(56c) Real-Time Design Modeling for Control Valve Size and Type Using Machine Learning and Big Data Streaming System

In the chemical processing facility, the control valve assumes a crucial function in the regulation of fluid flow and pressure. Also, the selection of an appropriate valve type is intricately linked to the operational efficiency and safety of the process. The selection of valve types relies on the expertise and experience of specialists, necessitating a considerable investment of time and financial resources.
To solve the problem, we proposed a method for predicting valve size and type based on machine learning approaches. To analyze the predictive performance of the model according to the phase of the fluid, we extracted the dataset by dividing the data of three fluid phases: liquid, gas, and steam. Additionally, the proposed method was evaluated using a F1 score and a receiver operating characteristic (ROC) curve as assessment metrics.
Moreover, in this study, process data is transmitted in real-time to a cloud server through Kafka, a data streaming system. For the prediction of control valve size and types, the machine learning analysis model is executed on a cloud server, and the prediction results are transmitted in real-time to the process and control engineering platforms through Kafka.
Therefore, the application of the proposed method in the design process, it is anticipated a reduction in the time and cost associated with determining control valve size and type.