SUMMARY

In design and optimization of a complex system, there exist various methods for defining the relationship between the system as a whole, the subsystems and the individual components. Traditional methods provide requirements at the system level which lead to a set of design targets for each subsystem. Meeting these targets is sometimes a simple task or can be very difficult and expensive, but this is not captured in the design process and therefore unknown at the system level. This work shows that under uncertainty, the use of design targets will lead to design choices that do not meet the targets and do not optimize the system. We propose Value Driven Design as an alternative that provides more information to the subsystem designer and ensures that if each subsystem is optimized, so too will the system as a whole. A computational experiment is proposed as a means of evaluating Requirement Allocation and VDD. A common preliminary design is determined by optimizing the utility of the system, and then a Subsystem of Interest (SOI) is chosen as the focal point of detail design. The SOI is design using both proposed approaches and evaluated under uncertainty. The expected utility of the resulting design of each method is quantitatively compared to determine which method yields the design with the highest expected utility. The computational experiment is carried out on a Series Hybrid Consumer Vehicle as an example system. The vehicle system is modeled at varying levels of fidelity and varying levels of the hierarchical structure. The electric motor is chosen as the SOI and is modeled from the geometry of the motor using First Principles. The contributions of this work are in the field of complex system design which can be applied to many technical domains. The deficiencies of current methods as well as possible improvements to the systems engineering process are discussed and best practices for the two proposed approaches are provided.