2.2 Prozesssimulation
Filtern
Dokumenttyp
- Vortrag (10)
- Posterpräsentation (8)
- Zeitschriftenartikel (6)
- Forschungsdatensatz (1)
Schlagworte
- Mixed Reality (5)
- Process Design (5)
- DEXPI (4)
- Parameter Estimation (4)
- Distillation Column (3)
- Generalized Disjunctive Programming (3)
- NRTL (3)
- Process Optimization (3)
- BPMN (2)
- Data Integration (2)
Organisationseinheit der BAM
Paper des Monats
- ja (2)
Eingeladener Vortrag (wissenschaftliche Konferenzen)
- nein (10)
Integration of Information Models Using Semantic Web Technologies Within the Mixed Reality Data Hub
(2026)
Engineering information is distributed across heterogeneous data models, ontologies, and reference data libraries (RDLs), which complicates interoperability, hinders workflow automation, and limits the deployment of industrial Mixed Reality (MR) applications. This work uses DEXPI Process semantics and Semantic Web technologies to align and integrate these heterogeneous sources into a single knowledge graph. The resulting graph provides a queryable semantic layer that enables consistent cross-source access to engineering information and supports MR application development. Beyond integration, the semantic graph enables automation of process design workflows. We demonstrate this using a basic engineering use case in which alternative flowsheet design options are generated from a semantically modeled block flow diagram (BFD).
Geräte und Maschinen, die zur bestimmungsgemäßen Verwendung in explosionsgefährdeten Bereichen gemäß 2014/34/EU konstruktiv vorgesehen sind, muss in der europäischen Union eine Zündgefahrenbewertung durchgeführt werden. Dabei müssen alle 13 Zündquellen betrachtet werden. Eine Teilmenge der 13 Zündquellen sind die Gefahren von nichtelektrischen Zündquellen, zu denen auch die mechanischen Schlag-, Schleif- und Reibvorgänge gehören.
Bei mechanischen Schlagvorgängen kommt es infolge des Zusammenstoßes zweier Werkstücke bzw. Bauteile zu einer Umwandlung der kinetischen Energie. Dabei erhöht sich die Temperatur der Werkstoffe an der Kontaktstelle und es kommt unter Umständen zu einem Abtrennvorgang kleiner Partikel erhöhter Temperatur. Sowohl die heißen Kontaktstellen (Zündquelle "heiße Oberflächen") als auch die abgetrennten Partikel (Zündquelle "mechanisch erzeugte Funken") können eine wirksame Zündquelle für ein explosionsfähiges Gasgemisch darstellen.
Zur Festlegung von Grenzwerten wurden in der Norm DIN EN ISO 80079-36:2016 die Gasgemische anhand ihrer Explosionsgruppe klassifiziert und zu jeder Gruppe die maximale Energie des Schlagvorgangs festgelegt, unter derer die Entstehung einer wirksamen Zündquelle als unwahrscheinlich angenommen werden kann. Weitere Festlegungen von Grenzwerten für die kinetische Schlagenergie und geeignete Werkstoffpaarungen finden sich u.a. in der DIN EN 1755 (ex-geschützte Flurförderzeuge) und in der DIN EN 14986 (ex-geschützte Ventilatoren).
This poster summarizes a comparative study of multi-fidelity modeling approaches for Bayesian optimization, focusing on how low-fidelity data can be systematically integrated with high-fidelity evaluations to improve optimization efficiency. Several Gaussian process–based methods, including multi-task models, prior mean approaches, and Kennedy–O’Hagan–type formulations, are evaluated on standard benchmark functions. The results highlight strengths and limitations of each approach, emphasizing challenges such as robustness of training, nonlinear inter-fidelity relationships, and potential bias introduced by prior mean models. The poster was presented at ESCAPE35 and subsequently at the BAM internal AI Symposium.
Der Vortrag stellt das Projekt „Handbuch2AR“ vor, das einen KI-gestützten, kontextbasierten Zugriff auf technische Dokumentation für Betriebs-, Wartungs- und Instandsetzungsaufgaben adressiert. Vorgestellt wird ein prototypischer Ansatz auf Basis von Retrieval-Augmented Generation (RAG), der Prozessmodelle der Wartungsaufgabe mit relevanten Inhalten aus Handbüchern verknüpft, um situationsabhängige Informationen bereitzustellen. Anhand mehrerer industrieller Anwendungsfälle wird der aktuelle Entwicklungsstand sowie das geplante Vorgehen zur qualitativen Evaluation der Ergebnisse erläutert.
Operation and maintenance tasks in an industrial context are complex and often time- consuming processes. The manufacturer-provided manuals form the foundation for these processes. However, finding the required information from manuals remains an arduous task. During a maintenance task, the relevant information from plant documentation should be directly available. Along the individual process steps, questions may arise that require information from documentation. To make plant documentation queryable, a Retrieval Augmented Generation (RAG) system was developed. This technique is ideal for cases with little training data, in which fine-tuning of large language models (LLM) is ill-advised. The plant documentation serves as an external knowledge base for answering incoming questions with links to the original documents. Before the RAG development, comprehensive pre-processing of documents takes place, which here is divided into three main blocks: (1) Bill of materials are recognized, page references are stored, and tables are extracted. (2) Technical drawings are also recognized, their page reference is saved, and then passed to a multi-modal LLM (MMLLM), which extracts title and drawing number. (3) The manual is divided into chapters so that each query can be thematically assigned to one or more chapters where appropriate answers are sought. To provide further context, the operation and maintenance tasks are modeled using Business Process Model and Notation (BPMN), an international standard for representation of business processes. As a case study, the maintenance of an industrial pump is used. Along the maintenance process, questions are passed to LLM and RAG to provide concise answers.
Der Einsatz von Mixed-Reality-(MR)-Anwendungen in der chemischen Prozessindustrie bietet das Potenzial für sicherere und effizientere Prozessauslegung, Anlagenwartung und Betriebsabläufe sowie für eine vereinfachte Schulung von Bedienpersonal. In der Praxis wird die Umsetzung jedoch häufig durch hohen Backend- und Frontend-Entwicklungsaufwand, die heterogene Landschaft an Datenquellen (z. B. Automatisierungssysteme und Engineering-Daten) sowie Anforderungen an Datenschutz, Pseudonymisierung und Rechtemanagement erschwert – insbesondere bei kleinen und mittleren Unternehmen. Diese Arbeit stellt den MR-Datenhub als wiederverwendbares Entwicklungs-Framework und Integrationsrückgrat für industrielle MR-Anwendungen vor. Der Ansatz kombiniert (1) ein vorkonfiguriertes, wiederverwendbares Backend- und Frontend-Framework, (2) standardisierte Schnittstellen zur Datenintegration sowie (3) einen Pseudonymisierungsdienst für personenbezogene und potenziell gesundheitsbezogene Daten, um DSGVO-konforme MR-Anwendungen zu unterstützen. Der Datenhub setzt auf eine GraphQL-basierte Serverarchitektur und ermöglicht die systematische Modellierung und Anbindung relevanter Datenquellen. Ergänzend wird eine Datenlandkarte vorgestellt, die domänenspezifische Standards und technische Frameworks systematisch erfasst, einordnet und strukturiert. Sie dokumentiert zudem die Mappings zwischen den Standards und spezifiziert deren Zuordnungsregeln. Der MR-Datenhub wird bereits in Projekten wie MR4SafeOperations und KeepCool eingesetzt und dient dort als Integrationsbasis für MR-gestützte Arbeitsabläufe.
BPMN 2.0 is a widely adopted standard across various industries, primarily used for business process management outside of the engineering sphere [1]. Its long history and widespread use have contributed to a mature ecosystem, offering advanced software tools for editing and optimizing business workflows.
DEXPI Process, a newly developed standard for early-phase chemical process design, focuses on representing Block Flow Diagrams (BFDs) and Process Flow Diagrams (PFDs), both crucial in the conceptual design phase of chemical plants. It provides a standardized way to document design activity, offering engineers a clear rationale for design decisions [2], which is especially valuable during a plant’s operational phases. While DEXPI Process offers a robust data model, it currently lacks an established serialization format for efficient data exchange. As Cameron et al. note in [2], finding a suitable format for DEXPI Process remains a key research area, essential for enhancing its usability and adoption. So far, Cameron et al. have explored several serialization formats for exchanging DEXPI Process information, including AutomationML, an experimental XML, and UML [2].
This work aims to map the DEXPI Process data model to BPMN 2.0, providing a standardized serialization for the newly developed standard. Mapping DEXPI Process to BPMN 2.0 also unlocks access to BPMN’s extensive software toolset. We investigate and validate the effectiveness of this mapping and the enhancements it brings to the usability of DEXPI Process through a case study based on the Tennessee-Eastman process, described in [3]. We then compare our approach with those of Cameron et al. in [2].
We conclude by presenting our findings and the key benefits of this mapping, such as improved interoperability and enhanced toolset support for chemical process engineers. Additionally, we discuss the challenges encountered during the implementation, including aligning the differences in data structures between the two models. Furthermore, we believe this mapping serves as a bridge between chemical process design engineers and business process management teams, unlocking opportunities for better collaboration and integration of technical and business workflows.
Usage of mixed reality (MR) applications in the chemical process industry can offer safer and more efficient process design, plant maintenance, and operational workflows, as well as easier operator training. Despite this potential, the development of MR applications is still slow and often impeded by high costs – especially for small and medium-sized companies. To this end, this contribution presents the MR DataHub, a framework that simplifies MR application development by providing a pre-configured backend environment for data integration and by automating frontend code generation.
The MR DataHub consists of a containerized backend server that enables connections to external data sinks and sources through standard interfaces such as REST API, MQTT, and OPC UA. Hence, it can link automation systems, enterprise resource management, and / or engineering data to the MR frontend environment. It enables the exchange of live process data and control commands, providing the MR application with essential operational information. Based on the linked sinks and sources, the MR DataHub automatically generates type and class definitions for a frontend Unity engine. Furthermore, MR DataHub offers predefined, extensible GraphQL schemas for various industry domain standards, including DEXPI P&ID and DEXPI Process, enabling faster integration of standardized data into industrial MR applications.
The benefits of the MR DataHub are demonstrated in the MR4SafeOperations project (https://mr4b.de/wir-buendnis-mr4b-projekte/mr4safeoperations/), where an MR application was developed using the MR DataHub to guide plant operators in liquid sampling procedures by providing step-by-step visual guidance integrated with live process and engineering data – enhancing operational safety. Preliminary evaluations indicate significant reductions in development time of the MR application by relying on MR DataHub’s pre-configured backend environment, predefined interfaces, and extensible GraphQL schemas.
An ML approach for parameter estimation of dynamic models is proposed, in which Time Series Extrinsic Regression (TSER) is used to learn the inverse mapping of the time series output to the underlying time-invariant parameters. To our knowledge, this is the first deliberate application of TSER and its methods to parameter estimation of dynamic models.
Dynamic real time optimization of chemical processes requires fast models. So, data-driven surrogate models are often used. However, these models do not contain information on the feasible region of the underlying rigorous model. Therefore, the data-driven regressor is combined with an additional classification model to prevent faulty extrapolations.