A machine learning-based framework for early fault prediction in high-pressure pumps of reverse osmosis desalination plants: A case study from Morocco

Abstract

Employing machine learning (ML) for predictive maintenance represents a promising approach for enhancing the reliability and efficiency of critical water infrastructure, particularly in reverse osmosis (RO) desalination plants where high-pressure pumps, which play a critical role in RO operation, operate under demanding industrial conditions to address growing water scarcity. Against this backdrop, the objective of this study is to develop and validate an industrial ML-based predictive maintenance framework for early fault prediction in high-pressure pumps operating in RO desalination plants. To achieve this, real-time sensor data from 5000-labeled observations including vibration, temperature, pressure, displacement, and electrical parameters were employed in combination with six classification algorithms: random forest (RF), support vector machine, k-nearest neighbors, decision tree, artificial neural network, and naive Bayes. Furthermore, a comparative analysis was conducted to assess the performance of these algorithms, considering accuracy, precision, recall, F1-score, and AUC metrics. Results show that the RF model demonstrated superior performance, achieving 97.3% overall accuracy, a weighted F1-score of 0.97, and an AUC of 0.96, with particularly strong performance on minority fault classes (F1-scores ranging from 0.85 to 0.94). These findings highlight the potential of integrating multisensor industrial monitoring and ML-based analysis for predictive maintenance applications in RO desalination infrastructures operating under real industrial conditions.