AI-based remaining useful life prediction of seawater desalination membranes

• An AI-based predictive model is proposed for RO membranes Remaining Useful Life (RUL). • Full list of the desalination system operational variables were considered including: SDI, water temperature, pump speed, and membrane age. • The model achieved an exceptional R² coefficient of 0.984. • Infuential variable identified through SHAP analysis were: SDI, water temperature, pump speed, and the membrane age. Remaining Useful Life (RUL) of Reverse Osmosis (RO) membranes is defined as the anticipated time until the RO membrane reaches a specified performance threshold, triggering maintenance actions to enhance system longevity and efficiency. We propose an artificial intelligence-based predictive model for modelling and estimating the RUL of RO membranes. Literature on the prediction of RUL for RO membranes is limited, solely based on historical water production data and does not incorporate operational conditions and parameters. We consider a detailed list of sensor and laboratory based system operational parameters such as Silt Density Index (SDI), water temperature, pump speed, and the water production history at the different stages of the desalination process to model and predict the membranes RUL. A case study from a local RO plant in the UAE is used to demonstrate the proposed work. Four machine learning based predictive models were examined. Random Forest algorithm demonstrated the best performance, achieving R² of 0.984, RMSE of 0.136, and MAE of 0.0997 indicating exceptional RUL prediction performance. SHapley Additive exPlanations (SHAP) importance analysis identified SDI, water temperature, pump speed, and the membrane age as the top RUL influential parameters. Understanding these key variables can provide valuable insights into optimizing operational conditions, thus extending membrane lifespan. Implementation of this AI-driven model optimizes clean-in-place scheduling maximizing the longevity and performance of membranes. Findings enhance the understanding of necessary operating conditions for effective seawater treatment and contribute to the broader literature on predictive maintenance in desalination, supporting more efficient and sustainable water production practices.