Prediction and optimization of emissions in cement manufacturing plant under uncertainty by using artificial intelligence-based surrogate modeling

Uncertainties in the process industries are the biggest challenge for smooth operation. This study is based on the use of artificial intelligence based surrogate modeling for predicting and optimizing emissions in combustion sections of cement manufacturing plants under uncertainty. Uncertainties in feed flow rate, kiln air flow rate, tertiary air flow rate, and coal flow rate in the combustion sections of the plant are the subject of the study. Initially, an Aspen Plus model of the kiln and calciner units of the cement plant was developed and converted into a dynamic mode to generate 700 data samples with 10% uncertainty. Then, genetic algorithm (GA), particle swarm optimization (PSO), and GA-PSO frameworks were used to optimize the process conditions using Artificial Neural Network (ANN) models as surrogates. The GA-PSO framework exhibited an advantage over GA and PSO in predicting and optimizing CO, and CO emissions. Besides, ANN models were used as a surrogate within the SOBOL and Fourier Amplitude Sensitivity Test frameworks for performing sensitivity analysis of the process. The sensitivity analysis was performed to identify the process conditions with the highest sensitivity toward the emissions. Based on sensitivity analysis, total coal and tertiary air were found to be more influencing parameters on process output. This study provides a baseline for real-time predicting and optimizing emissions at the cement plant.