Hybrid First Principles-Machine Learning Models
Aim: Bridge first-principles physical-chemical models with machine learning models.
Property Prediction: ANNs trained on quantum descriptors
Reactor Prediction and Control:Reinforcement learning for chemical process control
Stochastic Event Prediction:Deep learning to predict rare combustion events
Multi-component flash calculation with deep learning
Multiphase flows in geological formation are often needed to be solved numerically in reservoir engineering, as the subsurface oil and gas reservoirs typically contain multiple phases and multiple chemical species with complex phase behaviors. Flash calculation, a computational step to predict equilibrium properties of each phase for a given fluid mixture going through phase splitting, is a key step priori to multiphase flow simulation as the thermodynamic properties of the target mixture at the equilibrium sate, such as composition, densities and total phase amount, play a crucial role in designing physically-meaningful models and stable algorithms. The petroleum industry is increasingly requiring a reliable and efficient flash calculation method to handle the realistic reservoir fluid mixture containing hundreds of components, which challenges the previous methods and calls for a more efficient acceleration approach. Despite the suitability and necessity of developing deep learning-based methods in this field and their potential applications, limited attempts have been made on developing deep learning methods for accelerating flash calculation. Furthermore, all of the existing machine learning methods assume a fixed and given number of components in the mixture, which makes such models to be practically useless. In this project, we propose to develop the first self-adaptive deep learning method for general flash calculation, which can quantitatively predict the total phase amount in the mixture and related thermodynamic properties at equilibrium for various realistic reservoir fluids with a large number of compositions under different environment conditions. This research will overcome the previous component limitation of current machine learning methods used in flash calculation. By generating a larger dataset of more components and advanced network structures, our trained network can be extensively used to accelerate flash calculation for various realistic reservoir fluids.
Smart Breeding Platform: Phase 1. AI modeling for Better Crops
Since crop species have been domesticated to fit it to humanity’s needs, their genomes hold the secrets to ancient and modern agricultural practices, which can serve as an informative reference for future breeding practices. Now the world population is expected to reach 10 billion, and the agricultural community must manage to ensure a safe and sustainable food supply for our future generations. In this situation, we aim to enhance crop breeding processes by the means of AI-aided genome analysis platform, and contribute to food security as an initiative in KAUST, as well as in the Kingdom. As the long-term objective, by accelerating conventional and gene editing breeding methods in conjunction with our Smart Breeding Platform (Figure), we will create the next generation of green super crops that are high yielding and more nutritious, while at the same time having reduced ecological footprints – i.e. crops that can grow with less water, fertilizers and pesticides; can grow on marginal lands, and have reduced greenhouse gas emissions.
Machine Learning For Semiconductor Nano Technologies
Semiconductor nano technologies are among the most consequential ones for almost every aspect of the modern society. Conventionally, the design towards desired technological specifications is by no means trivial. It comprises interactions of simulations and experiments that often translate into lengthy periods and high costs. While ML has been successfully employed to tackle problems in numerous areas, few studies have been carried to investigate how to use ML to create semiconductor nano technologies comprising sophisticated structures. In this project, the PIs would utilize ML to design semiconductor nano technologies. Specifically, the structural variables will be inferred by regression models, genetic algorithms, and other generative models. To prove its effectiveness, experiments would be carried out using advanced semiconductor equipment in the PI lab and Core Labs.
Artificial Intelligence for Sustainable Synthesis of Separation Materials
Materials synthesis and optimization are tedious and labor-intensive. Consequently, we need a paradigm shift toward the prediction of materials performance without the necessity to experimentally optimize them. Developing methodologies using machine learning, a subfield of artificial intelligence, for the performance prediction and optimization of separation materials is high on the sustainability agenda. Experimental design in combination with machine learning algorithms is employed to develop materials with reduced environmental footprint.
Data Mining and Artificial Intelligence for the Prediction of Membrane Performance
There is an urgent need to develop energy-efficient separations through predictive methodologies that will fast-track the industrial implementation. The performance prediction of solvent-resistant membranes has been a daunting and challenging task, due to the complexity of the multicomponent system comprising of solvents, solutes and membranes with a high number of possible molecular interactions. Data mining is performed to build a database, which is used to explore artificial intelligence based predictive models for membrane performance.