Model based systems engineering to support failure mode avoidance for driver-assistance systems

The design and production of modern automotive systems requires addressing concerns of multiple stakeholders, ranging from regulatory to safety, supply chain and manufacturing. These concerns can sometimes make it challenging to maintain a consistent set of product specifications throughout the lifecycle. These challenges present opportunities for applying model-based methods, not only to formally capture the specifications, but also to help in minimizing inconsistencies and generating different work products that are essential to the completion of an automotive program. We present a case study where OMG SysML™ was used to support the development of a Driver-Assist Technology within Ford Motor Company. We followed the proposals of multi-abstraction and multi-view modeling to help make the process of producing the work products more effective and efficient. First, we aligned the modeling with the organizational structure by introducing multiple abstraction levels. This enabled a design team within the organization to infer what their specific constraints are, and how these constraints cascade to other levels/teams. This also enabled us to generate work products catered to each abstraction level. Second, we used multi-formalism modeling to formally capture both the structural and behavioral constraints, along with addressing concerns around failure mode avoidance and customer experience. Third, we performed meta-modeling to help incorporate the terminology and conventions from the automotive domain into our models. Our UML profile for failure model avoidance (FMA) also helps in generating the work products automatically from the model, such as FMEA P-diagrams, and interface specifications, which is very useful when working with suppliers. Overall, our approach improved traceability between work flows, and enhanced the effectiveness of design review discussions. On the other hand, the diagrams prevalent in system engineering are sometimes prone to misinterpretation and often to cluttering, which needs to be addressed to better the effectiveness of the approach.