Ranking Subsystem Targets According to Their Influence on System Performance Haitham Mahmoud Pierre T. Kabamba A. Galip Ulsoy College of Engineering, University of Michigan, Ann Arbor, MI 48109 Dr. Gerald A. Brusher Ford Motor Company, Dearborn, MI Abstract Large, complex systems are often decomposed into several subsystems that are designed in parallel. This process requires specifying design targets for the different subsystems such that meeting these targets yields an overall system with satisfactory performance. Methods for setting and balancing subsystem targets were described in previous publications [10], [11]. This paper addresses the question of how to rank, or set priorities on, different subsystem design targets according to their influence on overall system performance. The present work quantifies the influence of changes in the different subsystem design targets on overall system performance. Calculating the sensitivities of system- level performance with respect to subsystem targets rather than design variables, as traditionally done in the design literature, is a key component of the present study. The proposed procedure for ranking subsystem targets will enable product design teams to commit their engineering resources towards modification of only those subsystem targets that will bring about the greatest improvement in overall system performance. The present study deals specifically with Noise, Vibration and Harshness (NVH) design targets. 1 Introduction Product Development is the process of transforming a market opportunity into a final product [6]. The difficulties encountered by engineers during the design stage of the product development process motivate the present work. In particular, this work addresses the issue of identifying which subsystem design targets are most important in delivering overall system performance. This is especially critical when design teams are faced with hundreds of targets for a given subsystem [11]. Product Development starts with identifying customer, corporate and regulatory requirements. These requirements define the desired system-level specifications. Since the development of a new product usually involves the collaboration of several design teams, it is necessary to cascade the system-level requirements into performance targets for the different subsystems. Design teams can then be assigned the tasks of designing the subsystems to meet their respective targets. The process of cascading system-level requirements into design targets for the different subsystems should be performed in a ì consistentî and ì efficientî manner [4]. Consistent subsystem targets will ensure that, upon assembly, the different subsystems achieve the desired system-level performance. An efficient process is characterized by judicious decisions early in the Product Development process that avoid costly iterations late in the design cycle. Hence, a consistent and efficient target- cascading process ensures that subsystem targets correspond to the system-level performance specifications and that design decisions are made uniformly to meet the overall system requirements. A generic systems engineering approach to vehicle product development is shown in Figure 1. Design targets flow down, or cascade, from the vehicle level to the system, subsystem, and component levels. This cascade of targets is not unidirectional. Infeasible targets must be rebalanced at a higher level in the hierarchy. Thus, the cascading process is iterative. There is also a gradual freezing of design details and systematic verification against targets as the program proceeds through its milestones. The target cascading process often increases the number of subsystem targets as compared to the initial number of system-level targets. Though qualitatively understood, the relationships between subsystem targets are often not well-quantified. This hinders the ability of design engineers to ì balanceî targets among subsystems, i.e. to relax targets on some subsystems (that may be difficult to design or produce) by tightening the requirements on other subsystems [10]. Moreover, the lack of analytically rigorous relationships between the different system-level specifications and subsystem targets impedes the ability of design engineers to quantify the influence of changes in the subsystem targets on overall system performance. Consequently, the focus shifts toward relating system requirements to component design variables, effectively bypassing the intermediate subsystem-level targets. This encourages premature definition of design details, which in turn may lead to costly design changes late in the program. The aforementioned problems are common to several industries that use system decomposition and target cascading as part of their product development process. A new process that addresses these issues and improves the product development process will be of great value to these industries.