(40a) Hybrid Machine Learning and Fundamental Modeling for Real-Time Optimization of a Fluidized Bed Roaster

The advantage of hybrid modeling for the advanced process control (APC) applications is emphasized in [1] with advantages that utilize the strengths of physics-based [2] and data-informed modeling [3]. The hybrid approach uses ‘a priori‘ knowledge in the form of a nonlinear physics-based model with empirical model elements in a framework with several different methods and applications [4-7].

In the current study, a similar approach is used with process operation data from a fluidized bed roaster for gold ore processing. A certain type of gold ore, so-called refractory carbonaceous ore, contains naturally occurring carbonaceous material that harmfully affects the cyanidation gold recovery process by encapsulating the solubilized gold in it. The main purpose of the roaster is to oxidize these carbonaceous materials.

The reaction kinetics of the fluidized bed reactor is modeled with a shrinking core model (SCM) in conjunction with the heat and material balance equations. The unknown parameters in the model are effective diffusivity, particle size, and heat transfer coefficients that are empirically estimated with operation data.

The hybrid physics and machine learning model is used in a real-time optimization (RTO) application that sends target operating ranges to a model predictive controller (MPC). The RTO hybrid model demonstrates higher accuracy over traditional physics-based or empirical neural net approaches.

[1] Park, J., Hansen, B., Gates, N., Darby, M., Hedengren, J.D., Use of Nonlinear and Machine Learning Techniques for Improved APC Modeling, AIChE Spring Meeting, New Orleans, LA, April 2019.

[2] Park, J., Webber, T.R., Asgharzadeh Shishavan, R., Hedengren, J., Improved Bottomhole Pressure Control with Wired Drillpipe and Physics-Based Models, SPE-184610-MS, SPE/IADC Drilling Conference and Exhibition, The Hague, The Netherlands, 14-16 March 2017.

[3] Eaton, A.N., Beal, L., Thorpe, S., Hubbell, C., Hedengren, J.D., Nybø, R., Aghito, M., Real Time Model Identification Using Multi-Fidelity Models in Managed Pressure Drilling, Computers and Chemical Engineering, Vol 97, 2 February 2017, pp. 76–84, doi:10.1016/j.compchemeng.2016.11.008.

[4] Psichogios, D.C., Ungar, L.H., A hybrid neural network‐first principles approach to process modeling. AIChE Journal. 1992;38(10):1499-1511.

[5] Himmelblau, D.M., Applications of artificial neural networks in chemical engineering. Korean Journal of Chemical Engineering. 2000;17(4):373-392.

[6] Pathak J., Wikner A., Fussell R., Chandra, S., Hunt, B., Girvan, M., Ott, E., Hybrid forecasting of chaotic processes: Using machine learning in conjunction with a knowledge-based model. Chaos: An Interdisciplinary Journal of Nonlinear Science. 2018;28(4):41101.

[7] Von Stosch M, Oliveira R, Peres J, de Azevedo SF. Hybrid semi-parametric modeling in process systems engineering: Past, present and future. Comput Chem Eng. 2014;60:86-101.