Efficient estimation of thermal conductivity distribution during curing of thermoset composites

To precisely simulate the curing process of composite materials, many material constants have to be determined experimentally, because the model based on the data sheet from suppliers usually differs from the actual behavior of the material. The data assimilation technique, which combines a simulation and experiments, is effective in reasonably setting the model parameters of the simulation, but it requires many parallel calculations. The present study develops an efficient data assimilation using an ensemble Kalman filter to estimate the thermal conductivity distribution of carbon fiber-reinforced plastics, which decreases calculation cost and maintains a sufficient estimation accuracy. We quantitatively evaluated the complexity of the thermal conductivity distribution using its variogram. It was found that there is correlation between the range of the variogram and the required number of ensemble members, which directly affects the calculation cost. The method was demonstrated using grid thermal conductivity distribution problems.