Predictive Thermal Digital Twin Model for a Renewably Powered Offshore AUV Servicing Platform

Abstract Autonomous Underwater Vehicle (AUV) missions are limited in range and duration by the vehicle’s battery capacity. The Platform for Expanding AUV exploRation to Longer ranges (PEARL), described in this paper, aims to extend the range and endurance of AUVs while reducing data latency and operating costs by providing AUV battery recharging and data uplink. PEARL’s autonomous requirements and operating environment present challenges. Its high-bandwidth communication system consumes significant power, leading to localized heating and electromagnetic interference that may affect functionality. Self-prognosis of health conditions is highly desirable to ensure the reliability of PEARL. This paper focuses on developing a thermal digital twin model to monitor thermal conditions during field tests of PEARL on the Charles River, Massachusetts. To build this model, we acquired temperature data from thermistors attached to various components of PEARL, which we model via corresponding nodes in our thermal network model. Temperature measurements were made during mission operations from July to November 2024, ranging from 10 °C to 57 °C, which were used to calibrate the thermal digital twin. Temperature profiles of the nodes could be estimated within 1% RMSE in most cases with Simulink thermal models that had 80 to 90 signal lines in terms of complexity. The tradeoff between digital twin complexity and estimation performance was prominent throughout modeling iterations. The results of the model will be used to inform the design of the next-generation PEARL in an ocean setting.