Speaker
Description
Smartphone sensing technology offers a new next-generation system for structural health monitoring (SHM) of civil structures. Among smartphone built-in sensors, the tri-axial MEMS accelerometer is an appropriate and affordable choice for vibration monitoring with successful applications to real-world large-scale structures such as bridges and high-rise buildings. However, the implementation of a long-term SHM program via smartphone sensing technique is a big challenge. One of the critical issues is the impacts of environmental changes, particularly ambient temperature, on both the civil structure being monitored and the smartphone MEMS sensors. Under such circumstances, it is challenging that variability in measured accelerations is caused by structural changes or sensor performance. To investigate the impacts of temperature and humidity on smartphone MEMS sensors and their applications to long-term SHM programs, this research implements an experimental study and assess extreme environmental conditions of such sensors. In this investigation, a smartphone is placed and fixed in different areas with various temperatures varying -25⸰C to 40⸰C and humidity varying 28% to 96%. To indicate the capabilities of machine learning for dealing with the challenges regarding MEMS sensors under varied environmental conditions, three unsupervised outlier detectors in terms of discriminative models are introduced and compared to identify conditions that the smartphone MEMS sensors are significantly influenced by extreme environmental changes. Results demonstrate that the extremely cold and warm temperatures critically impact the performance of the smartphone MEMS sensor.
