The oil and gas industry faces increasing pressure to remove people from dangerous offshore environments. Robots present a cost-effective and safe method for inspection, repair, and maintenance of topside and marine offshore infrastructure. In this work, we introduce a new multi-sensing platform, the Limpet, which is designed to be low-cost and highly manufacturable, and thus can be deployed in huge collectives for monitoring offshore platforms. The Limpet can be considered an instrument, where in abstract terms, an instrument is a device that transforms a physical variable of interest (measurand) into a form that is suitable for recording (measurement). The Limpet is designed to be part of the ORCA (Offshore Robotics for Certification of Assets) Hub System, which consists of the offshore assets and all the robots (Underwater Autonomous Vehicles, drones, mobile legged robots etc.) interacting with them. The Limpet comprises the sensing aspect of the ORCA Hub System. We integrated the Limpet with Robot Operating System (ROS), which allows it to interact with other robots in the ORCA Hub System. In this work, we demonstrate how the Limpet can be used to achieve real-time condition monitoring for offshore structures, by combining remote sensing with signal-processing techniques. We show an example of this approach for monitoring offshore wind turbines, by designing an experimental setup to mimic a wind turbine using a stepper motor and custom-designed acrylic fan blades. We use the distance sensor, which is a Time-of-Flight sensor, to achieve the monitoring process. We use two different approaches for the condition monitoring process: offline and online classification. We tested the offline classification approach using two different communication techniques: serial and Wi-Fi. We performed the online classification approach using two different communication techniques: LoRa and optical. We train our classifier offline and transfer its parameters to the Limpet for online classification. We simulated and classified four different faults in the operation of wind turbines. We tailored a data processing procedure for the gathered data and trained the Limpet to distinguish among each of the functioning states. The results show successful classification using the online approach, where the processing and analysis of the data is done on-board by the microcontroller. By using online classification, we reduce the information density of our transmissions, which allows us to substitute short-range high-bandwidth communication systems with low-bandwidth long-range communication systems. This work shines light on how robots can perform on-board signal processing and analysis to gain multi-functional sensing capabilities, improve their communication requirements, and monitor the structural health of equipment.
Recent Comments