Development and application of machine learning‐based prediction model for distillation column

Abstract

Distillation is an energy‐consuming process in the chemical industry. Optimizing operating conditions can reduce the amount of energy consumed and improve the efficiency of chemical processes. Herein, we developed a machine learning‐based prediction model for a distillation process and applied the developed model to process optimization. The energy consumed in the distillation process is mainly used to control the temperature of the distillation column. We developed a model that predicted temperature according to the following procedure: (1) data collection; (2) characteristic extraction from the collected data to reduce learning time; (3) min–max normalization to improve prediction performance; and (4) a case study conducted to select the artificial neural network algorithm, optimization method, and batch size, which are the most appropriate elements for predicting production stage temperature. The result of the case study revealed that the most appropriate model was observed with a root mean squared error of 0.0791 and a coefficient of determination of 0.924 when the long short‐term memory algorithm, Adam optimization method, and batch size of 128 were applied. We calculated the amount of steam consumption required to consistently maintain the production stage temperature by utilizing the developed model. The calculation result indicated that the amount of steam consumption was expected to be reduced by approximately 14%, from an average flow rate of 2763–2374 kg/h. This study proposed a control method applying a machine learning‐based prediction model in the distillation process and confirmed that operation energy could be reduced through efficient operation.

Inequality distance hyperplane multiclass support vector machines

“…The key of the solving Equation ( 7) is the minimum of (6). So, we have to take the partial derivative of b ξ w , , m m i m in Equation (7).…”

Development of physical property prediction models for polypropylene composites with optimizing random forest hyperparameters

“…Recently, ML‐based prediction models are being developed in various industries as well as in the materials industry 14–18 . Coley et al developed a prediction model using a convolutional neural network for four properties of molecules: aqueous solubility, octanol solubility, melting point, and toxicity 14 .…”

“…They demonstrated the use of neural network‐based models that do not rely on exhaustive molecular descriptor calculations or experimental parameters. Kwon et al used an ML‐based prediction model to predict the temperature of the distillation column for controlling a chemical process 15 . They optimized the operating conditions of the distillation process using the prediction model reduced steam consumption by approximately 14% using the prediction model.…”

Methoxy‐methylheptane as a cleaner fuel additive: An energy‐ and cost‐efficient enhancement for separation and purification units

“…The net present value was compared for the modified configurations and with the base case such that those modified configurations showed a significant improvement. Kwon et al 26 carried out the energy optimization and satisfy the target product purity by using the developed machine learning-based prediction model into the n-butane separation process. The estimated reduction of the steam flow rate was then obtained by using the predicted model.…”