Parallel optimization by means of a Spectral-Projected-Gradient approach

• A parallel implementation of the Spectral Projected Gradient method was analyzed.
• Parallelism is addressed for time-demanding chemical engineering problems.
• For the parallel design, significant time bottlenecks were individualized.
• Efficiency was achieved without having to relax the problems.
• It works efficiently under clusters and multicore platforms.

The judicious exploitation of the inherent optimization capabilities of the Spectral-Projected-Gradient method (SPG) is proposed. SPG was implemented in order to achieve efficiency. The novel adjustments of the standard SPG algorithm showed that the parallel approach proves to be useful for optimization problems related to process systems engineering. Efficiency was achieved without having to relax the problems because the original model solutions were obtained in reasonable time.