Low-Power Low-Error Wireless Sensor Network for Health and Usage Monitoring Systems

Abstract

Condition-based maintenance (CBM) is an information-based just-in-time maintenance approach that is designed to replace traditional periodically-scheduled inspections to reduce cost without sacrificing safety/reliability. CBM relies on data collected under actual operating conditions by a network of sensors in a Health and Usage Monitoring System (HUMS). HUMS is used in a variety of applications, including aircrafts, built-structure monitoring, or systems whose maintenance cost is high, or their failure can have life-threatening consequences. The focus of this research is on a HUMS to wirelessly monitor the condition of laminated bearings linking rotor blades to the rotor axis in a helicopter. A minimum of two HUMS nodes per blade is required to collect inertial data synchronously from different locations with timing error <30 µs. This is needed to accurately extract blades’ angles and predict the bearings’ health. Wireless connectivity poses challenges in terms of low-error time synchronization and extended operational lifetime. Battery-powered and BLE (Bluetooth Low Energy)-enabled sensor nodes are custom-designed for the above application. Low-error and low-power synchronization protocols are proposed and tested on the developed nodes, including BlueSync, Discrete Adjustments, and AdaptSync. BlueSync is compatible with the BLE standard and reduces synchronization error to <1 µs per 60 s of measurement without requiring any wireless resynchronization between transmitter and receiver. This is the lowest reported synchronization error for BLE, which is achieved by placing the timestamping step at the end of packet transmissions and reducing the uncertainty in transmitter timestamping. Discrete Adjustments increases a node’s operational lifetime by reducing the overhead of synchronization calculations, which is the time needed for computing frequency adjustments and is indicative of energy efficiency of calculations. Calculations and adjustments are applied only when needed. Up to 15x reduction can be achieved in calculation times. AdpatSync enables both low-power consumption and low synchronization error for long sessions that could run for as long as one hour by using previous timing information as training data to estimate the error and apply the appropriate correction on each node, thus eliminating the need for resynchronizations. It. In a 10-minute recording session, AdaptSync reduces the error by 7x compared with standard synchronization methods. To further increase the node lifetime, a 915 MHz, -61 dBm, 2.8 nW wake-up radio (WUR) is designed in TSMC 65 nm CMOS technology, to listen for wake-up commands when the node is in the sleep state. With a passive front-end and only two <1.5 mm2 off-chip components for matching network, it achieves an estimated range of ~150 m, which is enough for many wireless sensor applications. It has a two-stage wake-up architecture that reduces the probability of false wake-ups, and consequently the power consumption of the node. A complete system is developed for in-flight tests and has been evaluated on a rotor test bench. Nodes with their required IP67 housing (4×4×2 cm3) weigh 29 g and provide a lifetime of ~110 days with 1~2 hours recording per week. Collected accelerometers and gyroscopes data are used to extract the blades’ angles. Results show angles errors of < 1°, compared with measurements obtained from an optical reference, which are well within the limit for the fault-detection algorithms of this application, and validate that the developed wireless system can be reliably used to monitor the laminated bearings. Integrating the WUR with the nodes could increase their lifetime to ~6 months.