Date of Graduation


Document Type


Degree Name

Doctor of Philosophy in Engineering (PhD)

Degree Level



Chemical Engineering


Jerry A. Havens

Committee Member

Mark E. Arnold

Second Committee Member

Greg Thoma


Liquefied natural gas is shipped across the oceans in large marine carriers. The carriers house the LNG using several different insulation systems. One of these systems involves large aluminum spheres insulated with polystyrene foam. Polystyrene foam is very susceptible to heat degradation. This raised issues as to the extent of possible insulation failure caused by a large ship fire. Experiments were done investigating the nature of polystyrene’s thermal degradation, notably by Brauman, Chen, and Matzinger and Butler. A large scale investigation was also performed by Sandia National Laboratory. However, computational modeling of the degradation was lacking. This work set out to build comprehensive models of the experiments performed by BCM, Butler, and Sandia. The models were created using COMSOL Multiphysics, a complex finite element method computational software. The thermal degradation of polystyrene is a complex process involving moving boundaries, phase transitions, and temperature dependent physical properties. The BCM and Butler experiments provided the most amount of detail and were used to test the accuracy of the models and methods used within them. The value of principal interest was the overall regression rate of the material, a variable that could be calculated by the models and compared directly with experimentally measured values. Modeling started with an extensive literature review of the physical properties of dense and foam polystyrene. The BCM experiments were first considered as they were the least complex. Both a transient and a steady-state model of the BCM experiments were built. Although detail of the transient heat up phase was lost, the steady-state model provided a better fit with experiment. With a good fit of the BCM experiments, the techniques were carried forward and adapted to the more complex Butler foam polystyrene experiments. Again, regression rates for the model were reasonably close to those measured by experiment. Next, the Sandia experiment was investigated. Here, the values of interest were no longer regression rates but insulation failure times and cargo tank heat fluxes. The Sandia experiment presented a new complication in the form of necessary information not being available, notably the surface emissivity of the aluminum scrim used. With the model, it was found that a very low surface emissivity, one near the lowest commercially available, was required to match the experimental results.