We present a new digital solution based on a novel technique to predict acid gas membranes remaining performance based on the field data. Gas membranes are widely used onshore and offshore for acid gas removal from natural gas due to their efficiency and compactness. These systems are proven and well accepted, however their performance is highly dependent on field operations practices and conditions of the natural gas stream that feeds the system. If operating conditions are not controlled, the system performance can deteriorate. The weakened performance can lead to undesirable product gas specifications, contractual penalties, unexpected downtime, and ultimately the risk of environmental impact. On the other hand, maintenance anxiety and uncertainty can lead to overspend on membrane elements replacements; increasing overall operating expenditures. We developed the new technique during the past two years to allow the system operator to anticipate performance upsets by predictive monitoring and active machine learning using field operations data of gas membrane systems. This technique has adopted one of recursive Bayesian estimation techniques, linear Kalman filtering, and allows operators to predict and manage remaining membrane performance in the field proactively thereby optimize the membrane replacement expenditure.