AI-Augmented Model-Based Systems Engineering

Ruslan Bernijazov; Roman Dumitrescu; Fabian Hanke; Oliver von Heißen; Lydia Kaiser; Denis Tissen

doi:10.1515/zwf-2024-0123

Enjoy 40% off

academic books on De Gruyter Brill *

Article Open Access

AI-Augmented Model-Based Systems Engineering

Ruslan Bernijazov
Ruslan Bernijazov, born in 1990) studied computer science (M.Sc.) at the University of Paderborn and has been working on systems engineering, the use of AI in engineering and the development of AI-based systems for almost ten years. He is currently responsible for systems engineering in the Advanced Systems Engineering department at the Heinz Nixdorf Institute at the University of Paderborn and is managing director of AI Marketplace GmbH in Paderborn.
, Roman Dumitrescu
Prof. Dr.-Ing. Roman Dumitrescu is director at the Fraunhofer Institute for Mechatronics Design (IEM) in Paderborn and head of the Advanced Systems Engineering department at the Heinz Nixdorf Institute at the University of Paderborn. After his studies and doctorate, he took on various responsible tasks, including as managing director of the technology network Intelligent Technical Systems OstWestfalen-Lippe (it’s OWL) and as initiator of the tech startups Two Pillars GmbH and AI Marketplace GmbH. He is also involved in various advisory committees, including as a member of the German Academy of Engineering Sciences (acatech)
, Fabian Hanke
Fabian Hanke, born in 1994, studied mechanical engineering (M.Sc.) at the University of Paderborn. Since February 2022, he has been a research associate in the Digital Engineering department at the Fraunhofer Institute for Mechatronics Design (IEM) in Paderborn.
, Oliver von Heißen
Oliver von Heißen, born in 1994n studied computer science (M.Sc.) at the University of Paderborn. Since July 2022, he has been a research associate in the Digital Engineering department at the Fraunhofer Institute for Design Technology and Mechatronics (IEM) in Paderborn.
, Lydia Kaiser
Prof. Dr.-Ing. Lydia Kaiser, born in 1982, is head of the Digital Engineering 4.0 department at the Technical University of Berlin and the Einstein Center Digital Future. She is also a member of the extended board of the Society for Systems Engineering. After studying physics and earning her doctorate in mechanical engineering at the University of Paderborn, she took on various management positions at Fraunhofer IEM in Paderborn.
and Denis Tissen
Denis Tissen, born in 1994 studied mechanical engineering (M.Sc.) at the University of Paderborn. Since September 2020 he has been a research assistant in the Advanced Systems Engineering department at the Heinz Nixdorf Institute at the University of Paderborn.

Published/Copyright: March 27, 2025

Published by

Become an author with De Gruyter Brill

Submit Manuscript Author Information

From the journal Zeitschrift für wirtschaftlichen Fabrikbetrieb Volume 120 Issue s1

Abstract

The increasing complexity of modern technical systems necessitates innovative approaches such as Model-Based Systems Engineering (MBSE). In this context, using Artificial Intelligence (AI) emerges as a key enabler for practical application and efficiency improvement. This article introduces a maturity model for AI-based assistance systems in MBSE. It helps companies assess their current automation level in MBSE activities, providing a foundation for strategic planning of process improvements.

Abstract

Die zunehmende Komplexität moderner technischer Systeme erfordert innovative Ansätze wie das Model-Based Systems Engineering (MBSE). In diesem Zusammenhang wird der Einsatz von Künstlicher Intelligenz (KI) zu einem wichtigen Enabler für die praktische Anwendung und Effizienzsteigerung von MBSE. Dieser Beitrag präsentiert ein Reifegradmodell für KI-basierte Assistenzsysteme im MBSE, das Unternehmen dabei hilft, den Automatisierungsgrad ihrer MBSE-Aktivitäten zu bewerten und eine Grundlage für die strategische Planung von Prozessverbesserungen bietet.

Keywords: Model-Based Systems Engineering; Digital Engineering; Artificial Intelligence; Systems Engineering; Assistant Systems

Schlüsselwörter: Modellbasierte Systementwicklung; Digitale Produktentstehung; Künstliche Intelligenz; Assistenzsysteme

Introduction

Advances in digitalization have significantly influenced both society and industry over recent decades. Many new digital technologies have impacted industrial companies, affecting their market offerings and internal business processes [1]. These advancements have led to increased system complexity, necessitating innovative engineering approaches. In this context, Model-Based Systems Engineering (MBSE) has emerged as a key instrument for mastering this complexity. MBSE employs formalized models to simplify complex systems, facilitating a holistic and interdisciplinary perspective of products and services [2]. Despite its numerous advantages, the practical implementation of MBSE in companies encounters significant challenges. High initial costs related to MBSE tools and training, a lack of organizational acceptance, and the inherent complexity of modeling processes hinder its adoption [3].

Artificial Intelligence (AI) presents promising opportunities to address these challenges by automating repetitive tasks and by extracting valuable insights from engineering data through advanced analytics techniques [4, 5]. The integration of MBSE and AI is typically realized through AI-based assistance systems [6] within MBSE processes. While initial research has explored the application of AI in specific areas of MBSE, these studies often focus on isolated aspects of the process or specialized tools [7]. There remains a lack of a comprehensive, systematic framework that enables companies and researchers to classify AI-based assistance systems within MBSE and identify opportunities for their future development.

This article introduces a maturity model for AI-based assistance systems in engineering. The model helps companies assess their current MBSE automation level and is a foundation for strategic planning activities and investments to increase automation within engineering processes. We illustrate the model with practical examples, demonstrating its applicability in real-world MBSE scenarios

State of the Art

Current research on the integration of AI in MBSE shows promising approaches for optimizing development processes. For example, Bretz et al. investigate how AI-based assistance systems can support product development [6]. Tissen et al. extend MBSE with data-driven methods and present data-driven model-based systems engineering (DDMBSE), which uses AI for data-based modeling [8]. Kharatyan et al. analyze the potential of AI in engineering from a human-centered perspective [4]. Schräder et al. present various applications of AI in MBSE [5].

In many application areas, maturity models are used to analyze and categorize AI capabilities systematically. Examples include the 7 Levels of Hybrid Decision-Making from Gartner Research [9] or the ADA-CMM model by Ginger et al. [10]. In the context of MBSE, Kaiser et al. presented a task-based categorization of AI assistants [11]. However, no comprehensive maturity model exists specifically for categorizing AI-based assistance systems based on their capabilities within MBSE. Existing models either address general AI applications or focus on isolated aspects of MBSE, highlighting the need for a specialized framework tailored to the intersection of AI and MBSE.

Methodology

The maturity model was developed using the Design Science Research (DSR) method [12], which involves iterative cycles of evaluation and refinement. The process began with identifying and discussing the need for a maturity model specific to AI-based assistance systems in MBSE. Two researchers with expertise in MBSE and AI created an initial version of the model based on existing literature and maturity models from other domains.

In the first refinement cycle, the initial model was reviewed by two additional researchers. Their feedback was incorporated into subsequent revisions. The second refinement cycle involved a more extensive expert workshop with nine participants from the authors’ organizations. Workshop participants identified misconceptions, inconsistencies, and areas needing further clarification, which were subsequently addressed in the final version of the maturity model.

Maturity Model for AI-Based Assistance Systems in MBSE

In the following, we present our maturity model for AI-based assistance systems in MBSE (Figure 1). The maturity model comprises six levels that classify AI systems according to their degree of autonomy in MBSE. It is inspired by the established Society of Automotive Engineers (SAE) J3016 model for vehicle automation and offers a qualitative classification of AI systems.

Figure 1

Maturity model for AI-based assistance systems in Model-Based System Engineering

To demonstrate the applicability of our maturity model, we apply it to different AI-based assistance systems that support the architecture development process, aiming to create an architecture model for a new system based on specific requirements. A proven method is the Requirements, Functional, Logical, and Physical (RFLP) approach [13], as shown in Figure 2. This approach structures the system architecture into functional, logical, and physical views. The Requirements view captures the system‘s needs, the Functional view outlines the system‘s functions, the Logical view details the system‘s logical components, and the Physical view describes the system‘s physical components.

Figure 2

Architecture development according to RFLP

Level 0: Engineering Tools

At Level 0, engineers perform tasks manually without the support of AI. While digital tools like architecture modeling tools may be utilized, they lack AI-assisted features. Engineers bear full responsibility for all decisions and executions. The dependence on human expertise results in extended development timelines and increases the likelihood of errors due to manual processing. Such challenges can lead to higher costs and reduced efficiency in engineering projects. Level 0 represents the de facto standard in engineering and serves as the baseline scenario.

Level 1: AI-supported Engineering

In the first level, AI supports engineers by handling specific, clearly defined tasks. The primary focus of AI at this stage is to speed up or simplify specific processes by automating recurring routine activities, such as automated translations of models or searches within MBSE tools. However, the execution of individual engineering tasks remains entirely within the engineers‘ remit. The effects on engineering include increased efficiency in routine tasks and reduced burden on engineers from repetitive activities, allowing specialists to focus on more complex and creative aspects of development work.

The MBSE Workshop Assistant [5] exemplifies this level by utilizing computer vision to formalize hand-drawn architectures and convert sketches into SysML models (Figure 3). It features a digital whiteboard for which engineers can input touch, gestures, and pen. Through image and handwriting recognition, the assistant identifies system elements and functions, translates them into formal models, and exports them into a SysML tool. This automation relieves engineers from repetitive formalization tasks and enhances process efficiency by reducing the time required to develop formal system models.

Figure 3

Workflow of the MBSE Workshop Assistant (based on [5])

Level 2: Dedicated Engineering Copilots

In Level 2, specialized AI co-pilots—intelligent systems designed to assist engineers by automating specific engineering activities and providing proactive support—are utilized. While the engineer retains overall responsibility, the AI performs more complex tasks, such as generating functional architectures based on requirements or optimizing designs. This delegation allows engineers to focus on higher-level decision-making and creative problem-solving. The integration of AI at this level significantly accelerates development processes by automating time-consuming tasks and enables more informed decision-making through AI-supported analyses. Consequently, this leads to improved product quality, reduced time-tomarket, and increased overall efficiency in engineering projects.

One example of an AI co-pilot in the architecture creation process is the iQBuddy assistance system [14]. iQBuddy utilizes a knowledge graph to help engineers identify suitable solution patterns. By analyzing existing system designs, iQBuddy offers real-time suggestions for the current architecture development by recognizing and prioritizing recurring patterns. For instance, when an engineer sketches a component, iQBuddy can automatically suggest similar components from previous projects, ensuring consistency and leveraging proven design practices. Figure 4 illustrates the concept of iQBuddy, showcasing how new projects are integrated into the knowledge graph. Model elements are mapped to nodes based on their labels and types, while connections between elements are represented as edges. Priority numbers indicate the frequency of pattern occurrences, enabling iQBuddy to generate relevant recommendations by searching for matching areas in the knowledge graph and utilizing adjacent elements to propose suggestions.

Figure 4

iQBuddy Workflow (based on [14])

Level 3: Continuous AI-integrated Engineering

In the third level, AI is fully integrated into development processes, autonomously handling complex process chains. Engineers retain responsibility by validating and supervising AI-generated results. This integration leads to significant efficiency gains by automating entire process chains, enabling quicker adjustments to new requirements, and enhancing consistency and traceability in system development. Consequently, projects benefit from reduced time-to-market, lower operational costs, and higher quality outcomes due to minimized human error. As AI assumes responsibility for complex tasks, engineers focus on validating AI outputs and contributing strategic expertise.

One example of an AI assistant at this level is the AI4Cameo Assistant described by Heißen et al. [15]. This plugin for the Cameo Systems Modeler automates all RFLP steps using generative AI. It generates functional and logical architectures from requirements, integrating seamlessly into the modeling tool. Engineers input system specifications, which large language models analyze to produce comprehensive architectural models for review and adaptation. This automation reduces manual effort, accelerates development timelines, and ensures the consistency and accuracy of architectural designs.

Level 4: AI-driven Engineering

In the fourth level, AI assistants independently handle extensive development tasks within hybrid teams of humans and AI assistants. While AI performs many tasks autonomously, engineers handle particularly complex or high-stakes tasks. Engineers act as coordinators and strategic decision-makers, defining goals and setting framework conditions while collaborating closely with AI assistants. Together, they develop system components, conduct complex simulations, and optimize the overall design.

Although AI assistants at this level are still emerging, rapid advancements in generative AI suggest feasibility in the following years. For example, AI assistants could jointly develop a complete system architecture—from creating artifacts to validating and verifying them—while engineers ensure alignment with requirements and handle any particularly intricate tasks that exceed AI capabilities.

Level 5: Autonomous AI-Engineering

AI assistants entirely automate development processes at the fifth and highest level, proactively consulting human experts when ethical, strategic, or unpredictable decisions require human input. The AI also proactively informs stakeholders by providing all necessary information in a stakeholder-oriented manner. Human roles shift towards training and managing AI assistants, emphasizing oversight rather than direct involvement in daily tasks. This fundamental change accelerates project realization by streamlining workflows and significantly reducing development timelines. AI‘s ability to efficiently handle complex tasks enables the execution of high-risk projects previously deemed unfeasible due to their inherent risks and resource demands. As AI takes on more autonomous roles, engineers focus on defining project objectives, setting framework conditions, and overseeing AI-driven processes to ensure alignment with organizational goals and ethical standards.

A prospective system at this level may be a platform for AI assistants capable of autonomously developing a comprehensive system architecture based on a predefined requirements specification. The platform would assemble teams of AI assistants tailored to user requests, enabling the execution of complex engineering tasks. These teams would include assistants dedicated to generating relevant architectural artifacts and assistants designed for validation and verification. Engineers would oversee the entire process by defining objectives, providing strategic guidance, and granting final approval to the AI-generated architectures, ensuring that all necessary standards and requirements are satisfactorily met.

Discussion

The maturity model allows organizations to evaluate their current AI automation in MBSE and align stakeholders with strategic objectives. By positioning an organization on the automation spectrum, the model aids in planning and communicating long-term goals and serves as a roadmap for investments in AI-enabled engineering tools and methods.

However, the model is inherently generic, focusing on classifying existing AI-based assistance systems rather than detailing methodologies for developing new ones. This limits its applicability for organizations needing specific guidance on algorithms, data requirements, or AI workflows. Future research should develop actionable engineering practices based on the maturity levels. Additionally, the model was primarily developed and reviewed internally.

External validation by independent experts must enhance its generalizability and applicability across diverse industrial contexts.

Conclusion and Outlook

AI‘s rapid advancement in engineering presents challenges and opportunities for industry. Companies need to engage with AI-based assistance systems to remain competitive and manage the increasing complexity of modern technical systems. By automating routine tasks and accelerating development cycles, AI improves efficiency, fosters engineering innovations, and enhances product quality.

This article presents a maturity model for AI-based assistance systems in MBSE, classifying AI systems by autonomy level to aid organizations in assessing and planning their automation strategies. Future work will focus on refining the model through external validation and applying it in various industrial settings. Additionally, we will explore emerging AI technologies and develop integration methodologies to support the next generation of AI-assisted engineering systems.

Note

This article is peer reviewed by the members of the ZWF Special Issue Advisory Board.

Phone: +49 (0) 5251 60-6233

About the authors

Ruslan Bernijazov

Ruslan Bernijazov, born in 1990) studied computer science (M.Sc.) at the University of Paderborn and has been working on systems engineering, the use of AI in engineering and the development of AI-based systems for almost ten years. He is currently responsible for systems engineering in the Advanced Systems Engineering department at the Heinz Nixdorf Institute at the University of Paderborn and is managing director of AI Marketplace GmbH in Paderborn.

Prof. Dr.-Ing. Roman Dumitrescu

Prof. Dr.-Ing. Roman Dumitrescu is director at the Fraunhofer Institute for Mechatronics Design (IEM) in Paderborn and head of the Advanced Systems Engineering department at the Heinz Nixdorf Institute at the University of Paderborn. After his studies and doctorate, he took on various responsible tasks, including as managing director of the technology network Intelligent Technical Systems OstWestfalen-Lippe (it’s OWL) and as initiator of the tech startups Two Pillars GmbH and AI Marketplace GmbH. He is also involved in various advisory committees, including as a member of the German Academy of Engineering Sciences (acatech)

Fabian Hanke

Fabian Hanke, born in 1994, studied mechanical engineering (M.Sc.) at the University of Paderborn. Since February 2022, he has been a research associate in the Digital Engineering department at the Fraunhofer Institute for Mechatronics Design (IEM) in Paderborn.

Oliver von Heißen

Oliver von Heißen, born in 1994n studied computer science (M.Sc.) at the University of Paderborn. Since July 2022, he has been a research associate in the Digital Engineering department at the Fraunhofer Institute for Design Technology and Mechatronics (IEM) in Paderborn.

Lydia Kaiser

Prof. Dr.-Ing. Lydia Kaiser, born in 1982, is head of the Digital Engineering 4.0 department at the Technical University of Berlin and the Einstein Center Digital Future. She is also a member of the extended board of the Society for Systems Engineering. After studying physics and earning her doctorate in mechanical engineering at the University of Paderborn, she took on various management positions at Fraunhofer IEM in Paderborn.

Denis Tissen

Denis Tissen, born in 1994 studied mechanical engineering (M.Sc.) at the University of Paderborn. Since September 2020 he has been a research assistant in the Advanced Systems Engineering department at the Heinz Nixdorf Institute at the University of Paderborn.

Literature

1 Dumitrescu, R.; Albers, A.; Riedel, O. et al. (eds.): Engineering in Deutschland – Status quo in Wirtschaft und Wissenschaft: Ein Beitrag zum Advanced Systems Engineering. Paderborn, 2021Search in Google Scholar

2 Friedenthal, S.; Moore, A.; Steiner, R.: A Practical Guide to SysML – The Systems Modeling Language. Elsevier, 2015Search in Google Scholar

3 Chami, M.; Bruel, J.-M.: A Survey on MBSE Adoption Challenges. The Systems Engineering Conference of the Europe, Middle-East and Africa (EMEA) Sector of INCOSE (EMEASEC 2018), Berlin, Germany, 2018Search in Google Scholar

4 Kharatyan, A.; Humpert, L.; Anacker, H. et al.: Künstliche Intelligenz im Engineering. Zeitschrift für wirtschaftlichen Fabrikbetrieb ZWF, 117 (2022) 6, pp. 427–431 DOI:10.1515/zwf-2022-107410.1515/zwf-2022-1074Search in Google Scholar

5 Schräder, E.; Bernijazov, R.; Foullois, M. et al.: Examples of AI-based Assistance Systems in Context of Model-Based Systems Engineering: 2022 IEEE International Symposium on Systems Engineering (ISSE), 2022, pp. 1–8 DOI:10.1109/ISSE54508.2022.1000548710.1109/ISSE54508.2022.10005487Search in Google Scholar

6 Bretz, L.; Foullios, M.; Hillebrand, M.; Bernijazov, R.: Engineering Intelligence – KI-Kompetenz wird für Entwickler immer wichtiger. IT&Production (12/2018)Search in Google Scholar

7 Bernijazov, R.; Dicks, A.; Dumitrescu, R. et al.: A Meta-Review on Artificial Intelligence in Product Creation. Proceedings of the 30^th International Joint Conference on Artificial Intelligence (IJCAI-21), 2021Search in Google Scholar

8 Tissen, D.; Wiederkehr, I.; Bernijazov, R. et al.: Spearhead Data-Driven Model-Based Systems Engineering: Interview Study on Definition, Preconditions, Challenges, Potentials, and Use Cases: Proceedings of NordDesign, Reykjavik, Iceland, 12^th–14^th August 2024, pp. 539–548 DOI:10.35199/NORDDESIGN2024.5810.35199/NORDDESIGN2024.58Search in Google Scholar

9 Gartner Research: 7 Levels of Hybrid Human and AI Decision Making. 07/2021Search in Google Scholar

10 Korsten, G.; Aysolmaz, B.; Turetken, O. et al.: ADA-CMM: A Capability Maturity Model for Advanced Data Analytics. In: Bui, T. (eds.): Proceedings of the 55^th Hawaii International Conference on System Sciences, 2022 DOI:10.24251/HICSS.2022.03210.24251/HICSS.2022.032Search in Google Scholar

11 Kaiser, L.; Schräder, E.; Bernijazov, R. et al.: Ein Ansatz zur Strukturierung von KI-Assistenzen im Model-Based Systems Engineering: Tag des Systems Engineering, Tagungsband Paderborn, 16.–18. November 2022. Online available https://www.researchgate.net/publication/388884726_Ein_Ansatz_zur_Strukturierung_von_KI-Assistenzen_im_Model-Based_Systems_EngineeringSearch in Google Scholar

12 Peffers, K.; Tuunanen, T.; Rothenberger, M. A.; Chatterjee, S.: A Design Science Research Methodology for Information Systems Research. Journal of Management Information Systems, 24 (2007) 3, pp. 45–77 DOI:10.2753/MIS0742-122224030210.2753/MIS0742-1222240302Search in Google Scholar

13 Pohl, K. (ed.): Model-based Engineering of Embedded Eystems – The SPES 2020 Methodology. Springer, Berlin, 2012 DOI:10.1007/978-3-642-34614-910.1007/978-3-642-34614-9Search in Google Scholar

14 Kulkarni, P. J.; Tissen, D.; Bernijazov, R.; Dumitrescu, R.: Towards Automated Design: Automatically Generating Modeling Elements with Prompt Engineering and Generative Artificial Intelligence. Proceedings of Nord-Design, Reykjavik, Iceland, 12^th–14^th August 2024, pp. 617–625 DOI:10.35199/NORDDESIGN2024.6610.35199/NORDDESIGN2024.66Search in Google Scholar

15 Heißen, O. von; Mpidi Bita, I.; Hanke, F. et al.: Toward Intelligent Generation of System Architectures. Proceedings of NordDesign 2024, pp. 504–513 DOI:10.35199/NORDDESIGN2024.5410.35199/NORDDESIGN2024.54Search in Google Scholar

Published Online: 2025-03-27

Published in Print: 2025-03-20

Dieses Werk ist lizensiert unter einer Creative Commons Namensnennung 4.0 International Lizenz.

Articles in the same Issue

https://doi.org/10.1515/zwf-2024-0123

Keywords for this article

Model-Based Systems Engineering; Digital Engineering; Artificial Intelligence; Systems Engineering; Assistant Systems

Creative Commons

BY 4.0