Employing Well Logging Data to Generate a Synthetic Model of the Formation Rock Density Applying the ANN Approach

The first challenge in conducting seismic surveys, petrophysical evaluations, assessments of the mechanical rock characteristics, and stress analyses in oil and gas fields is to ascertain the bulk density. It is a physical characteristic of rocks assessed in the laboratory on rock specimens or acquired from oil and gas wells through logging equipment. However, the rock samples are difficult to extract along the interested intervals to construct a rock density profile due to the cost and time-consuming. Additionally, most of the logging tools, especially the bulk density log, are usually not implemented in the shallow depth of the drilled borehole sections. Therefore, this study was motivated to synthesize bulk density from other well-logged data, i.e., gamma ray, neutron porosity, and sonic compressional waves. Two mathematical models of bulk density were created exploiting a dataset from a single well, employing artificial neural networks (ANNs) and multiple regression analysis (MRA) as predictive techniques. The outcomes indicated that the ANNs and MRA are comparable in predicting bulk density; however, the higher determination coefficient (۰.۹۲) and smaller root mean square error (۰.۰۶۳) of the ANNs illustrate superior accuracy compared to the MRA. Eventually, this study offers efficient and cost-saving approaches that combine traditional well logs to synthesize the rock density.

کلیدواژه ها:

Artificial neural network ، Multiple Regression Analysis ، Bulk density ، Well logging