Term 252

Automatic Classification of Proppant Using Computer Vision

Project Type: Self-Initiated

Project Description

Hydraulic fracturing relies on proppant to maintain fracture conductivity, and well performance is highly sensitive to particle size distribution, shape, and fines content. Current quality-control methods based on manual sieve analysis and microscopy are time-consuming, operator-dependent, and unsuitable for rapid field decisions, increasing the risk of accepting off-spec material. This project presents the conceptual design of a portable, computer-vision–based proppant classification system for real-time particle characterization at the frac site. The system uses a controlled imaging enclosure with uniform LED illumination, a removable single-layer sample tray, and edge computing for on-device processing. Particle contours are segmented and converted to equivalent diameters, which are mapped to API RP 56/60/19C size classes (20/40 and 40/70). Performance targets include classification of ≥90% of visible particles, mean size error ≤10%, and agreement within ±10% of laboratory sieve mass fractions. Validation will be conducted using certified sieve reference samples, with triplicate imaging runs per test and statistical comparison to laboratory cumulative mass retention data. Strict design constraints include maximum hardware cost of 5,500 SAR, power consumption <150 W, total mass <12 kg, footprint ≤40 × 30 × 30 cm, IEC-compliant insulated enclosure, and IPX4-equivalent dust resistance. The system is designed to operate under field conditions including airborne dust, vibration from nearby pumping equipment, and variable ambient lighting (25–45°C typical site range). Concept alternatives were evaluated based on accuracy, robustness, portability, and energy consumption, resulting in a compact tray-based batch imaging configuration. Prototype testing will include laboratory validation against sieve analysis and environmental stress assessment to verify performance under realistic field conditions.