Sandia Labs FY22 Laboratory Directed Research & Development Annual Report

ENVIRONMENTAL LEARNING METHODOLOGIES FOR REMOTE SENSING POWER OPTIMIZATION. Understanding the environment that an

throughout its required service duration. This research developed a fusion of unsupervised learning, parametric optimization, and anomaly detection algorithms to design a robust, adaptable framework for in-situ power optimization to preserve and extend service life of remote sensor systems in the field. This technology affords a reduction in deployment planning time and deployment risk, affording quicker response to opportunities. (PI: Aaron Hill)

unattended sensor is being placed in is critical to developing accurate models that drive design, especially when the sensor’s power use is directly coupled with the environment’s stimuli. Knowledge of the deployment environment is used to design and configure a sensor to meet a specific service life requirement. Sensors deployed in environments that are understudied or dynamic pose a problem to the design process of guaranteeing that a sensor will remain functional

Diagram of environmental learning model. The clustering algorithm produces a model in the feature space of the environment that informs the optimizer. If the environment has changed, the anomaly detection algorithm triggers the optimizers. The optimizer makes dynamic adjustments to system parameters that directly affect the energy utilization of the system to better match the present environment.

BENEFITS IN A COMPACT PACKAGE.

in heterogeneous integration to reduce high voltage power conversion to a single packaged component: a radiation-hard, high-voltage, chip scale power converter. This effort exploits Sandia’s microelectronics development capabilities to realize a compact converter with high-boost gain, reasonable conversion efficiency, good voltage stability, and resilience in extreme environments. To date, the team has demonstrated a feasible

conversion circuit at the printed circuit board level; designed, fabricated, and demonstrated a custom integrated circuit (IC) to control switching using one of Sandia’s resilient IC platforms; built and begun testing a chip-scale power inductor; and fabricated an interposer with novel embedded high-voltage components. Demonstration of the chip-scale power converter is expected in FY23. (PI: Jason Neely)

49

LABORATORY DIRECTED RESEARCH & DEVELOPMENT