Machine Learning Enabled Fracture Network Imaging Using The Wirelessly powered Smart Microchip Proppants Technology

Fracture mapping is a longstanding necessity in understanding the behavior of rocks before, during, and after hydraulic fracturing. High-resolution subsurface imaging (i.e borehole imaging) to interpret fracture networks enables improving the simulation model and ML-based surveillance tools to monitor the created fracture geometry. This paper seeks to expand the interpretation from borehole imaging using the geo-sensor data generated from 40MHz Wirelessly-Powered Smart Microchips Proppants or abbreviated as SMPs (using 180 nm CMOS technology). In this paper, an approach that uses the novel SMPs’ data to map fracture geometry more accurately is proposed, namely Intelligent Fracture Diagnostic Procedure (IGSFD). This workflow processes the transmissible data from the SMPs and generates the predictive fracture network image(s). I-GSFD is integrated between multi-dimensional projection, unsupervised clustering, and image processing algorithms. Affine transformation and a shallow ANN are integrated to control the quality of clustering.I-GSFD proves its efficacy in fracture mapping for the 3 in house designed 3D synthetic fracture networks with 100% consistency, between 75-100% in prediction capability, and between 75-100% in execution robustness. The low bounds of prediction capability (75%) and execution robustness (75%) scores are caused by the “joint” locations in the synthetic fracture network with the highest complexity. Affine transformation and the shallow ANN increase the quality of unsupervised clustering in I-GSFD. However, the joints” in one of the synthetic 3D fracture networks are still a challenge for the I-GSFD platform.