Data-Driven Collaborative Decision Making in Complex Industrial Systems

Group leader
Hendro Wicaksono
Data-Driven and Collaborative Decision Making in Complex Industrial Systems
Specific themes and goals
  • Semantic modeling: An Ontology or Knowledge Graph is an information model, which integrates different schema and data sources. We apply a W3C standard, called The OWL (Web Ontology Language), as a model representation, and this serves as a shared vocabulary and schema among the heterogeneous systems and humans. Thus, it creates a model that machines and humans can understand. It provides flexibility to express statements described in description logic and integrate first-order logic. Therefore, logical reasonings, including fuzzy reasoning, are also possible in the knowledge model.
  • Semantic extraction: Semantic uplifting and linking extracts the semantics of exchanged data, which have various formats such as structured, semi-structured, natural language texts, and images. A single industrial data exchange standard cannot fully fulfill an application’s data requirements in complex industrial systems. There is increased interest and research on sharing data and schemas from different domains on the internet (open data). We developed methods to reuse linked data and link available data and schemas.
  • Machine learning and knowledge extraction models: Such as linear, non-linear, ensemble, time series, and deep learning models — identify interesting patterns in data. Using these models, we extract the information or knowledge required to solve particular problems, convert the information into ontology elements or rules, integrate them with formalized human knowledge, and store it in a knowledge graph.
  • Explainable artificial intelligence (XAI): Explainable AI (XAI) aims to improve the explainability, interpretability, and usability of machine learning models generated from data so that people unfamiliar with AI can use the models to make decisions. It links machine learning models to the knowledge graph. Our research also focuses on modeling and extracting causal relationships among data instead of correlations to enable causal inferences.
  • Data-driven simulation and optimization: We have developed different approaches for optimization problems, such as production scheduling, logistics, and product configuration. Current approaches are not able to solve large problems timeously. Meta-heuristics, by contrast, work on general models that do not correspond to reality. For this reason, we developed a hyper-heuristic approach, creating a flexible but real problem-relevant model. Furthermore, we develop a deep reinforcement approach to provide generic and scalable solutions for industrial optimization problems.
  • Applications: We apply our approaches in various complex industrial settings. For example, we optimized renewable energy sources in manufacturing processes; assessed risk and optimized processes in the automotive supply chain; and developed automatic assessment and optimisation tools for environment and social government (ESG) in manufacturing and supply chain. We also assessed and improved the explainability and deployment of AI models in agri-food supply chain, as well as for industry 4.0 maturity and architecture implementations. Other applications included transforming the automotive supply chain towards the circular economy; and developing a data-driven collaborative platform for smart cities and digital twin creation.
Highlights and impact

We have been collaborating with various large, small, and medium enterprises across Europe in our research projects. The collaboration projects have been funded by the German Federal Ministry of Economics and Climate Protection (BMWK), and the German Federal Ministry for Digital and Transport (BMDV). Our role is to support enterprises, especially small and medium enterprises (SMEs), to develop innovations and solve their operational problems. For example, we developed semantic middleware that facilitates energy-related data exchanges between manufacturing SMEs and utility companies. We also worked on machine-learning models to forecast power consumption and generation for setting up dynamic electricity pricing. Furthermore, we are also developing a digital-twin platform to facilitate digital model exchanges among SMEs. We are also collaborating with a large automotive manufacturer to assess supply chain risks and optimize the supply chain processes using causal AI. 

Our works in semantic linking and knowledge graph contribute to open source energy management solutions (OpenEMS). We were also one of the initiators of the W3C LBD community group and are currently active as a member. Prof. Hendro Wicaksono has been a visiting professor in six different universities in Indonesia, and since 2018, he has supervised 146 Master and Bachelor theses.

Group composition & projects/funding

Currently, we have eight PhD students, two nonPhD research associates, and five student teaching and research assistants, and two PhD students. The group receives funding from BMWK, BMDV, and the private sector, among others.

Selected publications
  • Wicaksono, H.; Yuce, B.; McGlinn, K. (2022): Smart Cities and Buildings, in Buildings and Seman - tics. CRC Press. 
  • Wicaksono, H.; Boroukhian, T.; Bashyal, A. (2021): A Demand-Response System for Sustainable Manufacturing Using Linked Data and Machine Learning, Dynamics in Logistics. Springer, Cham, 2021. 
  • Farooq, Y.; Wicaksono, H. (2021). Advancing on the analysis of causes and consequences of green skepticism: causal inference methods. Journal of Cleaner Production. 
  • Fajrul Falah, M.; Sukaridhoto, S.; Udin Harun Al Rasyid, M.; Wicaksono, H. (2020): Design of Virtual Engineering and Digital Twin Platform as Implementation of Cyber Physical Systems, Procedia Manufacturing. 
  • Schneider G.F.; Wicaksono, H.; Ovtcharova, J. (2019): Virtual Engineering of Cyber-Physical Automation Systems: The Case of Control Logic, Advanced Engineering Informatics.