Bayesian Surrogates for Integrating Numerical , Analytical and Wearable Computers Experimental Data : Application to Inverse Heat Transfer In

Wearable computers are portable electronics worn on the (CMU). Figure 1 shows the wearable computer's evolution in body. The increasing thermal challenges facing these compact terms of the power dissipation per unit area available for heat systems have motivated new cooling strategies such as transfer to the ambient. The exponential increase in the power transient thermal management with thermal storage materials. density together with the complex interactions of a concurrent The ability of building models to quickly assess the effect of design process requires innovative design methodologies as different design parameters is critical for effectively well as improved thermal strategies. incorporating these innovative thermal strategies into new products. System models that enable design space exploration are built from different information sources such as numerical simulations, physical experiments, analytical solutions and heuristics. These models, called surrogates, are nonlinear and adaptive in nature and thus suitable for system responses where limited information is available and few realizations (experiments or numerical simulations) are feasible. In this paper, the surrogate framework is applied to estimate values for unknown physical parameters of an embedded electronics system. For this purpose, physical experiments and numerical simulations are performed on an embedded electronics prototype system of the TIA (Technical Information Assistant) wearable computer. Numerical models are studied which involve five and three unknown Parameten, Wearable computers, and more generally portable with and without thermal contact resistances, respectively. electronics, are good of how the rising product Through the use of orthogonal m a y s and optimal sampling9 an complexity and the restrictions on time to market introduce a efficient exPlomtion of the Parameter space is performd. The need for shorter design cycles. Simple models and previous objective is to determine system Parmetes such as thermal experience are commonly used to perform characterization and conductivities, thermal contact resistances and heat transfer assessment of different design alternatives in the early stages of coefficientsSurrogate are that combine a concurrent design process. The possibilities of using more information obtained from numerical simulations and detailed such as simulations or physical experimental model measurements as well as from a thermal prototypes during these early design stages remain prohibitive resistance network simplified model. The integration Of Several due to the large amount of design alternatives that to be information SourceS duces the number of numerical tested. This lack of reliable decision tools -during conceptual simulations needed to find reliable estimates of the system design __ limits many innovative approaches. paradoxically, parameters and allows for identification of the best numerical these innovative approaches are necessary to attain the sought model. For the embedded electronics case, the use of prior perfomancegoals, information f iom the thermal resistance network model reduces imposes Significantly the COmpUtatiOnal effort required to investigate the pressures on the development of innovative thermal solution space. management strategies, we take a close look at the design requirements imposed on wearable computers. First, we restrict INTRODUCTION our attention to passive thermal strategies due to battery power Wearable computers are Portable ad compact e l e c @ d c limitations. Second, we require a sealed computer housing due systems that merge information space with user workspace [l]. to the harsh operation ~ i ~ ~ l l , , , we to t2] and are designed to be ensure a reliable operation of the electronics and a safe lightweight, mgged and power-efficient. Several generations of operation for the user. The latter is attained by keeping the Wearable computers, of increasing comP1exitY~ have been surface of the wearable computer at temperatures below those designed and manufactured at Carnegie Mellon University by ergonomic requirements. The simultaneous Figure 1: Wearable computer evolution an example of how rising producr