Experimental verification of optimal experimental designs for the estimation of thermal properties of composite materials
The need to simultaneously estimate thermal properties stems from the desire to analyze complex structures which do not have the flexibility to be experimentally tested in multiple configurations. In order to produce reliable and accurate thermal property estimates, the experiments must be carefully developed. A carefully designed experiment maximizes the sensitivity of the temperature distribution with respect to the unknown thermal properties, as well as providing minimum correlation between the estimated properties.
Two objectives were set forth in this research. First to apply existing predicted optimal experimental designs developed by Moncman (1994) to simultaneously estimate the two-dimensional thermal properties of the carbon-fiber/epoxy-matrix composite, AS4/3502. Due to the anisotropic nature of the composite, the effective thermal conductivities through the thickness and in the plane of the composite needed to be estimated along with the volumetric heat capacity. After simultaneously estimating the properties, the second objective was to verify that the predicted optimal designs provided the most accurate estimates.
In accomplishing both objectives, the research plan developed in three distinct stages. As a starting point, the one-dimensional analysis was performed to gain confidence in the experimental setup and procedure. Due to the successful estimation of the one-dimensional properties, the experiments were expanded into a two-dimensional analysis. This analysis attempted to simultaneously estimate all three thermal properties from one optimal, transient, temperature measurement. But due to correlation problems invoked by experimental errors, it was unsuccessful. Therefore, the research focused on the estimation of the in-plane thermal conductivity and the volumetric heat capacity. After successfully estimating the properties, the optimal designs were verified through additional testing with perturbations applied to the optimal settings.
Complete success was not accomplished in this study due to partially satisfying the first objective. All three thermal properties were estimated for the anisotropic composite but due to near correlation between the thermal conductivities, they could not be determined from a single, optimal experiment. Therefore in an attempt to uncorrelate the thermal properties it is recommended that the experiments be performed with multiple sensors. In addition, alternative boundary conditions should be considered on their ability to provide more sensitive information on the thermal properties and practicality to experimentally maintain.
The second objective, verifying the optimal designs, was completely successful in demonstrating that the optimal parameter settings did produce the most accurate thermal property estimates. Therefore, an optimally designed experiment using multiple sensors should allow for the accurate and simultaneous estimation of the thermal properties for complex structures.