Thermal characterization of honeycomb core sandwich structures
Honeycomb core sandwich structures are an integral part of many of today's aerospace structures. When subjected to high-speed flight, thermal loading can induce significant stresses. The need for thermal properties to perform thermal stress analyses in these structures is the motivation behind this research. The thermal property estimation approach used here involves the minimization of a least-squares function containing both measured and calculated values. In addition, an applied heat flux is necessary at one boundary for the simultaneous estimation of thermal properties. The specific objectives are to develop a thermal model to describe honeycomb core sandwich structures, optimize experimental designs for use in parameter estimation, develop a finite element-based parameter estimation algorithm, and estimate the pertinent thermal properties of the structure.
A combined conductive/radiative heat transfer model was used for the analysis of the structure. Due to the composition of the structure, it was determined that a one-dimensional model would be sufficient. This model was used in both parameter estimation and experimental design.
Experimental design involves finding input variables for an experiment such that the response of the system contains the highest possible amount of information on the parameters of interest which characterize the response. In this study, the design was performed by using a combination of two methods. The first involved maximizing the temperature derivatives with respect to unknown thermal properties. The second involved a scaled confidence interval approach. The experimental parameters optimized were heating time and total experiment time.
A finite element program was used to perform transient temperature calculations because of the flexibility it has to analyze complex structures. Parameters estimated in this study exhibited a great deal of correlation, or interaction. This showed the need for a constrained parameter estimation algorithm. A penalty function method was developed for this purpose.
The last part of this study involved the actual estimation of thermal properties. An experimental apparatus was designed and built to record the transient temperature response of the test sample. A four-sheet SPF/DB sandwich was used as the test sample. Thermal properties were estimated using four combinations of sensors and boundary conditions.
It was found that in one case parameters could be simultaneously estimated despite the presence of correlation. These estimated parameters were shown to produce reasonably small errors when used in transient temperature calculations. It was also shown that large temperature gradients produce estimates with smaller confidence intervals. The importance of maintaining accurately known boundary conditions was also demonstrated.