An important part of converting heavy crude oil into gasoline and other products is the fluid catalytic cracking process. First used commercially in 1915, the process has undergone several improvements since then, but researchers in China say it can be safer and more effective. Answer? A.I. Researchers published April 7 in the journal Big Data Mining and Analytics how artificial intelligence could improve the catalytic cracking process that chemically breaks down the long molecules that make up heavy crude oil and separates them into the desired products.
“Safety, efficiency, and environmental protection are the main goals of oil refining,” said Fan Yang, a researcher at Sichuan University’s College of Information. Yang is also associated with the New Hope Liuhe Algorithm and the Big Data Center. “Abnormal operating conditions, early warning, product yield analysis and optimization, and flue gas desulfurization analysis and optimization are hot spots we are exploring to improve safety, efficiency, and environmental protection, respectively.”
According to Yang and co-author Mao Xu, a researcher at New Hope University’s Data Intelligence Lab, advances in data collection and artificial intelligence used to make sense of the data are the way forward. “With improvements in industrial data collection technology, more data can be obtained for analysis,” Xu said. “Advances in artificial intelligence will allow us to analyze this data more precisely.”
“The primary goals of petroleum processing are safety, efficiency, and environmental protection. We’re looking into abnormal operating conditions, early warning, product yield analysis and optimization, and flue gas desulfurization analysis and optimization to improve safety, efficiency, and environmental protection.”
Fan Yang, a researcher with the College of Computer Science at Sichuan University.
To better study and optimize the catalytic cracking process, the team studied neural networks. This type of artificial intelligence simulates the way the human brain processes data by using interconnected nodes to quickly analyze large amounts of data. As you process it, you learn and discover that combining seemingly disparate data points can reveal a larger problem or potential opportunity. This includes temperature, steam production, and desired product yield.
Neural networks are based on the human brain, but they are not limited to human concentration. They can collect thousands of data points about these three variables and see how they interact to produce different results.
“This kind of machine learning is data-driven and can automatically and efficiently solve large-scale problems,” Xu said. “By combining machine learning with mechanism models or algorithms that understand how mechanisms drive actions, uncertainty can be further reduced and predictive performance can be improved.”
Xu noted that other machine learning models that describe nonlinear relationships between chemical processes in catalytic cracking can be applied to neural networks. In this approach, researchers can select specific features or reduce the number of dimensions that the AI considers. “This paper provides a comprehensive overview of the analysis of catalytic cracking processes, mainly introducing traditional mathematical mechanisms and artificial intelligence-based methods,” Yang said. “Neural network methods have great advantages because they can effectively handle the high-order and nonlinear nature of the catalytic cracking process, yielding good results in process analysis and optimization studies.”
The researchers said they plan to test the neural network, which works well in simulations, in a real manufacturing process. “In future studies, hybrid models that combine mechanism models and artificial intelligence algorithms are expected to become powerful tools for comprehensive and accurate chemical process analysis and production outcome prediction,” Yang said. “These technologies will play an important role in the future development of the chemical industry and will be of great value.”
More information: Fan Yang et al, Artificial Intelligence Methods Applied to Catalytic Cracking Processes, Big Data Mining and Analytics (2023). DOI: 10.26599/BDMA.2023.9020002
