From Document-Based to Model-Based System and Software Engineering: Experience Report of a Selective Catalytic Reduction System Development

Morayo Adedjouma, Thibaud Thomas, Chokri Mraidha, Sebastien Gerard, Guillaume ZellerEduSymp/OSS4MDE@MoDELS 2016

TL;DR

Plastic Omnium (PO) transitioned from a document-heavy system development process to a model-based approach using SysML and AUTOSAR. This shift significantly improved traceability, consistency, and efficiency in developing complex automotive systems like Selective Catalytic Reduction (SCR) units, addressing challenges posed by traditional document-centric methods.

What Problem Does This Paper Address?

Automotive manufacturers like Plastic Omnium (PO) face increasing pressure to develop complex, customizable, and safety-compliant systems (e.g., emission reduction systems) while working with multiple suppliers. Traditionally, PO relied on a Document-Based System Engineering (DBSE) approach, where system specifications and designs were managed through numerous documents. This method led to significant challenges:

  • Manual Translation & Inconsistency: Translating textual specifications into software models (like AUTOSAR) was time-consuming, error-prone, and created inconsistencies.
  • Traceability Issues: Maintaining clear links between requirements, design elements, and final code was difficult, hindering compliance with standards like ISO 26262 and A-SPICE.
  • Workflow Inefficiencies: The sequential nature of document creation and updates led to delays and difficulties in managing parallel development activities.
  • Communication Barriers: Information spread across multiple documents made it hard for stakeholders to gain a unified understanding of the system.
  • Tooling Limitations: Integrating various commercial tools for requirements management, design, and simulation proved cumbersome, requiring manual data exchange.

The paper addresses the need for a more integrated, consistent, and traceable development process to overcome these limitations.

Methods

The paper details PO's adoption of a Model-Based System Engineering (MBSE) approach. Key components of this new approach include:

  • Modeling Languages:
    • SysML (Systems Modeling Language): Chosen for system-level activities, providing a standardized way to capture requirements, structural and behavioral design, and implementation.
  • AUTOSAR (Automotive Open System Architecture): Used for software architecture design, a de facto standard in the automotive industry.
  • Modeling Methodology: A five-abstraction-level methodology was defined, spanning from system requirements (SysTS) to software component specifications (SwCTS), with clear links between each level.
  • Modeling Tool:
    • Papyrus: An open-source tool that supports both SysML and AUTOSAR, offering customization capabilities, integration features, and the ability to generate documentation.
  • Traceability: The MBSE approach inherently supports automated traceability between model elements across different abstraction levels (e.g., from system functions to AUTOSAR runnables), a significant improvement over DBSE.
  • Integration: Efforts were made to integrate Papyrus with existing tools like IBM DOORS (for requirements) and MATLAB (for simulation), facilitating a smoother transition.

Datasets

Not discussed in the paper. The paper focuses on an experience report of applying MBSE to an existing product (SCR system) rather than introducing new datasets or benchmarks.

Key Insights and Ideas

  • Traceability as a Core Benefit: The most significant advantage of MBSE is the automated and integrated traceability it provides. This directly addresses the critical need for compliance with automotive standards like ISO 26262 and A-SPICE.
  • Unified Representation: Moving from disparate documents to a single, interconnected model reduces ambiguity and improves consistency in design and requirements.
  • Early Design Validation: MBSE allows for earlier and more iterative design and analysis, as models can be checked and refined throughout the development lifecycle, not just at document release points.
  • Tooling Synergy: The use of an open-source, customizable tool like Papyrus was crucial for adapting the MBSE approach to PO's specific needs and integrating with existing workflows.
  • Publication on Demand: The ability to generate documentation from models "on demand" rather than relying on fixed publication cycles offers greater flexibility.
  • Skill Shift Required: The transition necessitates a significant shift in engineer skills, requiring training and potentially impacting hiring practices.

Key Results

The paper reports on the successful re-engineering of a Selective Catalytic Reduction (SCR) system using the MBSE approach. While specific quantitative metrics (e.g., percentage reduction in errors, time savings) are not detailed, the qualitative findings highlight:

  • Improved Workflow and Planning: Project planning became less constrained, allowing engineers to work more concurrently across different levels.
  • Enhanced Design Semantics: SysML models replaced ambiguous natural language descriptions, leading to more formal, consistent, and traceable design elements.
  • Automated Traceability: The MBSE approach significantly improved traceability, facilitating justification and global consistency for certification.
  • Facilitated Communication: A unified model and dedicated viewpoints improved communication among stakeholders.
  • Smoother Standards Compliance: The MBSE approach enabled better integration between specification, design, and implementation, aligning with A-SPICE and ISO 26262 expectations.

Limitations and Future Work

Limitations:

  • Publication Effort: Converting models into documents for quality processes still requires effort.
  • Communication Challenges: Bridging the gap between engineers using technical SysML and document reviewers/approvers remains a challenge.
  • Skill Gap: A significant effort is needed to train existing staff and ensure new hires possess the necessary SysML skills.
  • Tool Chain Management: Establishing a dedicated tool chain management role and processes is necessary for large-scale, international deployment.

Future Work:

  • Integration of Further Concerns: Leveraging Papyrus to integrate additional aspects like dysfunctional analysis and design simulation.
  • Process and Product Certification: Exploring how MBSE can further facilitate process and product certification.
  • Domain Specific Languages (DSLs): Developing DSLs to simplify SysML usage and reduce the skill barrier.
  • Automated Checks: Exploiting traceability for automated checks like validation rules and change impact analysis.

Summary updated

BibTeX
@inproceedings{adedjouma2016from,
  author = {Morayo Adedjouma and Thibaud Thomas and Chokri Mraidha and Sebastien Gerard and Guillaume Zeller},
  title = {{From Document-Based to Model-Based System and Software Engineering: Experience Report of a Selective Catalytic Reduction System Development}},
  booktitle = {EduSymp/OSS4MDE@MoDELS},
  year = {2016},
}
Figure