Physics-informed AI improves predictions of how critical minerals move through rock

Understanding the movement of critical minerals through rock may hinge on comprehending fluid and chemical flow. These fluids can dissolve, transport, and concentrate valuable minerals beneath the Earth's surface. Traditionally, studying these processes required lengthy computer simulations or AI systems needing vast training data, often generating unrealistic results.

However, a novel AI-powered tool developed by researchers from the University of Houston and EMSL can predict fluid and dissolved chemical movement through rock and other porous materials underground. This advanced approach combines AI with established physical and chemical laws, enabling faster and more dependable predictions. The team's physics-informed machine learning framework, a neural network trained with data and physical laws, offers several benefits over conventional methods.

While traditional simulations are accurate but resource-intensive, and standard AI models predict swiftly but demand large datasets and may produce non-physical results, the new framework achieves rapid predictions while adhering to physical and chemical principles. By ensuring predictions remain within realistic bounds, the model effectively addresses the challenge of modeling critical minerals in trace amounts, preventing negative results due to computational errors.

This tool could aid critical mineral recovery efforts, like in situ mining and biomining, by predicting fluid injection strategies, movement duration, and conditions for maximizing mineral recovery, potentially reducing costs and improving yields.