Stochastic modeling of a wet acetylene reactor using the Monte Carlo method and mixed density networks

e356

Abstract

The chemical industry, being involved in processes that are subject to variability and uncertainty, stochastic modeling is a valuable tool, as it allows the simulation of complex systems, risk and uncertainty assessment, process design and optimization, and production planning. In the present investigation, the modeling of a wet acetylene reactor is proposed. It is carried out taking into account not only the fundamental reaction of calcium carbide with water, but also the existence of impurities in the raw material. Based on the chemical and energy balance models, the Monte Carlo method was used to evaluate the acetylene outlet temperature, depending on the amount of water supplied, taking in to account the variability of the chemical composition of the raw material. Finally, a mixture density network was used to model the probability distribution of the acetylene output temperature. The model obtained allows not only to predict the expected value of said temperature, but also the respective reliability limits, which contributes to a more realistic decision making during the control of the reactor operation.