Machine Learning Framework for the Methyl Chloride Production Process

Abstract

We report a framework for calculating the exergy and energy losses in the hydrochlorination of the methanol process for a methyl chloride production unit. Machine learning-based predictive maintenance models are identified to assess plant toxicity. The proposed novel framework integrates Hyprotech Systems (HYSYS) and machine learning models. It optimizes the operating conditions of the chloromethane plant. Both supervised and unsupervised machine learning models such as Bayesian Ridge regression (BRR), Nearest Neighbors regression (KNR), and Stochastic gradient descent regression (SGD) are used to forecast energy and exergy destruction. Among these models, the BRR exhibits exceptional accuracy in predicting thermodynamic losses, notably achieving an R2 value of 0.998. Thereby, an integrated framework could serve as a valuable diagnostic tool for assessing real-time plant data to enhance plant operational efficiency.

Keywords: energy loss, exergy destruction, machine learning framework, methyl chloride production process, Bayesian Ridge regression