SIMOAI

Research area 02

PCM + AI

Data-guided phase-change thermal systems.

Phase change materials store and release latent heat during melting and solidification. We are interested in combining experiments, phase-change CFD, and machine learning to study material selection, heat-exchanger geometry, and intelligent charging and discharging.

Conceptual visualization of a solid-to-liquid phase boundary inside a thermal-storage cell
Conceptual visualization · not simulation output

Data landscape

Material, phase, and system data.

PCM performance depends on thermophysical properties, geometry, operating conditions, and cycle history. AI models must preserve energy balance and remain reliable beyond their training materials and configurations.

Materials

Thermophysical properties

Transition temperature, latent heat, conductivity, density, heat capacity, supercooling, hysteresis, and cycle stability.

Phase state

Evolution through time

Temperature fields, liquid fraction, phase-front position, melt/solidification time, and natural convection.

System

Thermal performance

Stored and released heat, power density, peak temperature, uniformity, pressure drop, and pumping power.

Validation

Reliable generalization

Energy residual, phase-front agreement, held-out material families, unseen geometry, uncertainty, and cyclic performance.

Key descriptors

From material properties to storage value.

Tpc
Phase-change temperatureThe transition range must fit the intended thermal duty.
L
Latent heatEnergy stored per unit mass during phase change.
k
Thermal conductivityOften limits charging and discharging rate.
fl
Liquid fractionRanges from 0 for solid to 1 for liquid.
tc/d
Charge/discharge timeMeasures how quickly useful storage can respond.
Ebal
Energy balanceChecks consistency between heat input, storage, losses, and output.

Potential directions

Where AI can contribute.

  • Traceable property dataConnect experiments, literature, and simulation with provenance and uncertainty.
  • Melt-front surrogatesPredict temperature, phase fraction, and charging time while enforcing energy consistency.
  • Joint optimizationBalance PCM properties, fins, foams, geometry, pressure drop, and manufacturability.
  • Predictive controlCoordinate charge and discharge using thermal state, demand, and weather forecasts.

Published benchmarks · not SimoAI results

Examples from the literature.

19% / 88%

An embedded-PCM electronic-package study reported a 19% reduction in maximum temperature rise and up to 88% lower temperature fluctuations versus its studied no-PCM design.

Applied Thermal Engineering
Ternary salt

A study trained a neural-network interatomic potential for NaCl–MgCl₂–CaCl₂ and used deep-potential molecular dynamics to predict structure and thermophysical properties against experiments.

ACS Applied Materials & Interfaces

Selected literature

Research context.

  1. The contribution of artificial intelligence to phase change materials in thermal energy storageRenewable Energy, 2025
  2. Machine-learning assisted optimization strategies for phase change materials embedded within electronic packagesApplied Thermal Engineering, 2021
  3. Machine learning techniques to probe the properties of molten salt phase change materialsCell Reports Physical Science, 2024
  4. Development of NaCl–MgCl₂–CaCl₂ ternary salt for high-temperature thermal energy storage using machine learningACS Applied Materials & Interfaces, 2024