed method. This research underscores the enduring potential of domain adaptation in advancing photovoltaic power forecasting while providing valuable insights into overcoming challenges in transfer learning-based renewable energy forecasting.},
keywords = {energy management, green energy, machine learning, sustainability, transfer learning},
pubstate = {published},
tppubtype = {article}
}
The global transition towards renewable energy sources necessitates accurate forecasts of such energy sources for efficient grid management. While deep learning models offer effective solutions for intermittent renewable energy forecasts, they face challenges due to their inherent data intensity. Transfer learning methods have emerged as valuable tools to address such challenges. However, existing transfer learning frameworks used in renewable energy forecasting, require a significant amount of labelled training data for fine-tuning and knowledge transfer, limiting their applicability to scenarios where abundant data are available. This paper introduces a domain adaptation framework that enables seamless knowledge transfer from forecasting models trained with abundant data to models that need to be trained without labelled data. The proposed domain adaptation framework, leverages variational inference techniques to align feature spaces between source and target domains, utilizing a generative variational auto-encoder architecture. Experimental validation across solar parks with varying configurations demonstrates the replicability and adaptability of the proposed method. This research underscores the enduring potential of domain adaptation in advancing photovoltaic power forecasting while providing valuable insights into overcoming challenges in transfer learning-based renewable energy forecasting.
Fekete, Tamas; Petrone, Isabella Marquez; Wicaksono, Hendro
A comprehensive causal AI framework for analysing factors affecting energy consumption and costs in customised manufacturing Journal Article
In: International Journal of Production Research, pp. 1–38, 2025.
Abstract | Links | BibTeX | Tags: artificial intelligence, causal AI, causal inference, energy management, explainable AI, industry 4.0, industry 5.0, machine learning, manufacturing, sustainability
@article{fekete2025comprehensive,
title = {A comprehensive causal AI framework for analysing factors affecting energy consumption and costs in customised manufacturing},
author = {Tamas Fekete and Isabella Marquez Petrone and Hendro Wicaksono},
url = {https://hendro-wicaksono.de/a-comprehensive-causal-ai-framework-for-analysing-factors-affecting-energy-consumption-and-costs-in-customised-manufacturing-2/},
doi = {https://doi.org/10.1080/00207543.2025.2580541},
year = {2025},
date = {2025-10-29},
urldate = {2025-10-29},
journal = {International Journal of Production Research},
pages = {1–38},
publisher = {Taylor & Francis},
abstract = {The manufacturing sector is a major energy consumer, resulting in high operational costs and environmental impacts. In customised manufacturing, optimising energy use is especially challenging due to high variability and complex interdependencies between process factors. Meanwhile, the increasing availability of operational data presents opportunities for advanced analytics. Unlike traditional machine learning, which identifies correlations, causal AI uncovers cause-and-effect relationships – enabling more explainable and actionable decision-making. This paper presents a causal AI framework that combines causal discovery and inference methods to analyse drivers of energy consumption and process duration in customised manufacturing. We integrate three core components: DirectLiNGAM and RESIT for causal discovery, and DoWhy for causal inference. Applied to a real-world case study in a German energy-intensive manufacturing Small and Medium-sized Enterprise (SME), the framework demonstrates its ability to identify key causal drivers of inefficiency and energy use. Results show improved interpretability, revealing, for example, that increasing product weight can reduce energy consumption by up to 4.70 kWh per unit, enabling targeted, data-driven interventions for optimisation. Compared to correlation-based models, the framework reveals underlying causes, helping decision-makers focus on critical levers for sustainability and cost reduction. The findings lay a foundation for applying causal AI in industrial settings through a structured, data-driven approach.},
keywords = {artificial intelligence, causal AI, causal inference, energy management, explainable AI, industry 4.0, industry 5.0, machine learning, manufacturing, sustainability},
pubstate = {published},
tppubtype = {article}
}
The manufacturing sector is a major energy consumer, resulting in high operational costs and environmental impacts. In customised manufacturing, optimising energy use is especially challenging due to high variability and complex interdependencies between process factors. Meanwhile, the increasing availability of operational data presents opportunities for advanced analytics. Unlike traditional machine learning, which identifies correlations, causal AI uncovers cause-and-effect relationships – enabling more explainable and actionable decision-making. This paper presents a causal AI framework that combines causal discovery and inference methods to analyse drivers of energy consumption and process duration in customised manufacturing. We integrate three core components: DirectLiNGAM and RESIT for causal discovery, and DoWhy for causal inference. Applied to a real-world case study in a German energy-intensive manufacturing Small and Medium-sized Enterprise (SME), the framework demonstrates its ability to identify key causal drivers of inefficiency and energy use. Results show improved interpretability, revealing, for example, that increasing product weight can reduce energy consumption by up to 4.70 kWh per unit, enabling targeted, data-driven interventions for optimisation. Compared to correlation-based models, the framework reveals underlying causes, helping decision-makers focus on critical levers for sustainability and cost reduction. The findings lay a foundation for applying causal AI in industrial settings through a structured, data-driven approach.
Boroukhian, Tina; Supyen, Kritkorn; Samson, Jhealyn Bautista; Bashyal, Atit; Wicaksono, Hendro
Integrating 3D object detection with ontologies for accurate digital twin creation in manufacturing systems Journal Article
In: The International Journal of Advanced Manufacturing Technology, vol. 140, no. 9, pp. 4679–4711, 2025.
Abstract | Links | BibTeX | Tags: data science, deep learning, digital twins, image processing, industry 4.0, machine learning, manufacturing, object detection, ontologies, semantic web
@article{boroukhian2025integrating,
title = {Integrating 3D object detection with ontologies for accurate digital twin creation in manufacturing systems},
author = {Tina Boroukhian and Kritkorn Supyen and Jhealyn Bautista Samson and Atit Bashyal and Hendro Wicaksono},
doi = {https://doi.org/10.1007/s00170-025-16548-x},
year = {2025},
date = {2025-10-01},
urldate = {2025-10-01},
journal = {The International Journal of Advanced Manufacturing Technology},
volume = {140},
number = {9},
pages = {4679–4711},
publisher = {Springer London},
abstract = {The digitization of manufacturing resources through digital twins (DTs) enhances operational efficiency and resource management. Ontologies play a key role in maintaining semantic consistency within DT systems. However, existing ontology-based approaches face challenges, including limited adaptability, integration of heterogeneous data—such as 3D images—and high manual effort in ontology development. These limitations hinder the scalability of DT implementations. Traditional 2D imaging often lacks spatial accuracy in complex manufacturing environments, causing inefficiencies and higher costs. Integrating richer data with intelligent frameworks is crucial for improving production and adaptability. The proposed study addresses these challenges by introducing a methodology that integrates existing ontologies with advanced 3D object detection models. The proposed approach employs two fully automated pipelines: one for detecting manufacturing resources from 3D images and another for mapping them into ontologies, ensuring seamless integration into DT frameworks. By leveraging established ontologies, the methodology enhances interoperability, reduces implementation complexity, and facilitates scalable deployment of DT systems across various industrial applications. Additionally, a comparative analysis of multiple advanced 3D detection models provides valuable insights to guide the selection of optimal solutions for diverse industrial settings. Experimental results show that YOLOv8 achieved the highest performance, with 91% classification accuracy, 86% precision, 81% recall, and the fastest inference time of 0.66 s. For ontology population, four machine labels—Robot, MillingMachine, BandSaw, and Lathe—were successfully integrated using a semantic similarity-based mapping strategy, enabling automated class creation and merging. This innovative framework sets a new benchmark for DT implementations, offering enhanced accuracy, efficiency, and semantic coherence in modern manufacturing.},
keywords = {data science, deep learning, digital twins, image processing, industry 4.0, machine learning, manufacturing, object detection, ontologies, semantic web},
pubstate = {published},
tppubtype = {article}
}
The digitization of manufacturing resources through digital twins (DTs) enhances operational efficiency and resource management. Ontologies play a key role in maintaining semantic consistency within DT systems. However, existing ontology-based approaches face challenges, including limited adaptability, integration of heterogeneous data—such as 3D images—and high manual effort in ontology development. These limitations hinder the scalability of DT implementations. Traditional 2D imaging often lacks spatial accuracy in complex manufacturing environments, causing inefficiencies and higher costs. Integrating richer data with intelligent frameworks is crucial for improving production and adaptability. The proposed study addresses these challenges by introducing a methodology that integrates existing ontologies with advanced 3D object detection models. The proposed approach employs two fully automated pipelines: one for detecting manufacturing resources from 3D images and another for mapping them into ontologies, ensuring seamless integration into DT frameworks. By leveraging established ontologies, the methodology enhances interoperability, reduces implementation complexity, and facilitates scalable deployment of DT systems across various industrial applications. Additionally, a comparative analysis of multiple advanced 3D detection models provides valuable insights to guide the selection of optimal solutions for diverse industrial settings. Experimental results show that YOLOv8 achieved the highest performance, with 91% classification accuracy, 86% precision, 81% recall, and the fastest inference time of 0.66 s. For ontology population, four machine labels—Robot, MillingMachine, BandSaw, and Lathe—were successfully integrated using a semantic similarity-based mapping strategy, enabling automated class creation and merging. This innovative framework sets a new benchmark for DT implementations, offering enhanced accuracy, efficiency, and semantic coherence in modern manufacturing.
Bashyal, Atit; Boroukhian, Tina; Veerachanchai, Pakin; Naransukh, Myanganbayar; Wicaksono, Hendro
Multi-agent deep reinforcement learning based demand response and energy management for heavy industries with discrete manufacturing systems Journal Article
In: Applied Energy, vol. 392, pp. 125990, 2025.
Abstract | Links | BibTeX | Tags: artificial intelligence, data science, deep learning, demand response system, energy management, green energy, machine learning, manufacturing, operation research, reinforcement learning, sustainability
@article{bashyal2025multi,
title = {Multi-agent deep reinforcement learning based demand response and energy management for heavy industries with discrete manufacturing systems},
author = {Atit Bashyal and Tina Boroukhian and Pakin Veerachanchai and Myanganbayar Naransukh and Hendro Wicaksono},
doi = {https://doi.org/10.1016/j.apenergy.2025.125990},
year = {2025},
date = {2025-08-15},
urldate = {2025-01-01},
journal = {Applied Energy},
volume = {392},
pages = {125990},
publisher = {Elsevier},
abstract = {Energy-centric decarbonization of heavy industries, such as steel and cement, necessitates their participation in integrating Renewable Energy Sources (RES) and effective Demand Response (DR) programs. This situation has created the opportunities to research control algorithms in diverse DR scenarios. Further, the industrial sector’s unique challenges, including the diversity of operations and the need for uninterrupted production, bring unique challenges in designing and implementing control algorithms. Reinforcement learning (RL) methods are practical solutions to the unique challenges faced by the industrial sector. Nevertheless, research in RL for industrial demand response has not yet achieved the level of standardization seen in other areas of RL research, hindering broader progress. To propel the research progress, we propose a multi-agent reinforcement learning (MARL)-based energy management system designed to optimize energy consumption in energy-intensive industrial settings by leveraging dynamic pricing DR schemes. The study highlights the creation of a MARL environment and addresses these challenges by designing a general framework that allows researchers to replicate and implement MARL environments for industrial sectors. The proposed framework incorporates a Partially Observable Markov Decision Process (POMDP) to model energy consumption and production processes while introducing buffer storage constraints and a flexible reward function that balances production efficiency and cost reduction. The paper evaluates the framework through experimental validation within a steel powder manufacturing facility. The experimental results validate our framework and also demonstrate the effectiveness of the MARL-based energy management system.},
keywords = {artificial intelligence, data science, deep learning, demand response system, energy management, green energy, machine learning, manufacturing, operation research, reinforcement learning, sustainability},
pubstate = {published},
tppubtype = {article}
}
Energy-centric decarbonization of heavy industries, such as steel and cement, necessitates their participation in integrating Renewable Energy Sources (RES) and effective Demand Response (DR) programs. This situation has created the opportunities to research control algorithms in diverse DR scenarios. Further, the industrial sector’s unique challenges, including the diversity of operations and the need for uninterrupted production, bring unique challenges in designing and implementing control algorithms. Reinforcement learning (RL) methods are practical solutions to the unique challenges faced by the industrial sector. Nevertheless, research in RL for industrial demand response has not yet achieved the level of standardization seen in other areas of RL research, hindering broader progress. To propel the research progress, we propose a multi-agent reinforcement learning (MARL)-based energy management system designed to optimize energy consumption in energy-intensive industrial settings by leveraging dynamic pricing DR schemes. The study highlights the creation of a MARL environment and addresses these challenges by designing a general framework that allows researchers to replicate and implement MARL environments for industrial sectors. The proposed framework incorporates a Partially Observable Markov Decision Process (POMDP) to model energy consumption and production processes while introducing buffer storage constraints and a flexible reward function that balances production efficiency and cost reduction. The paper evaluates the framework through experimental validation within a steel powder manufacturing facility. The experimental results validate our framework and also demonstrate the effectiveness of the MARL-based energy management system.
Gupta, Ishansh; Martinez, Adriana; Correa, Sergio; Wicaksono, Hendro
In: Supply Chain Analytics, vol. 10, pp. 100116, 2025.
Abstract | Links | BibTeX | Tags: artificial intelligence, causal AI, causal inference, data science, decision support systems, industry 4.0, industry 5.0, machine learning, multi criteria decision making, resillience, supply chain management, technology adoption
@article{gupta2025comparative,
title = {A comparative assessment of causal machine learning and traditional methods for enhancing supply chain resiliency and efficiency in the automotive industry},
author = {Ishansh Gupta and Adriana Martinez and Sergio Correa and Hendro Wicaksono},
doi = {https://doi.org/10.1016/j.sca.2025.100116},
year = {2025},
date = {2025-06-01},
urldate = {2025-06-01},
journal = {Supply Chain Analytics},
volume = {10},
pages = {100116},
publisher = {Elsevier},
abstract = {Efficient supplier escalation is crucial for maintaining smooth operational supply chains in the automotive industry, as disruptions can lead to significant production delays and financial losses. Many companies still rely on traditional escalation methods, which may lack the precision and adaptability offered by modern technologies. This study presents a comparative analysis of decision-making strategies for supplier escalation, evaluating causal machine learning (CML), traditional machine learning (ML), and current escalation practices in a leading German automotive company. The study employs an explanatory sequential mixed method, integrating the Analytical Hierarchy Process (AHP) with in-depth interviews with 25 industry experts. These methods are assessed based on several performance metrics: accuracy, business impact, explanation capability, human bias, stress test, and time-to-recover. Findings reveal that CML outperforms traditional ML and existing approaches, offering superior risk prediction, interpretability, and decision-making support Additionally, the research explores the internal acceptance of these technologies through the Technology Acceptance Model (TAM). The results highlight the transformative potential of CML in enhancing supply chain resilience and efficiency. By bridging the gap between predictive analytics and explainable AI, this research offers valuable guidance for firms seeking to optimize supplier management using advanced analytics.},
keywords = {artificial intelligence, causal AI, causal inference, data science, decision support systems, industry 4.0, industry 5.0, machine learning, multi criteria decision making, resillience, supply chain management, technology adoption},
pubstate = {published},
tppubtype = {article}
}
Efficient supplier escalation is crucial for maintaining smooth operational supply chains in the automotive industry, as disruptions can lead to significant production delays and financial losses. Many companies still rely on traditional escalation methods, which may lack the precision and adaptability offered by modern technologies. This study presents a comparative analysis of decision-making strategies for supplier escalation, evaluating causal machine learning (CML), traditional machine learning (ML), and current escalation practices in a leading German automotive company. The study employs an explanatory sequential mixed method, integrating the Analytical Hierarchy Process (AHP) with in-depth interviews with 25 industry experts. These methods are assessed based on several performance metrics: accuracy, business impact, explanation capability, human bias, stress test, and time-to-recover. Findings reveal that CML outperforms traditional ML and existing approaches, offering superior risk prediction, interpretability, and decision-making support Additionally, the research explores the internal acceptance of these technologies through the Technology Acceptance Model (TAM). The results highlight the transformative potential of CML in enhancing supply chain resilience and efficiency. By bridging the gap between predictive analytics and explainable AI, this research offers valuable guidance for firms seeking to optimize supplier management using advanced analytics.
Wicaksono, Hendro; Trat, Martin; Bashyal, Atit; Boroukhian, Tina; Felder, Mine; Ahrens, Mischa; Bender, Janek; Groß, Sebastian; Steiner, Daniel; July, Christoph; others,
Artificial-intelligence-enabled dynamic demand response system for maximizing the use of renewable electricity in production processes Journal Article
In: The International Journal of Advanced Manufacturing Technology, vol. 138, pp. 247–271, 2025.
Abstract | Links | BibTeX | Tags: artificial intelligence, data management, data science, demand response system, energy management, green energy, industry 4.0, interoperability, machine learning, manufacturing, ontologies, reinforcement learning, semantic web, sustainability
@article{wicaksono2024artificial,
title = {Artificial-intelligence-enabled dynamic demand response system for maximizing the use of renewable electricity in production processes},
author = {Hendro Wicaksono and Martin Trat and Atit Bashyal and Tina Boroukhian and Mine Felder and Mischa Ahrens and Janek Bender and Sebastian Groß and Daniel Steiner and Christoph July and others},
url = {https://link.springer.com/article/10.1007/s00170-024-13372-7},
doi = {https://doi.org/10.1007/s00170-024-13372-7},
year = {2025},
date = {2025-05-01},
urldate = {2025-05-01},
journal = {The International Journal of Advanced Manufacturing Technology},
volume = {138},
pages = { 247–271},
publisher = {Springer London},
abstract = {The transition towards renewable electricity provides opportunities for manufacturing companies to save electricity costs through participating in demand response programs. End-to-end implementation of demand response systems focusing on manufacturing power consumers is still challenging due to multiple stakeholders and subsystems that generate a heterogeneous and large amount of data. This work develops an approach utilizing artificial intelligence for a demand response system that optimizes industrial consumers’ and prosumers’ production-related electricity costs according to time-variable electricity tariffs. It also proposes a semantic middleware architecture that utilizes an ontology as the semantic integration model for handling heterogeneous data models between the system’s modules. This paper reports on developing and evaluating multiple machine learning models for power generation forecasting and load prediction, and also mixed-integer linear programming as well as reinforcement learning for production optimization considering dynamic electricity pricing represented as Green Electricity Index (GEI). The experiments show that the hybrid auto-regressive long-short-term-memory model performs best for solar and convolutional neural networks for wind power generation forecasting. Random forest, k-nearest neighbors, ridge, and gradient-boosting regression models perform best in load prediction in the considered use cases. Furthermore, this research found that the reinforcement-learning-based approach can provide generic and scalable solutions for complex and dynamic production environments. Additionally, this paper presents the validation of the developed system in the German industrial environment, involving a utility company and two small to medium-sized manufacturing companies. It shows that the developed system benefits the manufacturing company that implements fine-grained process scheduling most due to its flexible rescheduling capacities.
},
keywords = {artificial intelligence, data management, data science, demand response system, energy management, green energy, industry 4.0, interoperability, machine learning, manufacturing, ontologies, reinforcement learning, semantic web, sustainability},
pubstate = {published},
tppubtype = {article}
}
The transition towards renewable electricity provides opportunities for manufacturing companies to save electricity costs through participating in demand response programs. End-to-end implementation of demand response systems focusing on manufacturing power consumers is still challenging due to multiple stakeholders and subsystems that generate a heterogeneous and large amount of data. This work develops an approach utilizing artificial intelligence for a demand response system that optimizes industrial consumers’ and prosumers’ production-related electricity costs according to time-variable electricity tariffs. It also proposes a semantic middleware architecture that utilizes an ontology as the semantic integration model for handling heterogeneous data models between the system’s modules. This paper reports on developing and evaluating multiple machine learning models for power generation forecasting and load prediction, and also mixed-integer linear programming as well as reinforcement learning for production optimization considering dynamic electricity pricing represented as Green Electricity Index (GEI). The experiments show that the hybrid auto-regressive long-short-term-memory model performs best for solar and convolutional neural networks for wind power generation forecasting. Random forest, k-nearest neighbors, ridge, and gradient-boosting regression models perform best in load prediction in the considered use cases. Furthermore, this research found that the reinforcement-learning-based approach can provide generic and scalable solutions for complex and dynamic production environments. Additionally, this paper presents the validation of the developed system in the German industrial environment, involving a utility company and two small to medium-sized manufacturing companies. It shows that the developed system benefits the manufacturing company that implements fine-grained process scheduling most due to its flexible rescheduling capacities.
Fekete, Tamas; Mengistu, Girum; Wicaksono, Hendro
Leveraging causal AI to uncover the dynamics in sustainable urban transport: A bike sharing time-series study Journal Article
In: Sustainable Cities and Society, vol. 122, pp. 106240, 2025.
Abstract | Links | BibTeX | Tags: artificial intelligence, causal AI, causal inference, industry 5.0, machine learning, sustainability, transportation
@article{nokey,
title = {Leveraging causal AI to uncover the dynamics in sustainable urban transport: A bike sharing time-series study},
author = {Tamas Fekete and Girum Mengistu and Hendro Wicaksono },
doi = {https://doi.org/10.1016/j.scs.2025.106240},
year = {2025},
date = {2025-03-15},
urldate = {2025-03-15},
journal = {Sustainable Cities and Society},
volume = {122},
pages = {106240},
abstract = {The importance of developing sustainable urban transportation systems to protect the environment is increasingly recognized worldwide, particularly within the European Union. In the era of digitalization, data-driven approaches are crucial for informed decision-making. This study introduces a methodology leveraging causal artificial intelligence (causal AI) to uncover cause-and-effect relationships in urban transport data. Unlike traditional methods relying on correlations, causal AI identifies the true drivers of transport dynamics. A case study using MOL Bubi bike-sharing data from Budapest demonstrates how the PCMCI (Peter and Clark Momentary Conditional Independence) algorithm revealed complex temporal dependencies within the data, with temperature emerging as the strongest causal factor positively influencing bike usage. Additionally, the reopening of the Chain Bridge led to a 10.7% increase in bike trips, as quantified by Causal Impact analysis. This case study can be extended to more complex scenarios with unpredictable outcomes. The insights gained provide policymakers with a deeper understanding, enabling them to design policies fostering sustainable urban mobility. These results showcase the potential of causal AI to guide policies that enhance sustainable urban mobility.},
keywords = {artificial intelligence, causal AI, causal inference, industry 5.0, machine learning, sustainability, transportation},
pubstate = {published},
tppubtype = {article}
}
The importance of developing sustainable urban transportation systems to protect the environment is increasingly recognized worldwide, particularly within the European Union. In the era of digitalization, data-driven approaches are crucial for informed decision-making. This study introduces a methodology leveraging causal artificial intelligence (causal AI) to uncover cause-and-effect relationships in urban transport data. Unlike traditional methods relying on correlations, causal AI identifies the true drivers of transport dynamics. A case study using MOL Bubi bike-sharing data from Budapest demonstrates how the PCMCI (Peter and Clark Momentary Conditional Independence) algorithm revealed complex temporal dependencies within the data, with temperature emerging as the strongest causal factor positively influencing bike usage. Additionally, the reopening of the Chain Bridge led to a 10.7% increase in bike trips, as quantified by Causal Impact analysis. This case study can be extended to more complex scenarios with unpredictable outcomes. The insights gained provide policymakers with a deeper understanding, enabling them to design policies fostering sustainable urban mobility. These results showcase the potential of causal AI to guide policies that enhance sustainable urban mobility.
Ghribi, Youssef; Graha, Ega Rudy; Wicaksono, Hendro
Comparative Analysis of Statistical and Machine Learning Models for Enhancing Demand Forecasting Accuracy in the Medical Device Industry Journal Article
In: Procedia CIRP, vol. 134, pp. 849–854, 2025.
Abstract | Links | BibTeX | Tags: artificial intelligence, data science, deep learning, demand forecasting, healthcare, machine learning, manufacturing, supply chain management
@article{ghribi2025comparative,
title = {Comparative Analysis of Statistical and Machine Learning Models for Enhancing Demand Forecasting Accuracy in the Medical Device Industry},
author = {Youssef Ghribi and Ega Rudy Graha and Hendro Wicaksono},
doi = {https://doi.org/10.1016/j.procir.2025.02.209},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {Procedia CIRP},
volume = {134},
pages = {849–854},
publisher = {Elsevier},
abstract = {Demand forecasting is a crucial instrument in the business strategy. The medical devices in the healthcare system are further significant as critical roles. Multiple businesses rely on traditional forecasting techniques due to their simplicity and understandable algorithm’s easy-to-use nature characteristics. The research conducted for each model analyzes how traditional statistical, Machine Learning (ML), and Deep Learning (DL) models can be used to make demand forecasting more accurate and valuable in the medical device industry. The work expands beyond prior research to demonstrate the enhanced effectiveness of DL models compared to statistical and ML models across multiple areas. However, research still needs to identify studies that adopt a business-centric perspective on the practical applicability of these models. Research utilizing SARIMAX, Exponential Smoothing, Linear Regression, Average, Support Vector Regression (SVR), Random Forest (RF), K-Nearest Neighbour Regression (KNR), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Convolution 1D (CONV1D) models to forecast what people demand to order. The data comes from a German medical device manufacturer’s past sales record. We evaluated the model’s performance using the weighted Mean Absolute Percentage Error (wMAPE) method. These showed that DL models needed a lot of knowledge and preprocessing, but they were the most accurate at predicting what would happen. The LSTM model exhibited outstanding performance, achieving an average wMAPE of 0.3102, surpassing all other models. The research results for more sophisticated models surpass traditional statistical models despite limited datasets, recommending that medical device businesses consider investing in advanced demand forecasting models.
publisher={Elsevier}
}},
keywords = {artificial intelligence, data science, deep learning, demand forecasting, healthcare, machine learning, manufacturing, supply chain management},
pubstate = {published},
tppubtype = {article}
}
Demand forecasting is a crucial instrument in the business strategy. The medical devices in the healthcare system are further significant as critical roles. Multiple businesses rely on traditional forecasting techniques due to their simplicity and understandable algorithm’s easy-to-use nature characteristics. The research conducted for each model analyzes how traditional statistical, Machine Learning (ML), and Deep Learning (DL) models can be used to make demand forecasting more accurate and valuable in the medical device industry. The work expands beyond prior research to demonstrate the enhanced effectiveness of DL models compared to statistical and ML models across multiple areas. However, research still needs to identify studies that adopt a business-centric perspective on the practical applicability of these models. Research utilizing SARIMAX, Exponential Smoothing, Linear Regression, Average, Support Vector Regression (SVR), Random Forest (RF), K-Nearest Neighbour Regression (KNR), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Convolution 1D (CONV1D) models to forecast what people demand to order. The data comes from a German medical device manufacturer’s past sales record. We evaluated the model’s performance using the weighted Mean Absolute Percentage Error (wMAPE) method. These showed that DL models needed a lot of knowledge and preprocessing, but they were the most accurate at predicting what would happen. The LSTM model exhibited outstanding performance, achieving an average wMAPE of 0.3102, surpassing all other models. The research results for more sophisticated models surpass traditional statistical models despite limited datasets, recommending that medical device businesses consider investing in advanced demand forecasting models.
publisher={Elsevier}
}
publisher={Elsevier}
}
2024
Fan, Xiaotong; Valilai, Omid Fatahi; Wicaksono, Hendro
Integrating Economic, Technological, and Consumer Factors for Enhanced Accuracy in Electric Vehicle Demand Forecasting: A Case Study in Germany Proceedings Article
In: Lecture Notes in Mechanical Engineering (LNME): Flexible Automation and Intelligent Manufacturing: Manufacturing Innovation and Preparedness for the Changing World Order , Springer, 2024.
Abstract | Links | BibTeX | Tags: automotive industry, demand forecasting, machine learning, supply chain management, sustainability, timeseries analysis
@inproceedings{nokey,
title = {Integrating Economic, Technological, and Consumer Factors for Enhanced Accuracy in Electric Vehicle Demand Forecasting: A Case Study in Germany},
author = {Xiaotong Fan and Omid Fatahi Valilai and Hendro Wicaksono
},
doi = {https://doi.org/10.1007/978-3-031-74485-3_50},
year = {2024},
date = {2024-12-13},
booktitle = {Lecture Notes in Mechanical Engineering (LNME): Flexible Automation and Intelligent Manufacturing: Manufacturing Innovation and Preparedness for the Changing World Order },
publisher = {Springer},
abstract = {The development of the electric vehicle industry has the potential to reduce CO
emissions significantly and overcome energy supply challenges. However, manufacturing electric cars is a complex process consisting of procurement, logistics, and assembly. Accurate demand forecasting plays an essential role in the industry’s long-term development because it effectively fulfills customer needs while mitigating the risks of overproduction. Forecasting electric vehicle demand presents a significant challenge due to limited data availability and multiple factors influencing it. Comprehensive research integrating economic, technological, and consumer dynamics into demand forecasts remains notably deficient. The main goal of this research is to develop demand forecasting of electric vehicles within Germany, considering those factors, including variables like gasoline prices, the count of installed charging stations, and the Google search index. The research involves experimentation with various models, including Prophet, SARIMA, and their variants, incorporating different combinations of exogenous variables. The results demonstrate that SARIMA and its variants outperform other models regarding predictive accuracy. This research equips electric vehicle manufacturing companies with invaluable insights into market trends and the potential impact of diverse influencing variables. With this knowledge, companies can adapt their production strategies to align with market dynamics, enhancing their competitiveness in the rapidly evolving electric vehicle landscape.},
keywords = {automotive industry, demand forecasting, machine learning, supply chain management, sustainability, timeseries analysis},
pubstate = {published},
tppubtype = {inproceedings}
}
The development of the electric vehicle industry has the potential to reduce CO
emissions significantly and overcome energy supply challenges. However, manufacturing electric cars is a complex process consisting of procurement, logistics, and assembly. Accurate demand forecasting plays an essential role in the industry’s long-term development because it effectively fulfills customer needs while mitigating the risks of overproduction. Forecasting electric vehicle demand presents a significant challenge due to limited data availability and multiple factors influencing it. Comprehensive research integrating economic, technological, and consumer dynamics into demand forecasts remains notably deficient. The main goal of this research is to develop demand forecasting of electric vehicles within Germany, considering those factors, including variables like gasoline prices, the count of installed charging stations, and the Google search index. The research involves experimentation with various models, including Prophet, SARIMA, and their variants, incorporating different combinations of exogenous variables. The results demonstrate that SARIMA and its variants outperform other models regarding predictive accuracy. This research equips electric vehicle manufacturing companies with invaluable insights into market trends and the potential impact of diverse influencing variables. With this knowledge, companies can adapt their production strategies to align with market dynamics, enhancing their competitiveness in the rapidly evolving electric vehicle landscape.
emissions significantly and overcome energy supply challenges. However, manufacturing electric cars is a complex process consisting of procurement, logistics, and assembly. Accurate demand forecasting plays an essential role in the industry’s long-term development because it effectively fulfills customer needs while mitigating the risks of overproduction. Forecasting electric vehicle demand presents a significant challenge due to limited data availability and multiple factors influencing it. Comprehensive research integrating economic, technological, and consumer dynamics into demand forecasts remains notably deficient. The main goal of this research is to develop demand forecasting of electric vehicles within Germany, considering those factors, including variables like gasoline prices, the count of installed charging stations, and the Google search index. The research involves experimentation with various models, including Prophet, SARIMA, and their variants, incorporating different combinations of exogenous variables. The results demonstrate that SARIMA and its variants outperform other models regarding predictive accuracy. This research equips electric vehicle manufacturing companies with invaluable insights into market trends and the potential impact of diverse influencing variables. With this knowledge, companies can adapt their production strategies to align with market dynamics, enhancing their competitiveness in the rapidly evolving electric vehicle landscape.
Shah, Muhammad Abdullah; Wicaksono, Hendro
Leveraging Machine Learning for Power Consumption Prediction of Multi-Step Production Processes in Dynamic Electricity Price Environment Journal Article
In: Procedia CIRP, vol. 130, pp. 226-231, 2024.
Abstract | Links | BibTeX | Tags: energy management, machine learning, manufacturing, sustainability
@article{Shah2024,
title = {Leveraging Machine Learning for Power Consumption Prediction of Multi-Step Production Processes in Dynamic Electricity Price Environment},
author = {Muhammad Abdullah Shah and Hendro Wicaksono},
url = {https://www.sciencedirect.com/science/article/pii/S2212827124012332},
doi = {https://doi.org/10.1016/j.procir.2024.10.080},
year = {2024},
date = {2024-11-27},
urldate = {2024-11-27},
journal = {Procedia CIRP},
volume = {130},
pages = {226-231},
abstract = {Rising energy costs drive a compelling demand for energy-efficient manufacturing across sectors, paralleled by increasing consumer preferences for eco-friendly products. To remain competitive, companies are actively enhancing their energy efficiency. Integrating dynamic pricing in manufacturing, aimed at optimizing renewable energy use, requires strategic adjustments in production planning for sustainability. This research highlights the importance of incorporating dynamic pricing into production planning, emphasizing the need to shift processes to time slots when the energy prices are low or optimal. This study focuses on predicting the power consumption of multi-step CNC machine operations within a production cycle. Utilizing advanced Machine Learning (ML), including neural networks, statistical, and additive models, this research found unique time series characteristics influencing model performance across production steps. A practical use case within a German manufacturing Small and Medium Enterprises (SME) demonstrates how prediction results can optimize production processes in a dynamic pricing environment, providing a blueprint for diverse machinery forecasting models. This research’s insights extend to any industry managing production schedules for multiple machines with various steps in a process cycle. Industries with high energy consumption will benefit significantly through aligning operational efficiency with environmental sustainability goals.},
keywords = {energy management, machine learning, manufacturing, sustainability},
pubstate = {published},
tppubtype = {article}
}
Rising energy costs drive a compelling demand for energy-efficient manufacturing across sectors, paralleled by increasing consumer preferences for eco-friendly products. To remain competitive, companies are actively enhancing their energy efficiency. Integrating dynamic pricing in manufacturing, aimed at optimizing renewable energy use, requires strategic adjustments in production planning for sustainability. This research highlights the importance of incorporating dynamic pricing into production planning, emphasizing the need to shift processes to time slots when the energy prices are low or optimal. This study focuses on predicting the power consumption of multi-step CNC machine operations within a production cycle. Utilizing advanced Machine Learning (ML), including neural networks, statistical, and additive models, this research found unique time series characteristics influencing model performance across production steps. A practical use case within a German manufacturing Small and Medium Enterprises (SME) demonstrates how prediction results can optimize production processes in a dynamic pricing environment, providing a blueprint for diverse machinery forecasting models. This research’s insights extend to any industry managing production schedules for multiple machines with various steps in a process cycle. Industries with high energy consumption will benefit significantly through aligning operational efficiency with environmental sustainability goals.
Almais, Agung Teguh Wibowo; Susilo, Adi; Naba, Agus; Sarosa, Moechammad; Juwono, Alamsyah Muhammad; Crysdian, Cahyo; Muslim, Muhammad Aziz; Wicaksono, Hendro
Characterization of Structural Building Damage in Post-Disaster Using GLCM-PCA Analysis Integration Journal Article
In: IEEE Access, vol. 12, 2024.
Abstract | Links | BibTeX | Tags: artificial intelligence, data science, machine learning
@article{nokey,
title = {Characterization of Structural Building Damage in Post-Disaster Using GLCM-PCA Analysis Integration},
author = {Agung Teguh Wibowo Almais and Adi Susilo and Agus Naba and Moechammad Sarosa and Alamsyah Muhammad Juwono and Cahyo Crysdian and Muhammad Aziz Muslim and Hendro Wicaksono},
url = {https://ieeexplore.ieee.org/abstract/document/10697160},
doi = {https://doi.org/10.1109/ACCESS.2024.3469637},
year = {2024},
date = {2024-09-27},
urldate = {2024-09-27},
journal = {IEEE Access},
volume = {12},
abstract = {Objective: To determine the characteristics of a building after a natural disaster using image input through the integration of image analysis techniques. Methods: Several image analysis techniques, including GLCM and PCA, were employed. The GLCM process converts image input into numerical values using 8 different angles and pixel distances of 1 and 0.5 pixels. The numerical values from GLCM are then processed by PCA to extract information stored in the images of buildings post-disaster. Results: The PCA process revealed different information between images processed with GLCM at 1 pixel distance and those at 0.5 pixel distance. Validation by surveyors confirmed that the accurate information corresponding to real images was obtained from GLCM with a 0.5 pixel distance, indicating severe damage. The PCA results using GLCM at 0.5 pixel distance produced 2D and 3D visualizations with dominant coordinates in the severely damaged cluster, with a value range (n) of n ≥ 2. Conclusion: Based on these findings, the integration of image analysis techniques, specifically GLCM and PCA, can be used to determine the level of damage to buildings after a natural disaster.},
keywords = {artificial intelligence, data science, machine learning},
pubstate = {published},
tppubtype = {article}
}
Objective: To determine the characteristics of a building after a natural disaster using image input through the integration of image analysis techniques. Methods: Several image analysis techniques, including GLCM and PCA, were employed. The GLCM process converts image input into numerical values using 8 different angles and pixel distances of 1 and 0.5 pixels. The numerical values from GLCM are then processed by PCA to extract information stored in the images of buildings post-disaster. Results: The PCA process revealed different information between images processed with GLCM at 1 pixel distance and those at 0.5 pixel distance. Validation by surveyors confirmed that the accurate information corresponding to real images was obtained from GLCM with a 0.5 pixel distance, indicating severe damage. The PCA results using GLCM at 0.5 pixel distance produced 2D and 3D visualizations with dominant coordinates in the severely damaged cluster, with a value range (n) of n ≥ 2. Conclusion: Based on these findings, the integration of image analysis techniques, specifically GLCM and PCA, can be used to determine the level of damage to buildings after a natural disaster.
Supyen, Kritkorn; Mathur, Abhishek; Boroukhian, Tina; Wicaksono, Hendro
Streamlining Manufacturing Resource Digitization for Digital Twins Through Ontologies and Object Detection Techniques Proceedings Article
In: International Conference on Dynamics in Logistics, pp. 419–430, Springer Nature Switzerland Cham 2024.
Abstract | Links | BibTeX | Tags: computer vision, digital twins, machine learning, ontologies, semantic web
@inproceedings{supyen2024streamlining,
title = {Streamlining Manufacturing Resource Digitization for Digital Twins Through Ontologies and Object Detection Techniques},
author = {Kritkorn Supyen and Abhishek Mathur and Tina Boroukhian and Hendro Wicaksono},
url = {https://link.springer.com/chapter/10.1007/978-3-031-56826-8_32},
doi = {https://doi.org/10.1007/978-3-031-56826-8_32},
year = {2024},
date = {2024-04-03},
urldate = {2024-04-03},
booktitle = {International Conference on Dynamics in Logistics},
pages = {419–430},
organization = {Springer Nature Switzerland Cham},
abstract = {Digital twins play an essential role in manufacturing companies to adopt Industry 4.0. However, their uptake has been lagging, especially in European manufacturing firms. This can be attributed to the absence of automated methods for digitizing physical manufacturing resources and creating digital representations accessible and processable by both humans and computers. Our research addresses this challenge by automating the digitization of manufacturing resources captured on the shop floor. We employ object detection techniques on a set of images and align the results with an ontology that standardizes the semantic description of digital representations. This research aims to accelerate digital transformation for manufacturing companies, providing digital representations to their physical resources. The ontology-based digital representation fosters interoperability among diverse equipment and machines from various vendors. It enables the automated deployment of digital twins, improving the efficiency of planning and control of manufacturing systems.
},
keywords = {computer vision, digital twins, machine learning, ontologies, semantic web},
pubstate = {published},
tppubtype = {inproceedings}
}
Digital twins play an essential role in manufacturing companies to adopt Industry 4.0. However, their uptake has been lagging, especially in European manufacturing firms. This can be attributed to the absence of automated methods for digitizing physical manufacturing resources and creating digital representations accessible and processable by both humans and computers. Our research addresses this challenge by automating the digitization of manufacturing resources captured on the shop floor. We employ object detection techniques on a set of images and align the results with an ontology that standardizes the semantic description of digital representations. This research aims to accelerate digital transformation for manufacturing companies, providing digital representations to their physical resources. The ontology-based digital representation fosters interoperability among diverse equipment and machines from various vendors. It enables the automated deployment of digital twins, improving the efficiency of planning and control of manufacturing systems.
Thapaliya, Suman; Valilai, Omid Fatahi; Wicaksono, Hendro
Power Consumption and Processing Time Estimation of CNC Machines Using Explainable Artificial Intelligence (XAI) Journal Article
In: Procedia Computer Science, vol. 232, pp. 861–870, 2024.
Abstract | Links | BibTeX | Tags: energy management, explainable AI, machine learning, manufacturing
@article{thapaliya2024power,
title = {Power Consumption and Processing Time Estimation of CNC Machines Using Explainable Artificial Intelligence (XAI)},
author = {Suman Thapaliya and Omid Fatahi Valilai and Hendro Wicaksono},
url = {https://www.sciencedirect.com/science/article/pii/S1877050924000863},
doi = {https://doi.org/10.1016/j.procs.2024.01.086},
year = {2024},
date = {2024-03-20},
urldate = {2024-03-20},
journal = {Procedia Computer Science},
volume = {232},
pages = {861–870},
publisher = {Elsevier},
abstract = {Due to environmental issues such as climate change, companies are required to optimize their resource and energy consumption in their production process. Predicting power consumption and processing time of all production facilities is essential for manufacturing to develop mechanisms to prevent energy and resource waste and optimize their use. Machine learning is a powerful tool for prediction tasks using data in digitalized environments. In this paper, we present power consumption and processing time prediction of CNC milling machines using five machine learning regression models, i.e., decision tree, random forest, support vector regression (SVR), extreme gradient boosting (XGBoost), and artificial neural network (ANN). Since most of those models are black-box, we applied two explainable artificial intelligence (XAI) approaches, SHapley Additive exPlanations (SHAP) and Local Interpretable Model-Agnostic Explanations (LIME), to give post-hoc explanations of the predictions given by the machine learning models. Our experiments indicated that random forest regression performed the best in predicting power consumption and processing time. The explanation showed that the number of axis rotations and the number of travels to the machine's zero point in rapid traverse were the most important factors that affected the processing time and power consumption. The companies using CNC milling machines can use our prediction models to optimally plan and schedule the operation of the milling machines in a time and energy-efficient manner. They can also optimize the factors that affect power consumption and processing time the most.
},
keywords = {energy management, explainable AI, machine learning, manufacturing},
pubstate = {published},
tppubtype = {article}
}
Due to environmental issues such as climate change, companies are required to optimize their resource and energy consumption in their production process. Predicting power consumption and processing time of all production facilities is essential for manufacturing to develop mechanisms to prevent energy and resource waste and optimize their use. Machine learning is a powerful tool for prediction tasks using data in digitalized environments. In this paper, we present power consumption and processing time prediction of CNC milling machines using five machine learning regression models, i.e., decision tree, random forest, support vector regression (SVR), extreme gradient boosting (XGBoost), and artificial neural network (ANN). Since most of those models are black-box, we applied two explainable artificial intelligence (XAI) approaches, SHapley Additive exPlanations (SHAP) and Local Interpretable Model-Agnostic Explanations (LIME), to give post-hoc explanations of the predictions given by the machine learning models. Our experiments indicated that random forest regression performed the best in predicting power consumption and processing time. The explanation showed that the number of axis rotations and the number of travels to the machine's zero point in rapid traverse were the most important factors that affected the processing time and power consumption. The companies using CNC milling machines can use our prediction models to optimally plan and schedule the operation of the milling machines in a time and energy-efficient manner. They can also optimize the factors that affect power consumption and processing time the most.
2023
Aikenov, Temirlan; Hidayat, Rahmat; Wicaksono, Hendro
Power Consumption and Process Cost Prediction of Customized Products Using Explainable AI: A Case in the Steel Industry Proceedings Article
In: International Conference on Flexible Automation and Intelligent Manufacturing, pp. 1183–1193, Springer Nature Switzerland Cham 2023.
Abstract | Links | BibTeX | Tags: energy management, explainable AI, machine learning, manufacturing, sustainability
@inproceedings{aikenov2023power,
title = {Power Consumption and Process Cost Prediction of Customized Products Using Explainable AI: A Case in the Steel Industry},
author = {Temirlan Aikenov and Rahmat Hidayat and Hendro Wicaksono},
url = {https://link.springer.com/chapter/10.1007/978-3-031-38165-2_135},
doi = {https://doi.org/10.1007/978-3-031-38165-2_135},
year = {2023},
date = {2023-08-25},
urldate = {2023-01-01},
booktitle = {International Conference on Flexible Automation and Intelligent Manufacturing},
pages = {1183–1193},
organization = {Springer Nature Switzerland Cham},
abstract = {Production shifted from a product-centered perspective (mass production of one article) to a customer-centered perspective (mass customization of product variants). It also happens in energy-intensive industries, such as steel production. Mass customization companies face a challenge in accurately estimating the total costs of an individual product. Furthermore, 20% to 40% of the costs related to steel products come from energy. Increasing the product variety can cause an inevitable loss of sustainability. This paper presents machine-learning approaches to improve the sustainability of the steel production industry. It is done by finding the most accurate way to predict the power consumption and the costs of customized products. Moreover, this research also finds the most energy-efficient machine mix based on the predictions. The method is validated in a steel manufacturing Small Medium Enterprise (SME). In this research, experiments were conducted with different machine learning models, and it was found that the most accurate results were achieved using regularization-based and random forest regression models. Explainable AI (XAI) is also used to clarify how product properties influence process costs and power consumption. This paper also discusses scenarios on how the prediction of costs and power consumption can assist production planners in performing workstation selection. This research improves the production planning of customized products by providing a trustable decision support system for machine selection based on explainable machine learning models for process time and power consumption predictions.
},
keywords = {energy management, explainable AI, machine learning, manufacturing, sustainability},
pubstate = {published},
tppubtype = {inproceedings}
}
Production shifted from a product-centered perspective (mass production of one article) to a customer-centered perspective (mass customization of product variants). It also happens in energy-intensive industries, such as steel production. Mass customization companies face a challenge in accurately estimating the total costs of an individual product. Furthermore, 20% to 40% of the costs related to steel products come from energy. Increasing the product variety can cause an inevitable loss of sustainability. This paper presents machine-learning approaches to improve the sustainability of the steel production industry. It is done by finding the most accurate way to predict the power consumption and the costs of customized products. Moreover, this research also finds the most energy-efficient machine mix based on the predictions. The method is validated in a steel manufacturing Small Medium Enterprise (SME). In this research, experiments were conducted with different machine learning models, and it was found that the most accurate results were achieved using regularization-based and random forest regression models. Explainable AI (XAI) is also used to clarify how product properties influence process costs and power consumption. This paper also discusses scenarios on how the prediction of costs and power consumption can assist production planners in performing workstation selection. This research improves the production planning of customized products by providing a trustable decision support system for machine selection based on explainable machine learning models for process time and power consumption predictions.
Almais, Agung Teguh Wibowo; Susilo, Adi; Naba, Agus; Sarosa, Moechammad; Crysdian, Cahyo; Basid, Puspa Miladin NSA; Hariyadi, Mokhamad Amin; Tazi, Imam; Arif, Yunifa Miftachul; Wicaksono, Hendro
SASSD: A Smart Assessment System For Sector Damage Post-Natural Disaster Using Artificial Neural Networks Proceedings Article
In: 2023 2nd International Conference on Computer System, Information Technology, and Electrical Engineering (COSITE), pp. 96–101, IEEE 2023.
Abstract | Links | BibTeX | Tags: data science, machine learning
@inproceedings{almais2023sassd,
title = {SASSD: A Smart Assessment System For Sector Damage Post-Natural Disaster Using Artificial Neural Networks},
author = {Agung Teguh Wibowo Almais and Adi Susilo and Agus Naba and Moechammad Sarosa and Cahyo Crysdian and Puspa Miladin NSA Basid and Mokhamad Amin Hariyadi and Imam Tazi and Yunifa Miftachul Arif and Hendro Wicaksono},
url = {https://ieeexplore.ieee.org/abstract/document/10249540},
doi = {https://doi.org/10.1109/COSITE60233.2023.10249540},
year = {2023},
date = {2023-08-02},
urldate = {2023-08-02},
booktitle = {2023 2nd International Conference on Computer System, Information Technology, and Electrical Engineering (COSITE)},
pages = {96–101},
organization = {IEEE},
abstract = {Smart Assessment System Sector Damage (SASSD) is an intelligent system for assessing the level of sector damage after natural disasters based on Machine Learning (ML) by applying the Artificial Neural Network (ANN) method. SASSD uses forward propagation in ANN. To measure the level of accuracy of the forward propagation algorithm, it is necessary to have a trial method using data pattern modelling. The optimal accrual level value can be achieved by applying 15 data pattern models and changing the structural values of the forward propagation, namely the hidden layer, and epoch. We used 100 training data and 50 testing data at the experimental stage. The training data is the processed data from Decision Support System (DSS), while the training data contains the level of damage to the sector after natural disasters collected by surveyors. The trial results demonstrate the E5 data pattern model’s ideal accuracy rate of 97 percent with a Mean Squared Error (MSE) value of 0.06 and a Mean Absolute Percentage error (MAPE) of 3 percent. This model uses five hidden layers and 125 epochs. The trial results demonstrate the E5 data pattern model’s ideal accuracy rate of 97 % with an MSE value of 0.06 and a MAPE of 3 %. This model uses five hidden layers and 125 epochs. Thus, the SASSD can use the 15th data pattern model (E5) to obtain optimal and accurate results.
},
keywords = {data science, machine learning},
pubstate = {published},
tppubtype = {inproceedings}
}
Smart Assessment System Sector Damage (SASSD) is an intelligent system for assessing the level of sector damage after natural disasters based on Machine Learning (ML) by applying the Artificial Neural Network (ANN) method. SASSD uses forward propagation in ANN. To measure the level of accuracy of the forward propagation algorithm, it is necessary to have a trial method using data pattern modelling. The optimal accrual level value can be achieved by applying 15 data pattern models and changing the structural values of the forward propagation, namely the hidden layer, and epoch. We used 100 training data and 50 testing data at the experimental stage. The training data is the processed data from Decision Support System (DSS), while the training data contains the level of damage to the sector after natural disasters collected by surveyors. The trial results demonstrate the E5 data pattern model’s ideal accuracy rate of 97 percent with a Mean Squared Error (MSE) value of 0.06 and a Mean Absolute Percentage error (MAPE) of 3 percent. This model uses five hidden layers and 125 epochs. The trial results demonstrate the E5 data pattern model’s ideal accuracy rate of 97 % with an MSE value of 0.06 and a MAPE of 3 %. This model uses five hidden layers and 125 epochs. Thus, the SASSD can use the 15th data pattern model (E5) to obtain optimal and accurate results.
Beibit, Rauan; Valilai, Omid Fatahi; Wicaksono, Hendro
Estimating the COVID-19 Impact on the Semiconductor Shortage in the European Automotive Industry using Supervised Machine Learning Proceedings Article
In: Proceedings of the 2023 10th International Conference on Industrial Engineering and Applications, pp. 302–308, 2023.
Abstract | Links | BibTeX | Tags: machine learning, supply chain management, timeseries analysis
@inproceedings{beibit2023estimating,
title = {Estimating the COVID-19 Impact on the Semiconductor Shortage in the European Automotive Industry using Supervised Machine Learning},
author = {Rauan Beibit and Omid Fatahi Valilai and Hendro Wicaksono},
doi = {https://doi.org/10.1145/3587889.3588215},
year = {2023},
date = {2023-06-09},
urldate = {2023-01-01},
booktitle = {Proceedings of the 2023 10th International Conference on Industrial Engineering and Applications},
pages = {302–308},
abstract = {The COVID-19 pandemic impacted different industrial sectors. It causes semiconductor shortage and, subsequently, on the industries downstream, such as the automotive industry. It is because of factory shutdown, increasing consumer electronic demands due to working from home, shifted focus of companies to consumer electronics, and limited logistic capacity. This research aims to analyze the influencing factors and estimate the extent of the impact of COVID-19 on the semiconductor and automotive industry in Europe using machine learning. We developed five regression models to predict the semiconductor sales and number of new passenger car registrations that reflect the development of sales in the automotive industry. Our research revealed that random forest regression is the best machine learning model for analyzing the relationship between COVID-19, semiconductor sales, and passenger car registrations. However, overall, our research found that the COVID-19 pandemic is not the only factor that impacts the semiconductor shortage in the automotive industry. The geopolitical landscape and the world’s reliance on Chinese exports are also essential influencing factors in many supply chains, including in the semiconductor and automotive sectors.
},
keywords = {machine learning, supply chain management, timeseries analysis},
pubstate = {published},
tppubtype = {inproceedings}
}
The COVID-19 pandemic impacted different industrial sectors. It causes semiconductor shortage and, subsequently, on the industries downstream, such as the automotive industry. It is because of factory shutdown, increasing consumer electronic demands due to working from home, shifted focus of companies to consumer electronics, and limited logistic capacity. This research aims to analyze the influencing factors and estimate the extent of the impact of COVID-19 on the semiconductor and automotive industry in Europe using machine learning. We developed five regression models to predict the semiconductor sales and number of new passenger car registrations that reflect the development of sales in the automotive industry. Our research revealed that random forest regression is the best machine learning model for analyzing the relationship between COVID-19, semiconductor sales, and passenger car registrations. However, overall, our research found that the COVID-19 pandemic is not the only factor that impacts the semiconductor shortage in the automotive industry. The geopolitical landscape and the world’s reliance on Chinese exports are also essential influencing factors in many supply chains, including in the semiconductor and automotive sectors.
Almais, Agung Teguh Wibowo; Susilo, Adi; Naba, Agus; Sarosa, Moechammad; Crysdian, Cahyo; Tazi, Imam; Hariyadi, Mokhamad Amin; Muslim, Muhammad Aziz; Basid, Puspa Miladin Nuraida Safitri Abdul; Arif, Yunifa Miftachul; others,
Principal Component Analysis-Based Data Clustering for Labeling of Level Damage Sector in Post-Natural Disasters Journal Article
In: IEEE Access, vol. 11, pp. 74590-74601, 2023.
Abstract | Links | BibTeX | Tags: data science, machine learning
@article{almais2023principal,
title = {Principal Component Analysis-Based Data Clustering for Labeling of Level Damage Sector in Post-Natural Disasters},
author = {Agung Teguh Wibowo Almais and Adi Susilo and Agus Naba and Moechammad Sarosa and Cahyo Crysdian and Imam Tazi and Mokhamad Amin Hariyadi and Muhammad Aziz Muslim and Puspa Miladin Nuraida Safitri Abdul Basid and Yunifa Miftachul Arif and others},
url = {https://ieeexplore.ieee.org/abstract/document/10123944},
doi = {https://doi.org/10.1109/ACCESS.2023.3275852},
year = {2023},
date = {2023-05-12},
urldate = {2023-01-01},
journal = {IEEE Access},
volume = {11},
pages = {74590-74601},
publisher = {IEEE},
abstract = {Post-disaster sector damage data is data that has features or criteria in each case the level of damage to the post-natural disaster sector data. These criteria data are building conditions, building structures, building physicals, building functions, and other supporting conditions. Data on the level of damage to the post-natural disaster sector used in this study amounted to 216 data, each of which has 5 criteria for damage to the post-natural disaster sector. Then PCA is used to look for labels in each data. The results of these labels will be used to cluster data based on the value scale of the results of data normalization in the PCA process. In the data normalization process at PCA, the data is divided into 2 components, namely PC1 and PC2. Each component has a variance ratio and eigenvalue generated in the PCA process. For PC1 it has a variance ratio of 85.17% and an eigenvalue of 4.28%, while PC2 has a variance ratio of 9.36% and an eigenvalue of 0.47%. The results of data normalization are then made into a 2-dimensional graph to see the data visualization of the results of each main component (PC). The result is that there is 3 data cluster using a value scale based on the PCA results chart. The coordinate value (n) of each cluster is cluster 1 ( $text{n} < 0$ ), cluster 2 ( $0le text{n} < 2$ ), and cluster 3 ( $text{n}ge 2$ ). To test these 3 groups of data, it is necessary to conduct trials by comparing the original target data, there are two experiments, namely testing the PC1 results based on the original target data, and the PC2 results based on the original target data. The result is that there are 2 updates, the first is that the distribution of PC1 data is very good when comparing the distribution of data with PC2 in grouping data, because the eigenvalue of PC1 is greater than that of PC2. While second, the results of testing the PC1 data with the original target data produce good data grouping, because the original target data which has a value of 1 (slightly damaged) occupies the coordinates of group 1 (n < 0), the original target data which has a value of 2 (moderately damaged) occupies group 2 coordinates ( $0le text{n} < 2$ ), and for the original target data the value 3 (heavily damaged) occupies group 3 coordinates ( $text{n}ge 2$ ). Therefore, it can be concluded that PCA, which so far has been used by many studies as feature reduction, this study uses PCA for labeling unsupervised data so that it has appropriate data labels for further processing.},
keywords = {data science, machine learning},
pubstate = {published},
tppubtype = {article}
}
Post-disaster sector damage data is data that has features or criteria in each case the level of damage to the post-natural disaster sector data. These criteria data are building conditions, building structures, building physicals, building functions, and other supporting conditions. Data on the level of damage to the post-natural disaster sector used in this study amounted to 216 data, each of which has 5 criteria for damage to the post-natural disaster sector. Then PCA is used to look for labels in each data. The results of these labels will be used to cluster data based on the value scale of the results of data normalization in the PCA process. In the data normalization process at PCA, the data is divided into 2 components, namely PC1 and PC2. Each component has a variance ratio and eigenvalue generated in the PCA process. For PC1 it has a variance ratio of 85.17% and an eigenvalue of 4.28%, while PC2 has a variance ratio of 9.36% and an eigenvalue of 0.47%. The results of data normalization are then made into a 2-dimensional graph to see the data visualization of the results of each main component (PC). The result is that there is 3 data cluster using a value scale based on the PCA results chart. The coordinate value (n) of each cluster is cluster 1 ( $text{n} < 0$ ), cluster 2 ( $0le text{n} < 2$ ), and cluster 3 ( $text{n}ge 2$ ). To test these 3 groups of data, it is necessary to conduct trials by comparing the original target data, there are two experiments, namely testing the PC1 results based on the original target data, and the PC2 results based on the original target data. The result is that there are 2 updates, the first is that the distribution of PC1 data is very good when comparing the distribution of data with PC2 in grouping data, because the eigenvalue of PC1 is greater than that of PC2. While second, the results of testing the PC1 data with the original target data produce good data grouping, because the original target data which has a value of 1 (slightly damaged) occupies the coordinates of group 1 (n < 0), the original target data which has a value of 2 (moderately damaged) occupies group 2 coordinates ( $0le text{n} < 2$ ), and for the original target data the value 3 (heavily damaged) occupies group 3 coordinates ( $text{n}ge 2$ ). Therefore, it can be concluded that PCA, which so far has been used by many studies as feature reduction, this study uses PCA for labeling unsupervised data so that it has appropriate data labels for further processing.
2022
Wicaksono, Hendro; Yuce, Baris; McGlinn, Kris; Calli, Ozum
Smart cities and buildings Book Section
In: Buildings and Semantics, pp. 25, CRC Press, 2022.
BibTeX | Tags: machine learning, ontologies, semantic web, smart cities, sustainability
@incollection{wicaksono2022smart,
title = {Smart cities and buildings},
author = {Hendro Wicaksono and Baris Yuce and Kris McGlinn and Ozum Calli},
year = {2022},
date = {2022-01-01},
urldate = {2022-01-01},
booktitle = {Buildings and Semantics},
pages = {25},
publisher = {CRC Press},
keywords = {machine learning, ontologies, semantic web, smart cities, sustainability},
pubstate = {published},
tppubtype = {incollection}
}
2021
Wicaksono, Hendro
Data Driven Manufacturing: Challenges and Opportunities Presentation
01.01.2021.
BibTeX | Tags: data science, machine learning, manufacturing
@misc{wicaksono2021data,
title = {Data Driven Manufacturing: Challenges and Opportunities},
author = {Hendro Wicaksono},
year = {2021},
date = {2021-01-01},
urldate = {2021-01-01},
publisher = {OSF},
keywords = {data science, machine learning, manufacturing},
pubstate = {published},
tppubtype = {presentation}
}
Wicaksono, Hendro
Accelerating Energy Transition to Green Electricity through Artificial Intelligence Journal Article
In: 2021.
BibTeX | Tags: artificial intelligence, energy management, machine learning, ontologies, semantic web, sustainability
@article{wicaksono2021accelerating,
title = {Accelerating Energy Transition to Green Electricity through Artificial Intelligence},
author = {Hendro Wicaksono},
year = {2021},
date = {2021-01-01},
urldate = {2021-01-01},
publisher = {OSF Preprints},
keywords = {artificial intelligence, energy management, machine learning, ontologies, semantic web, sustainability},
pubstate = {published},
tppubtype = {article}
}
Fritz, Simon; Srikanthan, Vethiga; Arbai, Ryan; Sun, Chenwei; Ovtcharova, Jivka; Wicaksono, Hendro
Automatic information extraction from text-based requirements Journal Article
In: Int. J. Knowl. Eng, vol. 7, no. 1, pp. 8–13, 2021.
BibTeX | Tags: machine learning, natural language processing
@article{fritz2021automatic,
title = {Automatic information extraction from text-based requirements},
author = {Simon Fritz and Vethiga Srikanthan and Ryan Arbai and Chenwei Sun and Jivka Ovtcharova and Hendro Wicaksono},
year = {2021},
date = {2021-01-01},
urldate = {2021-01-01},
journal = {Int. J. Knowl. Eng},
volume = {7},
number = {1},
pages = {8–13},
keywords = {machine learning, natural language processing},
pubstate = {published},
tppubtype = {article}
}
Wicaksono, Hendro; Boroukhian, Tina; Bashyal, Atit
A demand-response system for sustainable manufacturing using linked data and machine learning Book Section
In: Dynamics in Logistics: Twenty-Five Years of Interdisciplinary Logistics Research in Bremen, Germany, pp. 155–181, Springer International Publishing Cham, 2021.
BibTeX | Tags: energy management, machine learning, ontologies, semantic web
@incollection{wicaksono2021demand,
title = {A demand-response system for sustainable manufacturing using linked data and machine learning},
author = {Hendro Wicaksono and Tina Boroukhian and Atit Bashyal},
year = {2021},
date = {2021-01-01},
urldate = {2021-01-01},
booktitle = {Dynamics in Logistics: Twenty-Five Years of Interdisciplinary Logistics Research in Bremen, Germany},
pages = {155–181},
publisher = {Springer International Publishing Cham},
keywords = {energy management, machine learning, ontologies, semantic web},
pubstate = {published},
tppubtype = {incollection}
}
2020
Wicaksono, Hendro
Data analytics in supply chain management Journal Article
In: 2020.
BibTeX | Tags: data science, machine learning, supply chain management
@article{wicaksono2020data,
title = {Data analytics in supply chain management},
author = {Hendro Wicaksono},
year = {2020},
date = {2020-01-01},
urldate = {2020-01-01},
publisher = {OSF},
keywords = {data science, machine learning, supply chain management},
pubstate = {published},
tppubtype = {article}
}