In this study, we applied machine learning approach to estimate the mineralogical compositions based on elemental data acquired using x-ray fluorescence (XRF) instruments. Artificial neural networks (ANN) was used to develop new models to provide continues profiles of quartz, calcite, and clay minerals using profiles of Na, Al, Si, K, and Ca. Thereafter, the mineral-based brittleness index (MBI) was estimated using the predicted profiles of quartz, calcite, and clay minerals. The obtained results showed that the developed models can provide accurate predictions for the mineralogical profiles and brittleness index, with R2 of around 0.96. Finally, new empirical correlations were extracted from the ANN models, which can provide accurate and quick estimations for the mineralogical composition. The ANN-based equations were validated using testing data; very acceptable performance was obtained with R2 higher than 0.95.

This work therefore will have benefit in obtaining high-resolution mineralogy using XRF data without sending many samples for XRD laboratory measurements. Eventually, the predicted mineralogy can be used to quantify the BI and enable accurate predictions that result in better de-risking strategies and evaluating successful unconventional plays in terms of estimations of source-rock quality, identification of sweet spots, and designing/executing well placement and hydraulic fracturing stages.


Because of low porosity and insufficient permeability to allow the hydrocarbons flow naturally, hydrocarbons within organic-rich mudstones or referred to as "unconventional” typically extracted by a combination of vertical and horizontal drilling followed by multi-staged hydraulic fracturing. Fracking will increase effective permeability and allows the hydrocarbon to be released and economically produced (Lee et al., 2011; Sone and Zoback, 2013; Dong et al., 2017). Several key parameters were used to evaluate sweet spot as well as designing drilling, completion and stimulation parameters. Therefore, it is important to identify rock compositions and understand the factors affecting their properties, known as brittleness index (BI).

This content is only available via PDF.
You can access this article if you purchase or spend a download.