Fakultät Maschinenbau
Permanent URI for this collection
Browse
Recent Submissions
Item Approach to reference models for building performance simulation: establishing common understanding(2023-03-17) Arora, Sahil-Jai; Rabe, MarkusIn the fields of business process modeling, logistics, and information model development, Reference Models (RMs) have shown to enhance standardization, support the common understanding of terminology and procedures, reduce the modeling efforts and cost through the paradigm “Design by Reuse”, and enable knowledge transfer. Utilizing RMs in Building Performance Simulation (BPS) shows potential to achieve similar benefits. However, there is no universally agreed understanding of RMs. In a previous scientific publication, we provided a comprehensive overview of the diversely interpreted definitions, benefits, and attributes of RMs and related terms. Additionally, to transfer the approach of RMs to BPS, a definition for RMs applicable to BPS has been provided, and the identified RM qualities were matched with BPS’s challenges. However, a sound evaluation of the success of transferring RMs to BPS is lacking. Therefore, this scientific contribution firstly includes the analysis conducted in the previous scientific contribution constituting a common understanding about RMs and their elements for BPS. Secondly, by conducting expert interviews, the applicability and validity of the developed concept of RMs for BPS are surveyed. In total, ten experts (seven BPS experts and three RM experts) evaluated the quality of creating transparency about the understanding of RMs and the level of success of their transfer toward BPS. The experts consistently see a great benefit of RMs in BPS, but for BPS experts the transfer and possible application of RMs in BPS is not sufficiently clear. Accordingly, the key output of the conducted survey is that a clearer and more detailed application example, e.g., describing at a more easy-to-understand level of detail an exemplary class of the provided example of an RM, is required for a more profound transfer of RMs to BPS.Item Demystifying reinforcement learning approaches for production scheduling(2023) Rinciog, Alexandru; Meyer, Anne; Liebig, ThomasRecent years has seen a sharp rise in interest pertaining to Reinforcement Learning (RL) approaches for production scheduling. This is because RL is seen as a an advantageous compromise between the two most typical scheduling solution approaches, namely priority rules and exact approaches. However, there are many variations of both production scheduling problems and RL solutions. Additionally, the RL production scheduling literature is characterized by a lack of standardization, which leads to the field being shrouded in mysticism. The burden of showcasing the exact situations where RL outshines other approaches still lies with the research community. To pave the way towards this goal, we make the following four contributions to the scientific community, aiding in the process of RL demystification. First, we develop a standardization framework for RL scheduling approaches using a comprehensive literature review as a conduit. Secondly, we design and implement FabricatioRL, an open-source benchmarking simulation framework for production scheduling covering a vast array of scheduling problems and ensuring experiment reproducibility. Thirdly, we create a set of baseline scheduling algorithms sharing some of the RL advantages. The set of RL-competitive algorithms consists of a Constraint Programming (CP) meta-heuristic developed by us, CP3, and two simulation-based approaches namely a novel approach we call Simulation Search and Monte Carlo Tree Search. Fourth and finally, we use FabricatioRL to build two benchmarking instances for two popular stochastic production scheduling problems, and run fully reproducible experiments on them, pitting Double Deep Q Networks (DDQN) and AlphaGo Zero (AZ) against the chosen baselines and priority rules. Our results show that AZ manages to marginally outperform priority rules and DDQN, but fails to outperform our competitive baselines.Item A scalable machine learning system for anomaly detection in manufacturing(2023) Schlegl, Thomas; Deuse, Jochen; Müller, RainerBerichte über Rückrufaktionen in der Automobilindustrie gehören inzwischen zum medialen Alltag. Tatsächlich hat deren Häufigkeit und die Anzahl der betroffenen Fahrzeuge in den letzten Jahren weiter zugenommen. Die meisten Aktionen sind auf Fehler in der Produktion zurückzuführen. Für die Hersteller stellt neben Verbesserungen im Qualitätsmanagement die intelligente und automatisierte Analyse von Produktionsprozessdaten ein bislang kaum ausgeschöpftes Potential dar. Die technischen Herausforderungen sind jedoch enorm: die Datenmengen sind gewaltig und die für einen Fehler charakteristischen Datenmuster zwangsläufig unbekannt. Der Einsatz maschineller Lernverfahren (ML) ist ein vielversprechender Ansatz um diese Suche nach der sinnbildlichen Nadel im Häuhaufen zu ermöglichen. Algorithmen sollen anhand der Daten selbständig lernen zwischen normalem und auffälligem Prozessverhalten zu unterscheiden um Prozessexperten frühzeitig zu warnen. Industrie und Forschung versuchen bereits seit Jahren solche ML-Systeme im Produktionsumfeld zu etablieren. Die meisten ML-Projekte scheitern jedoch bereits vor der Produktivphase bzw. verschlingen enorme Ressourcen im Betrieb und liefern keinen wirtschaftlichen Mehrwert. Ziel der Arbeit ist die Entwicklung eines technischen Frameworks zur Implementierung eines skalierbares ML-System für die Anomalieerkennung in Prozessdaten. Die Trainingsprozesse zum Initialisieren und Adaptieren der Modelle sollen hochautomatisierbar sein um einen strukturierten Skalierungsprozess zu ermöglichen. Das entwickelt DM/ML-Verfahren ermöglicht den langfristigen Aufwand für den Systembetrieb durch initialen Mehraufwand für den Modelltrainingsprozess zu senken und hat sich in der Praxis als sowohl relativ als auch absolut Skalierbar bewährt. Dadurch kann die Komplexität auf Systemebene auf ein beherrschbares Maß reduziert werden um einen späteren Systembetrieb zu ermöglichen.