Erich Devendorf

The Impact of Process Architecture on Equilibrium Stability in Distributed Design

In distributed design processes, individual design subsystems have local control over design variables and seek to satisfy their own individual objectives, which may also be influenced by some system level objectives. The resulting network of coupled subsystems will either converge to a stable equilibrium or diverge in an unstable manner. In this paper, we study the dependence of system stability on the solution process architecture. The solution process architecture describes how the design subsystems are ordered and can be either sequential, parallel, or a hybrid that incorporates both parallel and sequential elements. In this paper, we demonstrate that the stability of a distributed design system does indeed depend on the solution process architecture chosen, and we create a general process architecture model based on linear systems theory. The model allows the stability of equilibrium solutions to be analyzed for distributed design systems by converting any process architecture into an equivalent parallel representation. Moreover, we show that this approach can accurately predict when the equilibrium is unstable and the system divergent when previous models suggest that the system is convergent. [DOI: 10.1115/1.4004463]

Using Network Theory to Model Distributed Design Systems

In this paper a network based model for distributed design is presented. In distributed design systems, the transient response is a related to the connections between designers. There are two types of designer connections. The first are intrinsic connections which are established when a single design subsystem controls a design variable that is present in one or more other subsystem’s objective function. The second are structural connections which depend on the intrinsic connections and the solution process architecture. Previous research in distributed design has limited its examination to intrinsic connections. Through the application of social network theory, a new model for distributed design is proposed. A case study is presented that demonstrates the construction of this social network model. The model is then analyzed using network theory concepts like ‘distance,’ ‘bridging,’ and ‘degree centrality’ to identify preferable solution process architectures.

EXAMINING INTERACTIONS BETWEEN SOLUTION ARCHITECTURE AND DESIGNER MISTAKES

When designing complex systems, it is often the case that a design process is subjected to a variety of unexpected inputs, interruptions, and changes. These disturbances can create unintended consequences including changes to the design process architecture, the planned design responsibilities, or the design objectives and requirements. In this paper a specific type of design disturbance, mistakes, is investigated. The impact of mistakes on the convergence time of a distributed multi-subsystem optimization problem is studied for several solution process architectures. A five subsystem case study is used to help understand the ability of certain architectures to handle the impact of the mistakes. These observations have led to the hypothesis that selecting distributed design architectures that minimize the number of iterations to propagate mistakes can significantly reduce their impact. It is also observed that design architectures that converge quickly tend to have these same error damping properties. Considering these observations when selecting distributed design architectures can passively reduce the impact of mistakes.

PLANNING ON MISTAKES: AN APPROACH TO INCORPORATE ERROR CHECKING INTO THE DESIGN PROCESS

Mistakes in the design process have been recognized as a major source of product quality loss. There are several methods currently used to identify and quantify these mistakes. However, these methods typically do not provide a useful context within which to quantitatively incorporate mistakes into the design process in a beneficial way. This paper presents an approach to determine when it is appropriate to perform error checking to eliminate a potential mistake. The proposed approach is intended to be used when time is a limited design resource and design goals are technically attainable. It is proposed that the cost of a mistake can be quantified as the amount of time a mistake adds or subtracts from the overall time required to achieve the design’s objectives. To determine this, an optimization problem is formulated which minimizes time spent in the design process. In this optimization problem the design variables are the binary choice whether or not to perform an error check. The approach is demonstrated in two case studies, one a simple theoretical design problem and the other the design of an I-beam. The results of these case studies demonstrate the approach’s effectiveness, and present several avenues for future work.Copyright

USING PRODUCT ARCHAEOLOGY TO INTEGRATE GLOBAL, ECONOMIC, ENVIRONMENTAL, AND SOCIETAL FACTORS IN INTRODUCTORY DESIGN EDUCATION

Design education has traditionally been incorporated into the engineering curriculum in the junior or senior year through upper level mechanical design courses and capstone design projects. However, there is a general trend in engineering education to incorporate design activities at the freshman and sophomore level. The design aspects of these courses provide a unique opportunity to integrate global, economic, environmental, and societal factors with traditional design considerations. Incorporating these early in an engineering curriculum supports a broad engineering education in accordance with ABET required Outcome h. In this paper we introduce global, economic, environmental, and societal factors into a sophomore level engineering design course using strategies adapted from a Product Archaeology paradigm. Specifically, functional modeling is synthesized with a product dissection platform to create a foundation to demonstrate the broader impacts of engineering design decisions. The effectiveness of using Product Archaeology-based educational strategies to facilitate the learning objectives of Outcome h is evaluated using student surveys taken over a two year period.Copyright

Incorporating Process Architecture in the Evaluation of Stability in Distributed Design

In distributed design processes, individual design subsystems have local control over design variables and seek to satisfy their own individual objectives, which may also be influenced by some system level objectives. The resulting network of coupled subsystems will either converge to a stable equilibrium, or diverge in an unstable manner. In this paper, we study the dependence of system stability on the solution process architecture. The solution process architecture describes how the design subsystems are ordered and can be either sequential, parallel, or a hybrid that incorporates both parallel and sequential elements. In this paper we demonstrate that the stability of a distributed design system does indeed depend on the solution process architecture chosen and we create a general process architecture model based on linear systems theory. The model allows the stability of equilibrium solutions to be analyzed for distributed design systems by converting any process architecture into an equivalent parallel representation. Moreover, we show that this approach can accurately predict when the equilibrium is unstable and the system divergent when previous models suggest the system is convergent.Copyright

Development of a Distributed Design Toolkit for Analyzing Process Architectures

The distribution of a complex system design problem into smaller, coupled subsystem problems creates a number of research and implementation challenges, including determining the most appropriate process architecture. Here the term process architecture refers to the order in which design subsystems solve their individual optimization problem. While methods have been proposed to model specific distributed process architectures, analyzing the impact of a wide range of process architectures on solution quality, convergence, and stability still remains a difficult task. To address this, we introduce a toolkit that enables the modeling and analysis of distributed design process architectures using basic sequential, parallel, and hybrid elements. Features of the toolkit are described and its use in studying solution characteristics including quality, convergence, and stability is discussed. The toolkit allows for users to create and analyze possible processes architectures using graphical interfaces.

Optimal Process Architectures for Distributed Design Using a Social Network Model

Distributed design systems fundamentally preserve individual design subsystem secrecy by limiting communication across subsystems. The natural secrecy of distributed design makes it difficult for design process managers to determine the appropriate order of subsystems in the design process. In this paper, we discuss a social network theory based heuristic to prescribe the optimal order of design subsystems. We call the order of the design subsystems process architecture and we leverage concepts like ‘distance,’ ‘bridging,’ and ‘degree centrality’ to analyze the aggregate design system and identify preferable solution process architectures. Our network theory approach only requires a manager to know which subsystems share design information. We distinguish this research from previous work by empirically validating the heuristic against a genetic algorithm for 80 randomly generated distributed design systems. The heuristic performs well against the genetic algorithm and beats it in the majority of cases. Moreover, it does so without requiring any function evaluations.

Exploring the effectiveness of parallel systems in distributed design processes subjected to stochastic disruptions

Abstract During the design of complex systems, a design process may be subjected to stochastic disruptions, interruptions, and changes, which can be described broadly as “design impulses.” These impulses can have a significant impact on the transient response and converged equilibrium for the design system. We distinguish this research by focusing on the interactions between local and architectural impulses in the form of designer mistakes and dissolution, division, and combination impulses, respectively, for a distributed design case study. We provide statistical support for the “parallel character hypothesis,” which asserts that parallel arrangements generally best mitigate dissolution and division impulses. We find that local impulses tend to slow convergence, but systems also subjected to dissolution or division impulses still favor parallel arrangements. We statistically uphold the conclusion that the strategy to mitigate combination impulses is unaffected by the presence of local impulses.

Examining Interactions Between Process Architecture and Architectural Impulses

When designing complex systems, a design process may be subjected to a variety of stochastic inputs. These stochastic inputs can alter the systems converged equilibrium, stability, transient response or convergence path. In this paper, we broadly characterize stochastic inputs into a design process as design impulses. We categorize them into three classes based on how they interact with the design system: local impulses, external impulses and architectural impulses. We explicitly define design impulses and each of these impulse types. We then empirically investigate architectural impulses using simulation. Based on these simulations, we suggest and examine strategies to mitigate the impact of architectural impulses on a design process.