A new concept — fluid substitution by integrating rock physics and machine learning

Abstract

This study presents a new and patented concept for fluid substitution that can be integrated with Machine Learning to provide robust and simple fluid substitution with approximately the same or better accuracy as Gassmann theory. The method is called ‘ROck physics Fluid Substitution’ (ROFS) and integrates machine learning and rock physics. ROFS allows for rapid and simple fluid substitution that in many cases give more physically consistent results than applied Gassmann theory. A stepwise workflow for the method is given. Comparison with Gassmann theory shows that the ROFS approach better predicts velocities in core plugs that are substituted from dry to brine filled. When applying the method on well logs, it is also demonstrated that for high porous rocks where the Gassmann assumptions are met, the methods give very similar results. But for intermediate-to-low porosity rocks, Gassmann theory seems to overpredict the fluid effect while the new model is more realistic. The method can be applied for both siliciclastic rocks and carbonates. By using a rock physics model for carbonates, the new method can account for the effect of microstructure variations such as pore shape variations and cracks when performing fluid substitution.