Autonomous Experimentation to Accelerate Boiling Heat Transfer Research
Boiling heat transfer stands out as a highly efficient method for dissipating heat, utilized across a spectrum of engineering systems ranging from nuclear reactors to computer processing units. Over the course of decades, researchers have heavily relied on direct visualization of boiling phenomena t...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Published: |
Massachusetts Institute of Technology
2024
|
| Online Access: | https://hdl.handle.net/1721.1/155613 |
| _version_ | 1826198707159171072 |
|---|---|
| author | Ravichandran, Madhumitha |
| author2 | Bucci, Matteo |
| author_facet | Bucci, Matteo Ravichandran, Madhumitha |
| author_sort | Ravichandran, Madhumitha |
| collection | MIT |
| description | Boiling heat transfer stands out as a highly efficient method for dissipating heat, utilized across a spectrum of engineering systems ranging from nuclear reactors to computer processing units. Over the course of decades, researchers have heavily relied on direct visualization of boiling phenomena through photography and videography to understand this phenomenon and develop modeling tools. However, due to the complexities of the boiling process, particularly near the surface, extracting essential information from these techniques has proven challenging. The primary impetus behind this thesis stems from the pressing need to advance current knowledge in boiling heat transfer, coupled with the potential offered by advancements in high-resolution imaging, e.g., infrared thermometry, and machine learning. The primary focus of this research work is on creating online platforms capable of real-time measurement of essential bubble dynamic parameters, including nucleation site density, bubble growth time, wait time, departure diameter, dry area fraction and distribution, and contact line density. To this end, a machine learning (ML) methodology was developed to accelerate the analysis of infrared data obtained during boiling heat transfer investigations. Deep neural network models were developed to directly measure bubble growth time, period, and nucleation site density from radiation recorded by a high-speed infrared camera. Further, a comprehensive physical framework utilizing Binary Neural Networks (BNN) supported by direct memory access (DMA) of binary data recorded by an IR camera was established to implement ML prediction models seamlessly in real-time during experimental procedures.
A machine learning-based model was developed to predict the Departure from Nucleate Boiling Ratio (DNBR) using unprocessed, time-dependent infrared radiation distributions without prior knowledge of heat flux, achieving over 95% accuracy. This methodology enabled online and quasi-real-time monitoring of DNBR and estimation of Critical Heat Flux (CHF) during boiling experiments using infrared thermometry, a crucial step toward implementing intelligent, autonomous experiments. Further, leveraging explainable AI models, a hypothesis-free data-driven methodology was developed to elucidate the importance of fundamental boiling parameters in predicting the boiling crisis. The analysis revealed that parameters such as nucleation site density, bubble departure frequency, growth time, and footprint radius are all necessary and equally important.
Finally, high-resolution infrared thermography and the developed online measurement framework were utilized to demonstrate the presence of self-organized criticality (SOC) in nucleate boiling. SOC was observed to emerge from bubble interactions near the wall and was characterized by nucleation site density, average bubble footprint radius, and the product of average bubble growth time and frequency. A new approach to estimate CHF from subcritical heat flux input conditions, alongside relevant critical quantities, was presented, providing insights into the onset of SOC and its implications for boiling heat transfer. |
| first_indexed | 2024-09-23T11:08:51Z |
| format | Thesis |
| id | mit-1721.1/155613 |
| institution | Massachusetts Institute of Technology |
| last_indexed | 2024-09-23T11:08:51Z |
| publishDate | 2024 |
| publisher | Massachusetts Institute of Technology |
| record_format | dspace |
| spelling | mit-1721.1/1556132024-07-11T03:40:38Z Autonomous Experimentation to Accelerate Boiling Heat Transfer Research Ravichandran, Madhumitha Bucci, Matteo Massachusetts Institute of Technology. Department of Nuclear Science and Engineering Boiling heat transfer stands out as a highly efficient method for dissipating heat, utilized across a spectrum of engineering systems ranging from nuclear reactors to computer processing units. Over the course of decades, researchers have heavily relied on direct visualization of boiling phenomena through photography and videography to understand this phenomenon and develop modeling tools. However, due to the complexities of the boiling process, particularly near the surface, extracting essential information from these techniques has proven challenging. The primary impetus behind this thesis stems from the pressing need to advance current knowledge in boiling heat transfer, coupled with the potential offered by advancements in high-resolution imaging, e.g., infrared thermometry, and machine learning. The primary focus of this research work is on creating online platforms capable of real-time measurement of essential bubble dynamic parameters, including nucleation site density, bubble growth time, wait time, departure diameter, dry area fraction and distribution, and contact line density. To this end, a machine learning (ML) methodology was developed to accelerate the analysis of infrared data obtained during boiling heat transfer investigations. Deep neural network models were developed to directly measure bubble growth time, period, and nucleation site density from radiation recorded by a high-speed infrared camera. Further, a comprehensive physical framework utilizing Binary Neural Networks (BNN) supported by direct memory access (DMA) of binary data recorded by an IR camera was established to implement ML prediction models seamlessly in real-time during experimental procedures. A machine learning-based model was developed to predict the Departure from Nucleate Boiling Ratio (DNBR) using unprocessed, time-dependent infrared radiation distributions without prior knowledge of heat flux, achieving over 95% accuracy. This methodology enabled online and quasi-real-time monitoring of DNBR and estimation of Critical Heat Flux (CHF) during boiling experiments using infrared thermometry, a crucial step toward implementing intelligent, autonomous experiments. Further, leveraging explainable AI models, a hypothesis-free data-driven methodology was developed to elucidate the importance of fundamental boiling parameters in predicting the boiling crisis. The analysis revealed that parameters such as nucleation site density, bubble departure frequency, growth time, and footprint radius are all necessary and equally important. Finally, high-resolution infrared thermography and the developed online measurement framework were utilized to demonstrate the presence of self-organized criticality (SOC) in nucleate boiling. SOC was observed to emerge from bubble interactions near the wall and was characterized by nucleation site density, average bubble footprint radius, and the product of average bubble growth time and frequency. A new approach to estimate CHF from subcritical heat flux input conditions, alongside relevant critical quantities, was presented, providing insights into the onset of SOC and its implications for boiling heat transfer. Ph.D. 2024-07-10T20:18:42Z 2024-07-10T20:18:42Z 2024-05 2024-06-13T16:26:28.608Z Thesis https://hdl.handle.net/1721.1/155613 In Copyright - Educational Use Permitted Copyright retained by author(s) https://rightsstatements.org/page/InC-EDU/1.0/ application/pdf Massachusetts Institute of Technology |
| spellingShingle | Ravichandran, Madhumitha Autonomous Experimentation to Accelerate Boiling Heat Transfer Research |
| title | Autonomous Experimentation to Accelerate Boiling Heat Transfer Research |
| title_full | Autonomous Experimentation to Accelerate Boiling Heat Transfer Research |
| title_fullStr | Autonomous Experimentation to Accelerate Boiling Heat Transfer Research |
| title_full_unstemmed | Autonomous Experimentation to Accelerate Boiling Heat Transfer Research |
| title_short | Autonomous Experimentation to Accelerate Boiling Heat Transfer Research |
| title_sort | autonomous experimentation to accelerate boiling heat transfer research |
| url | https://hdl.handle.net/1721.1/155613 |
| work_keys_str_mv | AT ravichandranmadhumitha autonomousexperimentationtoaccelerateboilingheattransferresearch |