Investigating Emergent Design Failures Using a Knowledge-Action-Decision Framework

Abstract

Cost growth and schedule delays frequently impact the design of naval vessels as well as other physically-large and complex products. In vessel design, many instances of cost growth, performance loss, and delays occur suddenly and are unexpected, causing significant damage to an acquisition program. These outcomes are not caused by physical product failure. Instead, they are symptoms of design failures and are characterized by emergent rework, inability to integrate disparate information, and unexpected design difficulty. Design failures emerge from learning and decision-making that occurs during a design activity. In naval acquisitions, failures are frequently caused by the complexity of the naval design activity. Design approaches, processes, methods, and tools attempt to manage this complexity, but no methods exist to address it directly. To address the root-cause of design failures, new capabilities are needed to understand the knowledge-based relationships that drive their emergence. This thesis proposes a Knowledge-Action-Decision (K-A-D) Framework that enables a knowledge-centric perspective of design. The K-A-D Framework decomposes design into an endogenous knowledge structure that describes the changing relationships between the ideas, concepts, and evidence designers use to progress a design activity. Analysis of the K-A-D Framework through a network representation allows the importance of knowledge structure elements to be measured, the trajectory of a design activity to be identified, and opportunities to prevent design failures to be found. The K-A-D Framework, its network representation, and network analysis provide insight into the complex design dynamics that create unexpected design outcomes. First, results and methods from design science and the study of path dependence are synthesized to create the theoretical basis of the K-A-D Framework. The framework is constructed to capture the temporal relationships in knowledge structure development that can ingrain and lock-in design outcomes through path dependence. Second, the framework is applied to investigate the complex behaviors of an agent designing a solution to the Traveling Salesman Problem. Using network analysis of the K-A-D Framework, the agent's solution development and decision-making behaviors are identified and, in some cases, predicted. Third, the knowledge-centric perspective is applied to a naval distributed system design scenario by using the framework to define an approximate knowledge structure for distributed system integration. Finally, ensemble analysis is used to investigate the potential distributed system design knowledge structure characteristics under different early-stage design conditions. Examining the knowledge structure characteristics identifies design conditions that have elevated risk for late-stage design failures. Analysis of path dependence in the knowledge structure development identifies conditions where the risk of failure may be reduced by decisions made by the designer during the distributed system design activity.