Hybrid Algorithms for Wave Propagation

New Jersey Institute Of Technology, Newark NJ

Investigators

Abstract

This project is focused on the development of innovative and efficient algorithms dedicated to solving problems of acoustic and electromagnetic wave propagation. The strategy consists of using domain decomposition to design advanced numerical techniques for obtaining high computational efficiency, and improved convergence and accuracy properties. The proposed approach also allows for suitable utilization of parallel computing. The investigator is concerned with two classes of problems. The first consists of using domain decomposition methods to suitably combine, (1) finite element methods with boundary element methods, and (2) finite element methods with asymptotic techniques. In the second class, the investigator proposes to couple domain decomposition methods with a specific integral equation method for problems concerning multiple scatterers in the high frequency regime. The resulting algorithm bypasses the need to resolve at the wavelength scale while retaining error-controllability. A new Krylov-subspace method that significantly improves convergence of the iterative procedure will be investigated. This approach will decrease the computational time required to obtain a given accuracy. A careful mathematical analysis will be conducted to help achieve these goals. The main focus of this proposal is the development of innovative and efficient algorithms dedicated to problems of acoustic and electromagnetic wave propagation. The strategy consists of careful mathematical analysis and design of high efficiency algorithms. The work is highly interdisciplinary and has impact on advanced technological applications including problems arising in areas such as telecommunications, aircraft design, and oceanography. For example, the response of large, geometrically complex structures to incoming electromagnetic radiation is a topic of great interest to the aerospace industry, which seeks to improve stealth capabilities of airborne vehicles. Engineers regularly employ computational tools to predict the radar signature of aircraft, which helps to minimize design costs. Clearly, such efforts will be greatly aided by the development of computationally efficient and rigorous numerical methods. Some aspects of this project will be supported by an active collaboration with engineers.