Middleware in Distributed Multidisciplinary Optimization (MDO)

Abstract

In distributed MDO, the MDO system is divided into parallel subsystems, where each subsystem is disciplined by constraints, collaboration of other subsystems and environmental variations. It’s desirable for the subsystem to adapt its behavior in accordance with the collaboration and environmental variations. However, the system coordination of distributed MDO usually suffers from imbalanced resource allocation among the subsystems. In contrast, the System Coordination is so greedy in the overall system benchmark that it neglects the needs of individual subsystems. Hence, the adaptation tasks of system coordination and environment variations are better provided by adaptive middleware, which is capable of catering for individual subsystem needs on a fair ground, while maintaining the optimal overall system benchmark. This paper describers the collaboration and the adaptation of middleware in distributed MDO. Current effort is to deploy the middleware based distributed MDO for conventional and unconventional aircraft optimization, targeting for a large sets of design variables and constraints. These experiments have shown the advantages and the potential of fair collaboration in distributed MDO.