Data-Driven Bayesian Modeling Using Approximate Bayesian Computation (ABC) for Heat Exchanger Monitoring

• Grant number: 24/05412-0
• Support Opportunities: Scholarships abroad - Research Internship - Scientific Initiation
• Start date: June 28, 2024
• End date: July 27, 2024
• Field of knowledge: Engineering - Mechanical Engineering
• Principal Investigator: Samuel da Silva
• Grantee: Vitória Batista Godoy
• Supervisor: Rafael de Oliveira Teloli
• Host Institution: Faculdade de Engenharia (FEIS). Universidade Estadual Paulista (UNESP). Campus de Ilha Solteira. Ilha Solteira , SP, Brazil
• Institution abroad: Institut Franche-Comté Electronique Mécanique Thermique Et Optique - Sciences Et Technologies, France
• Associated to the scholarship: 23/13794-7 - Data-Driven Bayesian Modeling for Heat Exchanger Monitoring, BP.IC

Abstract

This project aims to present a method for fault detection in heat exchangers based on monitoring temperature data that are measured and used in control processes in industrial environments. In this process, it is essential to have a numerical model as a digital replica that communicates with the process control. Bayesian calibration tools oriented by this data are a welcome way to carry out this process, as a heat exchanger has many uncertain parameters and requires stochastic modeling, especially when estimating the overall heat transfer coefficient. The undergraduate research project of the scholarship holder, Vitória, proposes to perform this calibration using the Markov Chain Monte Carlo (MCMC) method to sample the parameter distributions to be calibrated. However, the MCMC method requires knowledge and a proposal for a likelihood function; this motivated the development of more recent calibration methods that do not need this knowledge in their implementation. The Approximate Bayesian Computation (ABC) method is one of these techniques that has become quite popular. In this sense, this internship at Institut FEMTO-ST aims to implement this method in the problem of this undergraduate research. Professor Teloli (FEMTO-ST, Besançon) has recently worked with this formulation and has extensive experience in this approach. A significant scientific and cultural enrichment is expected for the scholarship holder, Vitória, during the 4-week internship in July. Finally, it is essential to highlight that during this period, the writing of a joint scientific paper is expected, considering that it has been noted that the ABC method is not yet used in solving inverse problems in heat exchangers, to the knowledge of those involved in this project. The proposal presents all the work plan details to be developed during the internship.