CIFRE - Agile Model-Based System Engineering for multi-disciplinary optimization in future vehicle development

1. Title:

Agile Model-Based System Engineering for multi-disciplinary optimization in future vehicle development

2. Key words:

Design Space Exploration, Surrogate Modelling, Agile Model-Based System Engineering (MBSE)

3. Context of the research project

Bus, Trucks and Light Vehicle markets have high-complexity product portfolios. In this sector, we are often challenged to explore possible design options for our new vehicles’ development. Especially with the increasing need for improved sustainability, we are investing more in our future products.

3.1. Company presentation

Iveco Group, a company committed to transforming their business and leading the change to a more sustainable future. Iveco Group is the house of eight unique, yet unified Brands. Our product range includes Light, Medium and Heavy Commercial Vehicles (IVECO), Powertrain (FPT Industrial), Buses (HEULIEZ, IVECO BUS), Financial Services (IVECO CAPITAL), Specialty Vehicles (IDV, ASTRA and MAGIRUS).

3.2. Research laboratory presentation

The DISP Lab (Decision & Information Systems for Production systems, UR4570), gathers researchers from the “Université de Lyon” around a double expertise in Industrial Engineering and Enterprise Information Systems. The DISP lab brings to this project specific competences in models engineering for products at the design phase. :

• Simulation and federal simulation modelling
• System modelling,
• System architectures
• Model-Based System Engineering
• Model exploration
• Model mapping

The DISP lab will provide the scientific, technical, and material supports (dedicated IT working-space) to achieve the research project tasks and obtain the targeted results.

4. The research approach.

The targeted research aims at proposing a model-based approach to optimize vehicles engineering models and accelerate the release of future vehicles designs.

4.1. Research interest

When we want to design new products, we need to integrate multiple simulation models (e.g., Finite Elements Model, Cost, Control, etc.) developed with different simulation tools. However, these models are heavy. They include enormous quantities of data for different scenarios. To overcome the crucial computational burden brought by these models, we need make them lighter. To do so, we generate representative data from these models (i.e., Surrogate Modelling) that enable us to have light, quick, inexpensive; but low-fidelity models. We then link our product requirements to our models to ensure that the simulations satisfy our requirements. Running all these models at the same time for a defined objective and optimizing them to get the best scenario for our future vehicle design, nevertheless, is a challenge in automotive industry.

4.2. Research objectives and related issues

The proposed research project is organised around 3 complementary research objectives:

• Obj1: propose a model-based system engineering architecture to drive design process of new products. The related research issues are:
  • Definitin and validation of high-quality requirements models
  • The cnnexion between several modelling perspectives
  • The definitin of agile reference architecture for the generated models
• Obj2: propose a surrogate modelling approach to lightweight complex engineering models and facilitate their exploitation. Some related research issues can be listed below:
  • The apprximation method for surrogation
  • The trade-ff between objectives in surrogate modelling
  • The management f noise in data.
• Obj3: propose design space exploration (DSE) methodologies to achieve a near-optimal system. The related research issues are:
  • The definitin of the methodology steps
  • The alignment between design iteratin, design parameters, design problems and near optimality.

The proposed research objectives will be developed in the next section. The achievement of the expected results will be driven by the respective methodologies.

4.3. Proposed methodology and expected results.

To support the development of new products, we consider innovative Model Based System Engineering methods, Design Space Exploration, and Multi-Disciplinary Optimization to approach the problem of “best” design scenario identification for vehicles.

4.3.1. Axis 1: Model-Based System Engineering to drive product design process.

Systems engineering is an independent engineering discipline that focuses on system properties [1]– including functionality, structure, performance, safety, reliability, and security. Model-based System Engineering (MBSE) is a model-centric approach to performing systems engineering. Systems engineering is largely independent of the engineering disciplines used to implement these properties. Systems engineering is an interdisciplinary activity that focuses more on this integrated set of system properties than on the contributions of the individual engineering disciplines [2, 3]. It is an approach to developing complex and technologically diverse systems.

The emergence of Model-based System Engineering (MBSE) and supporting system-modelling languages (i.e. SysML [4]) aims at providing a product representation structure, through a unique and timeless model, which potentially drives the whole product lifecycle, as the single and ubiquitous information source to stakeholders [5, 6]. In such approach, several building blocks need to be defined [7]:

• The system specification allows the capturing and the analysis of requirements. Functional analysis generates “high-quality requirements” (e.g., complete, accurate, correct, consistent, and verifiable), use cases, and user stories – all means to understand what the system must be.
• The modelling artefacts and related processes covering behaviour modelling, requirements modelling, structure modelling, parametric modelling, Requirements traceability and changes control, etc.
• The development of system architectures covering the development of pattern-driven architecture, subsystem and component architectures, inputs and preconditions, outputs and postconditions, subsystem logical and physical interfaces, and an agile reference architecture (for effectiveness, interoperability, and regulatory compliance) as the set of strategic design optimization decisions for the system.
• The verification and validation of model compositions using model-based testing, traceability modelling, model versioning, etc.

Several proprietary and open-source modelling tools will be used to support the development of the MBSE approach.

4.3.2 Axis 2: Surrogate Modelling

The basic idea in the ‘surrogate model’ approach is to avoid the temptation to invest one’s computational budget in answering the question at hand and, invest in developing fast mathematical approximations to the long running computer codes [8]. Given these approximations, many questions can be posed and answered, many graphs can be made, many trade-offs explored, and many insights gained [9]. One can then return to the long running computer code to evaluate the ideas so generated and, if necessary, update the approximations and iterate.

While the basic idea of the surrogate model approach sounds simple, the devil is in the details. What points do you sample to use in building the approximation? What approximation method do you employ? How do you use the approximation to suggest new, improved designs? How do you use the approximations to explore trade-offs between objectives? What do you do if your simulation has numerical noise in it? And, equally important: Where do I get the computer code to do all these things? [10-13]

The surrogate modelling process may cover the following steps [14]:

• Define the sampling plan with experiments.
• Quantitative evaluation of designs (computer simulations of physical experiments)
• Construct surrogates, design sensitivities and identify multi-fidelity data.
• Visualize to help comprehend design landscapes.
• Optimize using the surrogates to search infill criteria, verify present constraints, noise in data, integrate multi-design objectives by adding new designs.

As supporting tool, MATLAB may be used to support surrogate modelling.

4.3.3.Axis 3: Design Space Exploration

The Design Space Exploration (DSE) methodologies search the options of the architecture, the components, the interfaces and the data mapping to achieve a near-optimal system, which satisfies the constraints and minimizes the complete trade-off of the multidimensional space [15]. During the DSE, the input and output requirements, the storage requirements, the processing requirements and the system control are explored [16]. The academia research focuses more on DSE methodologies that provide near-optimal designs. The ideal DSE methodology is to address all mapping problems simultaneously in a single phase. However, the ideal DSE methodology cannot be achieved. The design of automotive systems is an overly complex process, which consists of several mapping phases, thus no good way exists to optimally formulate the single-phase solution. Hence, the DSE methodology must be divided into several sub-tasks to be manageable [17].

The division of the DSE methodology into steps is usually performed in an ad hoc way and thus the steps usually have constraints that bidirectionally affect each other [18]. In the case of bidirectionally connected steps, iterations between the design steps are required to search for near-optimal solutions. Due to the bidirectional correlation of the steps, no guarantee exists that the DSE will be finalized in a reasonable execution time with a near-optimal design. When the number of software and hardware parameters of the design is increased, the costly design iterations of the bi-directionally correlated DSE steps lead to not scalable approaches [19]. For instance, an iterative DSE methodology starts from the designer’s base configuration, changes the value of one parameter each time and uses the results to predict the optimal design. This DSE methodology may lead to less efficient designs within the available exploration time when a high number of parameters and interdependencies exist [20]. Each DSE step consists of tasks which are solving NP-hard problems . The conventional techniques applied in each step can achieve near-optimality only for small design problems in the available exploration time. When the complexity of the design problem is increased, which is usually the case in industrial contexts, the conventional techniques are incapable of identifying the near-optimal solution within reasonable search time.

5. Research project impacts

The proposed research project aims at supporting IVECO engineering teams in the development of future vehicles starting from existing, improved, adapted, or new models, new requirements, new constraints, etc. Following the proposed methodology, we may measure different levels of impacts at the three proposed research axis.

5.1. Scientific impact

From the implementation of the MBSE approach, the functional analysis effort will help to release new high-quality requirement models, as well as other modelling efforts connected in a reference architecture and then assessed and validated.

The surrogate modelling will help to faster the exploitation of the design models. The proposed architecture will help to integrate multi-design objectives by adding gradually new designs. The optimization capabilities will be valorised through the rapid ingest of lightweight models.

The Design Space Exploration methodology will help to define the adequate level of sub-tasks and steps to achieve a near-optimal system.

The proposed and validated artefacts as well as their consumption workflows (to be connected with IVECO systems) will be valorised in scientific publications to be submitted in international journals (JIMS, EIS, etc.)

5.2. Socio-economic impact

The IVECO engineering team will take benefit from this project results at several levels:

• Get support in the validation of the completeness of their constraints and requirements related to different modelling viewpoints (through the MBSE approach).
• Get support in mapping the different models’ viewpoints (through the surrogate modelling approach)
• Get support in the identification of the near-optimal system (through the DSE approach).

The IVECO vehicles will gain in sustainability by design through the earlier ingest of diverse sustainability constraints at the MBSE approach.

6. Application procedure

For application in this PhD position, applicants are invited to communicate:

• An updated CV
• A motivation letter with explicit interest in this research project
• The last academic transcripts
• The last produced report
• At least two recommendation letters

For applications and further request of information, you can contact: Nejib.Moalla(at)univ-lyon2.fr, and hazal.aktekin(at)ivecogroup.com

7. References

1.            Douglass, B.P., Agile Model-Based Systems Engineering Cookbook. Second edition ed. 2022: Packt Publishing Ltd. 601.

2.            Long, D.I. and S. Ferguson, Assessing Lifecycle Value Using Object-Based Modeling by Incorporating Excess and Changeability. Journal of Mechanical Design, 2021. 143(5).

3.            Gebreegziabher, T.G., et al., A Model-based Method for Assisting Decision Making Process in Product Development. 2017 5th International Conference on Enterprise Systems (Es), 2017: p. 93-98.

4.            Hinckel, E., et al., Driving Product Design and Requirements Management with SysML. Transdisciplinary Engineering: Crossing Boundaries, 2016. 4: p. 1071-1080.

5.            Haughey, B., The Impact of Model-Based Systems Engineering on Reliability Growth. 2020 Annual Reliability and Maintainability Symposium (Rams 2020), 2020.

6.            Li, S.N., et al., Implementation of Systems Engineering Model into Product Lifecycle Management Platform. Product Lifecycle Management in the Era of Internet of Things, Plm 2015, 2016. 467: p. 601-608.

7.            WEILKIENS, T., et al., MODEL-BASED SYSTEM ARCHITECTURE. 2016: JohnWiley & Sons. 400.

8.            Alizadeh, R., J.K. Allen, and F. Mistree, Managing computational complexity using surrogate models: a critical review. Research in Engineering Design, 2020. 31: p. 275-298.

9.            Jiang, P., Q. Zhou, and X. Shao, Surrogate model-based engineering design and optimization. 2020: Springer.

10.         Zhang, X., et al., Multi-fidelity deep neural network surrogate model for aerodynamic shape optimization. Computer Methods in Applied Mechanics and Engineering, 2021. 373: p. 113485.

11.         Hwang, J.T. and J.R. Martins, A fast-prediction surrogate model for large datasets. Aerospace Science and Technology, 2018. 75: p. 74-87.

12.         Zhang, J., S. Chowdhury, and A. Messac, An adaptive hybrid surrogate model. Structural and Multidisciplinary Optimization, 2012. 46: p. 223-238.

13.         Audet, C., et al. A surrogate-model-based method for constrained optimization. in 8th symposium on multidisciplinary analysis and optimization. 2000.

14.         Forrester, A.I.J., A. Sóbester, and A.J. Keane, Engineering Design via Surrogate Modelling A Practical Guide. 2008: www.wiley.com. 238.

15.         Xie, Y., et al., Security/Timing-Aware Design Space Exploration of CAN FD for Automotive Cyber-Physical Systems. Ieee Transactions on Industrial Informatics, 2019. 15(2): p. 1094-1104.

16.         Graf, S., et al., Design Space Exploration for Automotive E/E Architecture Component Platforms. 2014 17th Euromicro Conference on Digital System Design (Dsd), 2014: p. 651-654.

17.         Lukasiewycz, M., et al., Efficient symbolic multi-objective design space exploration. 2008 Asia and South Pacific Design Automation Conference, Vols 1 and 2, 2008: p. 661-666.

18.         Rathfux, T., et al., Efficiently Finding Optimal Solutions to Easy Problems in Design Space Exploration: A* Tie-breaking. Icsoft: Proceedings of the 14th International Conference on Software Technologies, 2019: p. 595-604.

19.         Rathfux, T., et al., An Experimental Evaluation of Design Space Exploration of Hardware/Software Interfaces. Proceedings of the 14th International Conference on Evaluation of Novel Approaches to Software Engineering (Enase), 2019: p. 289-296.

20.         Kritikakou, A., F. Catthoor, and C. Goutis, Scalable and Near-Optimal Design Space Exploration for Embedded Systems. 2014: © Springer International Publishing Switzerland. 287.

Funding category: Cifre

PHD title: Informatique
PHD Country: France