| dc.description.abstract | Phase Change Materials (PCMs) have garnered significant attention over recent years due to their efficacy for thermal energy storage (TES) applications. High latent heat of PCMs enable enhanced storage densities which in turn translate into compact form factors of TES platforms utilizing PCMs. TES platforms mitigate the deviations between energy demand and supply, i.e., they absorb heat during periods of surplus in energy supply and release the stored energy during periods of deficit in energy supply. PCMs are envisioned for futuristic applications and are also deployed in a wide range of TES platforms in the industry - ranging from solar power plants, building energy management, HVAC, chemical process industries, waste heat recovery systems (such as desalination and food processing), domestic water heating, to thermal management of electronics (to list a few). Inorganic PCMs have high latent heat values (compared to organic PCMs) but suffer from several reliability issues. A major reliability issue with inorganic PCMs is the high degree of supercooling needed to initiate nucleation (which compromises the reliability, net energy storage capacity and power rating of the TES platform). “Cold Finger Technique (CFT)” can obviate these issues wherein a small fraction of the total mass of PCM is left in a solid phase to aid spontaneous nucleation (thus reliability is enhanced at a marginal expense to the net storage capacity while power rating of the TES remains unaffected). In this study, a machine learning (ML) technique, more specifically artificial neural networks (ANN), are implemented to enhance the efficacy of CFT. Temperature transients from PCM melting experiments are used as inputs to an ANN to predict the time required to attain a predefined melt percentage in real time. Performance and efficacy of the machine learning techniques that utilize surface temperature transients (instead of measurement of PCM temperatures) is also investigated. Surface temperature-based monitoring strategies can be applied to PCM filled heat sinks to enhance operational reliability of thermal management platform using non-intrusive approaches (such as for electronics applications). The results show that an artificial neural network is capable of providing predictions apriori regarding the time to attain a chosen melt-fraction (e.g., 90% melt-fraction) with a considerable accuracy. | |
