Cost estimation of functional and physical changes made to complex systems

Abstract

Current cost estimation practices rely on statistically relating physical parameters of a system to historical cost data. Unfortunately, this method is unable to effectively communicate the increasing complexity of system design to cost data. Additionally, current cost estimation techniques have had a historical inability to produce credible and explainable results. It is often considered to be a "black art" with the recurring question: "Where did that number come from?" This thesis systematically links design and cost information together, and demonstrates the utility of that link by estimating the impact of functional and physical design changes on the life-cycle cost and determining key cost drivers. The ability to quickly estimate the cost impact of design changes is important for decision makers and serves as a medium of communication between customers and developers. Credible estimation is gained by intimately linking the axiomatic design framework to the already existing costing unit (or component) domain and providing design traceability. Development cost is predicted by determining the functional requirements (FRs) affected by a change in customer needs or constraints, then by determining the propagation of that change from FRs to design parameters (DPs) to costing units. The list of affected components and the magnitude of the impact on each component is found and then used to determine through a parallel iteration process model how much development labor will be necessary to implement those changes. The labor is directly related to development costs. A formal method to designing operations using axiomatic design is presented in this thesis. Operations exist due to the time-variant combinatorial complexity of FRs.

(cont.) Operations implement reinitialization procedures in order to maximize the probability of success of FRs. This provides the way that axiomatic design can derive operations and the related cost parameters. This information could then be plugged into the cost impact model of a design change to determine the list-of affected operations. A new method of estimating the change in cost parameters due to a design change will be the focus of future research. Two main forms of key cost drivers are identified: the most expensive FRs and design iteration. A method of mapping estimates from the costing unit domain to the FR-DP map is suggested in order to cost out FRs. Design iteration as a key cost driver can be seen from two points of view. Axiomatic design identifies small design ranges, coupling and imaginary complexity as contributors to cost. Design structure matrices identify the most iterative set of tasks in the development process and offer procedures to reduce or speed up the iteration.