Crude oil properties such as viscosity, molecular composition, and saturate, aromatic, resin, and asphaltene (SARA) fractions are crucial parameters for evaluating reservoir quality, producibility, and compartmentalization. In the past, physical and empirical models that relate oil properties to NMR measurements have been developed. How-ever, the existing models are too simplistic to accurately predict properties of crude oils which are complex mixtures of hydrocarbon and non-hydrocarbon molecules.
This paper introduces a model-independent technique for quantitative predictions of live-oil properties from NMR measurements. The technique assumes that the physics connecting NMR measurements to oil properties is implicitly contained within a database of NMR and fluid-property measurements made on a representative suite of live oils. The input measurements are mapped to oil properties using a mapping function that is a linear combination of Gaussian radial basis functions. The parameters of the mapping function are determined from the database. The mapping function predicts properties from input measurements made on live oils that are not in the database.
To validate the technique, an extensive database of NMR and fluid-property measurements made on live oils at elevated temperatures and pressures was acquired. Viscosities, molecular compositions, and SARA fractions were accurately determined from NMR measurements using the mapping function technique.