Aero-Astro Magazine Highlight

Strategic engineering: designing systemns for an uncertain future

By Olivier de Weck

In engineering, customer requirements are traditionally gathered and frozen to help generate implementable designs. This is a proven approach for systems of low to moderate complexity. However, for large-scale engineering systems such as those critical to our aerospace, energy, and manufacturing industries, this can be a recipe for failure. Here is a striking example:

"Motorola unveils new concept for global personal communications: base is constellation of low-orbit cellular satellites." Motorola Press Release on the Iridium Satellite System, London, June 26, 1990,

and

“Iridium LLC seeks bankruptcy protection … Iridium has sapped more than $5 billion in resources from investors world-wide.” Wall Street Journal, New York, August 16, 1999

Olivier de Weck with Experimental Projects (16.622) course students Brittany Baker (left) and Noelle Steber in Aero-Astro’s Gelb Laboratory with their planetary rover
reconfigurable wheel testbed. The wheel can change its shape to optimize drawbar pull and power consumption for varying soil conditions.

These quotes illustrate the dilemma. Iridium pioneered mobile satellite communications in the 1990s. It achieved many technological breakthroughs. Unfortunately, its original market forecasts and chosen capacity were confounded by the unexpected success of competing terrestrial cellular systems. The system’s architecture was technically successful, but contributed to its commercial failure. In essence, the system became “locked” into a rigid configuration that had been chosen based on a “best guess” about an uncertain future.

I have found that the issue of having to design systems for uncertain future requirements is important and ubiquitous in many large-scale aerospace projects. This is why the focus of my research is the study of how complex systems and products evolve over time, and how they can be deliberately designed for changeability. Changeability is the degree to which a system can undergo modifications in its configuration without incurring large increases in its complexity and cost. Flexible systems are those that exhibit a high degree of changeability. Reconfigurable systems are the most flexible since they allow fast and reversible configuration changes.

My experience in working on the redesign of the Swiss F/A-18 aircraft (1991-96) taught me that systems inevitably change. While the upgrading of avionics and software was relatively easy in this particular aircraft, modifications to the airframe turned out to be difficult. To certify the aircraft for 5,000 flight hours and flight loads up to 9g, we made changes to the baseline configuration. Some of these changes were well-behaved. Others, such as the substitution of titanium for aluminum in the three carry-through bulkheads, rippled through the system in complex ways. Unanticipated change propagation to other structures, manufacturing processes, and flight-control software added significant costs to the program. This experience sparked my interest in systems engineering and inspired my current research.

Increasingly, complex engineering systems have to account for partially unknown future requirements rather than being designed to a single “optimal” point. How can we identify changes that are likely to propagate? Are there generalizable principles across multiple domains, or must each engineering project be treated as a unique undertaking? Therein lays the challenge of my scientific quest.

Framework for strategic engineering

Typically, the first step in designing a new system such as a passenger aircraft, satellite system, or launch vehicle is systems architecting. This involves understanding the underlying stakeholder structure and value flows, mission needs and possible range of requirements. During this phase various concepts are generated, evaluated and selected. The technical performance, lifecycle cost, value, and risk of a particular concept are usually only known approximately at this stage. The purpose of Integrated Modeling and Simulation is therefore to provide better estimates of system performance under both nominal (expected) and off-nominal conditions. Typically initial concepts are then further refined, and Multidisciplinary Design Optimization is used to fine-tune system configurations such that optimal designs, designated as
x*, can be found.

An overview of the “big picture,” which articulates Olivier de Weck’s view of his research areas in strategic engineering of systems and how they fit together. -enlarge-

Increasingly large aerospace systems are experiencing unexpected events and requirements shifts during their operations. Examples include the high price of aviation fuels, shifting customer needs from voice and video to data in satellite communications or an increased use of unmanned aerial vehicles for combat rather than pure surveillance missions. These uncertainties often require changes to the original system designs; that is, at some time t=tº+Δt the original configuration x* is no longer optimal. If a system was designed such that it is very difficulty to change, it essentially becomes locked in to its original configuration. Systems are often changed retrospectively through retrofits or block upgrades in a reactive manner. This can be expensive and slow due to extensive change propagation effects as I experienced them on the F/A-18 program. Design for Changeability is a proactive approach to designing systems such that they can be “easily” changed by embedding flexibility in the original designs. This may in turn affect the choice of architecture and system decomposition as some architectures tend to be inherently more changeable than others.

A second important observation is that many aerospace systems are no longer designed and produced individually, but as part of a larger ensemble such as a product family. Examples include the Boeing 787 family with variants such as the 787-8, 787-3 and 787-9, which differ primarily in terms of their range and passenger capacity. Satellite manufacturers are increasingly developing system variants from a common “bus” or platform. Design for Commonality attempts to maximize the overall value of systems and projects by identifying opportunities for commonality and partial standardization. This in turn feeds back on system architecture choices through platforms and the choice of system modularization.

Designing for an uncertain future

While strategic engineering is still an evolving area of research, the theoretical foundations and industrial applications have grown significantly in recent years. This is important when three conditions are present: long lifecycles, large irreversible investments, and significant requirements uncertainty. The core of my research is now focused on the development of generalizable methods that allow the incorporation of changeability and commonality considerations during systems architecting and design.

An early result was the development of staged deployment strategies for satellite constellations in response to the type of uncertainty that had been experienced by Iridium. Previous work had focused only on static constellation optimization for global Earth coverage with a minimum number of satellites. We adopted a very different approach from the traditional design method and found optimal evolution paths for satellite constellations based on calibrated technical-economic models. These results have been incorporated in courses such as 16.861 and ESD.36 and are assisting firms in the evaluation of the next generation of satellite constellations (e.g., Iridium Next).

A staged deployment strategy for constellations of communications satellites is an example of planning for an uncertain future. The figure shows the optimal evolution path of a satellite constellation across three stages in terms of its lifecycle cost versus duplex channel capacity. (Olivier de Weck image)

As part of a project supported by Raytheon Integrated Defense Systems, we conducted a detailed analysis of more than 41,000 change requests on an engineering project. To everyone’s surprise, we discovered a network of 2,600 changes that evolved over time during the eight-year development project of a complex radar system. This allowed us to isolate which subsystems were acting as multipliers or absorbers of change with the help of a new Change Propagation Index. Our time-expanded decision network methodology is perhaps one the most unifying contributions; it allows simulation and optimization of a system’s evolution over time, by representing both the exogenous uncertainties and the decisions to reconfigure as an acyclic time-expanded network.

Future vision

Over the coming decades, billions of dollars will be invested in infrastructure renewal of vital large-scale systems. Globalization, technological innovation as well as regulatory, environmental and demographic changes drive large uncertainties, which must be reflected by the choice of appropriate system architectures and design solutions. A recent National Science Foundation workshop on complex systems acknowledged the changing and uncertain exogenous environment as an important phenomenon. The following research question was identified as one of the most important: “How can architectures enable resilient, adaptive, agile, evolvable systems?”

Together with my students and colleagues, I will continue to address the question of evolvability of systems and products over time. In addition to developing methods for designing new systems for changeability I want to study in more detail the past evolution of existing complex technical systems to understand which ones tended to be stable and which ones underwent significant changes, and why. Thinking deeply about changeability will help us be more proactive about the design of future systems.

“The engineer of 2020 will be faced with myriad challenges, creating offensive and defensive solutions at the macro- and microscales in preparation for possible dramatic changes in the world.” The Engineer of 2020: Visions of Engineering in the New Century, National Academy of Engineering, (2004), p.24

Olivier L. de Weck is an Associate Professor in the MIT Aero-Astro Department and Associate Director of the MIT Engineering Systems Division. His research interests are in Systems Engineering and Space Systems Design and Logistics. He has a Diplom Ingenieur degree in industrial engineering from the Swiss Federal Institute of Technology (1993) and a Ph.D. in Aerospace Systems from MIT (2001). From 1993 to 1997 he served as liaison engineer and, later, as engineering program manager for the Swiss F/A-18 program at McDonnell Douglas (now Boeing) in St. Louis. More information on his work may be found at http://strategic.mit.edu. He may be contacted at deweck@mit.edu