(194d) Optimization of Fluidized Bed Incineration Process for Explosive Waste Treatment Via Artificial Neural Network Surrogate Modeling Method

In this study, the operating conditions for a novel method of explosive waste disposal were optimized to minimize the formation of NOx air emissions and the operating cost. Previous methods of disposal include open burning and detonation and the more recent use of rotary kilns. However, these methods have disadvantages, particularly those associated with NOx emissions. To address this, development of a fluidized bed reactor explosive waste incinerator was proposed. Because this is a new application of the technology, little experimental process data are available. In such cases, computational fluid dynamics (CFD) simulations are typically used to evaluate the preferred operating conditions for the reactor. However, as the number of variables to consider increases, the optimization of the design through CFD simulations alone becomes challenging because of the significant computational time required. Thus, a more efficient method is needed.
Herein, we proposed a surrogate model of CFD output using an artificial neural network (ANN) that can be used to locate the optimal operating conditions. To perform the optimization on seven variables, namely, inlet gas velocity, temperature, pressure, feed particle size, TNT/water mass ratio, filled sand ratio and inlet gas composition, finite ranges of these variables were selected and sampled using a Latin hypercube method, yielding 500 random input samples. CFD simulations were performed using each of these conditions, and the surrogate model was designed to match the results. Using this surrogate model, 50% reduction in NOx emissions or 35% reduction in operating cost from the process was achievable. The surrogate model was then used to perform a sensitivity analysis for each variable, and the inlet gas velocity and reaction temperature had the greatest effect, whereas the TNT/water mass ratio had the smallest effect.
Overall, the fluidized bed reactors show promise with regard to the reduction of NOx emissions and operating cost from explosive waste disposal. In addition, surrogate modeling with ANNs represents a useful method for reducing the CFD computation time used for multi variable optimization of such a fluid bed reactor.