Electrocaloric (EC) cooling offers great potential to build efficient solid-state cooling devices that are quiet, low weight, and compact. With reported coefficient of performance (COP) as high as 13, EC cooling devices are a promising replacement for noisy, vapor compression systems (COP \< 4.5) and a candidate for microsystems (electronics, medical) cooling and self-cooling garments. However, the physics behind the EC temperature change are elusive.  The current search for effective EC materials relies on the instincts of experts and brute force experimental synthesis of ceramics, polymers, and composites. There is no robust physical model to predict the EC temperature change based on material properties. Furthermore, inconsistencies between the EC temperature change obtained from direct and indirect measurements plague the community. The direct measurement records the EC film temperature directly and is sensitive to unaccounted heat losses. The indirect measurement tracks the polarization and uses the Maxwell relations to calculate the EC temperature change. Although polarization is easier to measure than temperature, there is debate as to if the indirect measurement is applicable to non-ergodic relaxor ferroelectrics.

 

In this regard, we are developing a physics-informed machine learning model that can predict the EC temperature change given the material composition and easily-measured material properties (e.g., dielectric constant, density, Curie temperature, polarization, and pyroelectric coefficient). This work is the first application of machine learning to EC cooling. We gather experimental data on the hundreds of EC ceramics, polymers, and composites that have been synthesized and characterized from literature. The data is then cleaned and organized into human-readable json files, which will later be made public to the international research community via GitHub. The physical origin of the EC effect is considered as we select material properties as descriptors. These descriptors contain information from a macroscopic perspective (i.e., material properties) and an atomic level (for example, ionization potential, electron affinity, electronegativity, atomic radius, and position on the periodic table). Feature engineering techniques such as principal component analysis are used to prune material properties that are less relevant to the EC performance. We explore and compare different machine learning methods, such as random decision forests, neural networks, and a nonlinear model of support vector regression on the gathered data set. The resulting predictive model will help to accelerate the exploration of new EC materials by enabling researchers to predict the EC temperature change from properties available in the literature, without having to synthesize and measure the samples themselves. Moreover, this model will help to build the growing area of physics-informed machine learning, where domain-specific knowledge is leveraged with advanced algorithms.


Physics-Informed Data-Driven Approach for Optimizing Electrocaloric Cooling

Division / Sponsors: 
Aerospace, Bioengineering, Applied Mechanics, Design Engineering Division, Advanced Energy Systems, Manufacturing Engineering Division, Noise Control and Acoustics, Nondestructive Evaluation, Diagnosis, and Prognosis (NDPD), Nuclear Engineering Division, Process Industries Division, Fluids Engineering, Heat Transfer, Materials, Microelectromechanical Systems, Safety Engineering and Risk Analysis, Computers & Information in Engineering


Paper Proceedings: https://asmedigitalcollection.asme.org/IMECE/IMECE2020/volume/84478

IMECE2020 - Virtual Conference

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.