Application of machine learning in monitoring fouling in heat exchangers in chemical engineering: A systematic review

The present work consists of a systematic literature review that examines studies on using machine learning to monitor fouling in heat exchangers in the chemical engineering area. The research was conducted in four renowned databases: SCOPUS, Science Direct, IEEE, and Web of Science. The main objective of the investigation was to identify the most prevalent machine learning methods, evaluate their performance, and analyze the challenges associated with their implementation and prospects. Using the StArt software, seven relevant scientific papers from the established review protocol. The most frequently identified methods were support vector machine (SVM) and k-nearest neighbours (k-NN), followed by decision tree. However, long-term and short-term predictors and long short-term memory (LSTM) and non-linear autoregressive with exogenous inputs (NARX) algorithms were the most successful, followed by Gaussian process regression (GPR), SVM, k-NN, and improved grey wolf optimization-support vector regression (IGWO-SVR) algorithms. Although these methods inspire confidence, it is important to highlight that they are still in the software testing phase. Key gaps identified include the need for further studies on real industrial applications and the integration of advanced sensors and measurement systems. Future directions point to developing more robust and generalized algorithms capable of dealing with the complexity of real systems.