Optimal Control of the Process Systems Using Graphic Processing Unit
School of Engineering and Energy, Murdoch University, Western Australia, 6150
The 18th IFAC World Congress, 2011
In this paper the Graphic Processing Unit (GPU) is applied in order to improve the computational performance of process systems optimal control calculations. To apply GPU massive parallel architecture, a simplified version of interior point optimisation algorithm was selected and modified to fulfil special hardware requirements of GPU architecture. In this algorithm, a damped nonlinear Newton with preconditioned iterative linear solver was used to solve Dual-Primal equations. The comparison of results between this implementation and standard CPU-based implementation shows considerable improvement in computational performance.
October 31, 2011 by hgpu
