Blockmodeling and the Estimation of Evolutionary Architectural Growth in Major Defense Acquisition Programs

Abstract

It is well-known that cost overruns in Major Defense Acquisition Programs (MDAPs) are endemic, and requirements volatility is at least partially to blame. In particular, when the desired capabilities of a system change during its life cycle, substantial reengineering can result, especially when a new subsystem must be incorporated into an existing architecture. Of course, the likelihood and specifics of such additions are rarely known ahead of time, and predicting integration costs is challenging. In this paper, we present a novel algorithm to address this issue. In particular, leveraging an integer programming implementation of the social network analysis technique blockmodeling, we optimally partition the subsystems represented in Department of Defense Architecture Framework (DoDAF) models into architectural positions. Using this abstracted structure, we subsequently grow the architecture according to its statistical properties, and we estimate this unforeseen cost of evolutionary architectural growth via the Constructive Systems Engineering Cost Model (COSYSMO). We illustrate this process with a real-world example, discuss limitations, and highlight areas for future research.