Summer Engineering Innovation Program - Levarix - Rice Center for Engineering Leadership

This project aims to develop an AI agent to optimize Electric Submersible Pump (ESP) designs. The project has deadline of 10 weeks.

The AI agent will optimize publicly defined ESP designs based on several key parameters: the energy needed to rotate stages, the capacity to allow different particle sizes through impellers, the delta pressure generated by the stage to lift fluid, and suitability for traditional oil/gas or geothermal wells.

A significant component is creating an interactive interface. This interface will allow users to select specific design spaces and will generate links to the 3D geometry of the stages, along with a spec sheet detailing the operating envelope.

Project goals include conducting a literature review, defining requirements, defining parameters, collecting data, developing the optimization algorithm, and building the AI agent. The project will culminate in testing and validation of the AI agent.

Key deliverables include a fully functional AI agent with an interactive interface, a comprehensive project report, and a final presentation. Expected outcomes are an AI agent capable of effectively optimizing ESP designs and an enhanced understanding of ESP optimization techniques for energy sector applications.

The primary challenge this project aims to solve is the complex, multi-objective optimization of Electric Submersible Pump (ESP) designs. Currently, designing an ESP involves balancing several competing performance parameters: minimizing the energy needed to rotate stages, maximizing the ability to handle various particle sizes, achieving sufficient delta pressure to lift fluid, and ensuring suitability for specific applications like oil and gas or geothermal wells.

Manually finding a design that optimally satisfies all these criteria simultaneously is difficult because these objectives often conflict. For instance, a design that offers high delta pressure might consume more energy or be less tolerant of large particles.

The project addresses this by developing an AI agent that uses Pareto front analysis. This approach allows for the identification of a set of optimal designs where no single parameter can be improved without degrading at least one other parameter. This provides engineers with a range of superior design choices, rather than a single, potentially compromised solution, thereby overcoming the challenge of navigating these complex design trade-offs more effectively.