Tailor-made ML solution for a hydrocracking plant in a refinery company

DogwoodAI developed the ML model to predict the yield and quality from a hydrocracking plant. The cutting-edge solution is capable of predicting the products within an error margin of less than 1% in real time. This solution was actively collaborated with a partner company to monitor and improve the product prediction in the real field. This solution was validated for 6 months in a Korean refinery to predict products in real time.

Tailor-made ML solution for a hydrocracking plant

Hybrid Simulator based on machine
learning and mathematical model for the fine chemical production

Hybrid Simulator for a fine chemical plant

Leveraging collaborations with a global platform company, DogWoodAI is developing a Physics-informed data-driven Hybrid AI model that integrates mathematical and ML models. This breakthrough approach addresses the limitations of mathematical modeling in complex industrial settings, as well as the challenges of insufficient data for ML model training. Our mission is to create ML models that complement the shortcomings of each approach, providing tailored solutions for factories facing these challenges. The developed hybrid simulator with less than 0.5 deviation is set to a domestic fine chemical company to predict product quality and yield. Then, it will be applied for designing and constructing the second plant.

Hybrid Simulator for a fine chemical plant: Interview with AVEVA

Jae-Chul Lee, Head of SimSci APAC Technical team at AVEVA, discusses a revolutionary seven-month project with DogWoodAI, aimed at enhancing a chemical plant's operations through 'Hybrid Simulation'. This method merges AVEVA's first-principle simulations with machine learning to predict product quality and yield. The interview sheds light on the potential of Hybrid Simulations to transcend traditional commercial simulators, promising to expand into various manufacturing sectors with DogWoodAI's continued collaboration.

Computational Fluid Dynamic models for three-phase reactor in a fine chemical company

This project is to simulate and analyze the flow patterns within a reactor producing solid particles via liquid-gas reaction in a fine chemical company, using advanced computational fluid dynamics models (CFD). By investigating the correlation among the three-phase (gas-liquid-solid) flows as well as the mixing patterns inside the reactor, the optimal impeller design and operating conditions for maintaining the best mixing performance are determined. It will progress further development for CFD-based ML model.

CFD results for the mixing performance of the reactor
(Gas-Liquid-Solid phases)

Consulting Project: Xenon recovery process

DogWoodAI established a model-based framework for a cyclic adsorption process to separate He–Xe, employing a three-step approach: constructing physicochemical database from literature and experimental data, developing a breakthrough simulator validated with experimental results, and process simulations. This integrated strategy confirmed the process feasibility and provided guidelines for large-scale design. Additionally, DogWoodAI’s modeling expertise demonstrates a strong capacity to develop advanced model-based solutions, forming a solid foundation for further model-based AI solutions.

Xe recovery process results

Advanced AI Solutions for Fine Chemical Plants: DogWood Autonomous Manufacturing

DogWoodAI successfully launched DogWood_Autonomous Manufacturing Plant for a fine chemical plant, including real-time prediction, operating data analysis, optimization guidance, and safety & maintenance analysis. The AI solution was meticulously designed to address the dynamic and interconnected nature of a total plant, including scenarios influenced by aging catalysts. The AI solution is implemented in a fine chemical plant starting from Jan. 2025.

DogWood_Autonomous Manufacturing Plant, a unified framework of this solution, utilizes real-time plant data to analyze information for each tag and accurately predicted product quality in real time. It provides actionable insights on the optimal operating conditions needed to achieve the desired quality. Additionally, the solution deliveresreal-time analysis of anomalies in the plant, issued safety alerts, and generated detailed analytical reports, enabling safe and efficient plant maintenance.

DogWood Autonomous Manufacturing

Project

Sub menu