Machine Learning-Based Dynamic Modeling for Process Engineering Applications: A Guideline for Simulation and Prediction from Perceptron to Deep Learning

Abstract

A misusage of machine learning (ML) strategies is usually observed in the process systems engineering literature. This issue is even more evident when dynamic identification is performed. The root of this problem is the gradient explode and vanishing issue related to the recurrent neural networks training. However, after the advent of deep learning, these issues were mitigated. Furthermore, the problem of data structuration is often overlooked during the machine learning model identification in this field. In this scenario, this work proposes a guideline for identifying ML models for the different applications in process systems engineering, which are usually for simulation or prediction purposes. While using the proposed guideline, the work also identifies a virtual analyzer for a pressure swing adsorption unit. In these types of adsorption separations, it is usual that the measurement of the main properties is not done online. Therefore, the virtual analyzer is another contribution of this manuscript. The overall results demonstrate that even though the test provides good performance during the ML model identification, its quality might degenerate over the application domain if the model application is overlooked.

Mapping Uncertainties of Soft-Sensors Based on Deep Feedforward Neural Networks through a Novel Monte Carlo Uncertainties Training Process

“…Initial learning rate On the other hand, it is necessary to ensure that the model used is suitable for what is desired. In this sense, several works in the literature have pointed to the deep neural network as the most suitable machine learning solution to model complex dynamic systems [10,13,29]. For instance, Rebello et al, 2022 [13] and Oliveira et al, 2020 [10] compared several machine learning approaches, concluding that deep learning was able to better describe the dynamics of a pressure swing adsorption unit.…”

“…In this sense, several works in the literature have pointed to the deep neural network as the most suitable machine learning solution to model complex dynamic systems [10,13,29]. For instance, Rebello et al, 2022 [13] and Oliveira et al, 2020 [10] compared several machine learning approaches, concluding that deep learning was able to better describe the dynamics of a pressure swing adsorption unit. Schweidtmann et al, 2021 [29] points out that among ML techniques-such as random forests, support vector machines, spline functions, among others-deep learning is the most suitable for learning complex dependencies.…”

“…In general, this step uses fewer epochs to allow the method to explore a wider region of the hyperspace. This is a usual approach in the literature [9,13,[26][27][28]. Once the hyperparameters are defined, the epochs are increased for the training of the final structure.…”

“…Thus, imperfections, hypotheses, and idealizations imposed during building a model contribute to its uncertainty [2]. The soft sensor literature has pointed out the issue of forecast uncertainty as an unresolved issue in this field [13].…”

Novel Framework for Simulated Moving Bed Reactor Optimization Based on Deep Neural Network Models and Metaheuristic Optimizers: An Approach with Optimality Guarantee

“…Each method has its own advantages and disadvantages. The literature in ANN-based surrogates presents relevant usages of both methods. ,− …”