Researchers at the Aerospace Vehicle Systems Institute (AVSI) foresaw a problem with building the next generation of complex, software-intensive, safety-critical aircraft systems; as the complexity of the avionics systems continues to increase, they have identified a need for a fundamental change in developing the software and systems for the next generation system aircraft. Through Georgia Tech, AVSI conducted a pre-study of existing technologies that could help with software-intensive systems construction, and the Georgia Tech study recommended adoption of the Architecture Analysis and Design Language (AADL), which was developed at the SEI as a means to conduct model-based development.

AVSI: The Aerospace Vehicle Systems Institute (AVSI) is a consortium comprising aerospace companies—including Boeing, Lockheed Martin, Rockwell Collins, and others—the Department of Defense, and the Federal Aviation Administration. AVSI works to improve the integration of complex subsystems in aircraft.

“The AVSI project Systems Architecture Virtual Integration (SAVI) focuses on establishing a new way of specifying and integrating increasingly complex aerospace systems. This would reduce the cost and schedule of new airplane development while improving quality, safety, and performance,” says Jörgen Hansson of the SEI. Traditionally, subcontractors responsible for a part of the system would independently develop code or pieces of the system. When the pieces are brought together, the system has already gone far into development, but when you try to integrate all the pieces from the different subcontractors, the integration problems appear.

“So the question they are asking,” says Hansson, “is whether there is a way to conduct integration earlier using a model-based approach before the system is being built.” This is where AADL comes in. Using AADL, individual subcontractors can model their pieces of the system with large amounts of implementation detail. “Now I can take that model together with everyone else’s models and integrate them and make sure I get the system behavior I want for areas I determine to be critical,” says Hansson.

This process will allow AVSI to capture many integration faults as early in the development process as possible. The cost of fixing a fault escalates dramatically the later it is uncovered in the development process.

Studies have shown that 60 percent to 75 percent of all system defects are introduced in the system-lifecycle development phases preceding the code development— requirements engineering, system architecture design, and component designs. Yet only a small fraction of these defects, about 3 percent to 8 percent, are detected before code development and system realization; the majority of defects are detected at the time of system integration or later phases.

Costs of correcting defects in the system-integration phase or after the system has been deployed into operation are 15 to 30 times, and 30 to 110 times higher, respectively, compared to the cost of removing the defects early.

Correcting late-detected defects incurs significant costs. For example, the costs of correcting defects in the system-integration phase or after the system has been deployed into operation, are 15 to 30 times, and 30 to 110 times higher respectively compared to the cost of the removing the defects early—in the phase in which they were introduced.

“The goal,” says Hansson, “is to do more up-front modeling of the system to mitigate risks and integration problems, save money and time, and possibly allow construction of even larger, more complex systems with this technique.”