Digital Intelligence & Forensics
Measurement & Analysis
Performance & Dependability
Process & Performance Improvement
Security & Survivability
In fields as diverse as currency-exchange-rate forecasting and protein-folding simulations to better understand Alzheimer’s and other diseases, computational engineering (CE) has demonstrated that high-performance computing can address difficult, large-scale, scientific and engineering problems—the kind the Department of Defense (DoD) faces in the acquisition of air vehicles, ships, and other major assets.
To realize the potential CE offers, the DoD knows it must overcome some challenges for using CE in acquisitions of software-reliant systems. In FY2009, the SEI began work that can lead to extending best software engineering practices to CE through a study and a workshop for the DoD-sponsored Computational Research and Engineering Acquisition Tools and Environments (CREATE) program.
CE often requires the development of sizeable amounts of complex software to accurately represent multi-dimensional mathematical models and physical phenomena. However, the scientists who are developing that software typically are not formally trained in software engineering. They develop application codes with an eye towards solving an immediate problem, but without a systematic approach to assuring critical quality attributes such as performance, portability, and robustness or to evolving codes over the 20- to 50-year life spans of major DoD assets.
“Computational engineering software development lacks investment in costeffective infrastructure, doesn’t use state-of-the-art development tools or best software engineering practices, and has not dealt with production quality issues,” says David Fisher of the SEI, who acted as the CREATE program’s chief engineer in FY2009.
In addition, CE scientists need to become software developers for DoD asset design engineers, rather than being solely user-developers, according to Lisa Brownsword of the SEI, who conducted the study with colleagues Jim Smith and Phil Boxer. “Although developing for users with a different background is the typical case for most professional software developers, it is a new paradigm for scientist-developers,” Brownsword says.
The use of CE in modeling climate change, processing seismic data for oil and gas discovery, simulating aero-acoustic effects in defense and aerospace work, and the like is somewhat remote for most people. But CE also predicts behavior for more common things, such as the tires on the family car or diapers on the baby riding in the back seat of that car. Tires are made of composite materials (rayon, steel, nylon) that react differently to temperature and load conditions among other factors. Predicting how a composite “system” of materials will behave requires complex modeling. And computational fluid dynamics can model how well a diaper’s absorbent material will wick moisture away.
In its study, the SEI looked into existing CE software development capabilities, current and potential uses of CE for engineering decisions across the acquisition life cycle of DoD assets, and the gaps between current capabilities and the future needs of DoD asset design engineers. In the workshop, the SEI gathered 18 experts to develop a strategy and define software engineering practices for CREATE products.
“Our work with CREATE provides a unique opportunity for the SEI to engage with a community that is moving from developing and using complicated software as tools to support its own science agenda to providing tools to design engineers trying to solve complex DoD asset design and analysis problems,” says Brownsword.