Capacities of the largest new recovery boilers are steadily rising, and there is every reason to expect this trend to continue. However, the furnace designs for these large boilers have not been optimized and, in general, are based on semiheuristic rules and experience with smaller boilers. We present a multiobjective optimization code suitable for diverse optimization tasks and use it to dimension a high-capacity recovery boiler furnace. The objective was to find the furnace dimensions (width, depth, and height) that optimize eight performance criteria while satisfying additional inequality constraints. The optimization procedure was carried out in a fully automatic manner by means of the code, which is based on a genetic algorithm optimization method and a radial basis function network surrogate model. The code was coupled with a recovery boiler furnace computational fluid dynamics model that was used to obtain performance information on the individual furnace designs considered. The optimization code found numerous furnace geometries that deliver better performance than the base design, which was taken as a starting point. We propose one of these as a better design for the high-capacity recovery boiler. In particular, the proposed design reduces the number of liquor particles landing on the walls by 37%, the average carbon monoxide (CO) content at nose level by 81%, and the regions of high CO content at nose level by 78% from the values obtained with the base design. We show that optimizing the furnace design can significantly improve recovery boiler performance.
Comments on: Farkas’ Lemma: three decades of generalizations for mathematical optimization