Off-the-shelf modal analysis: Structural health monitoring with Motes

Breakthrough strides in micro-electro-mechanical systems (MEMS) demand a paradigm shift in traditional data acquisition and signal processing methodologies used for structural health monitoring (SHM). One such device which embodies MEMS technology is the Mote. Motes integrate a microprocessor, memory, and a radio transmitter together and can be outfitted with a plethora of industry standard sensory devices with little or no modification. The implication of this conglomerated setup is that it can be used to reduce, store, and ship data at the acquisition site. Moreover, the commercial development of these devices has reduced production costs and increased product development. As such, the intrinsic versatility of a Mote system can be affordably harnessed and dense sensor arrays can be utilized to enhance real-time modal parameter identification, the backbone of global vibration based SHM techniques. However there are limitations which need to be addressed. For example, significant data loss is attributed to the low bandwidth of the low power radio transmitter employed on Mica2 Motes. This issue of contention has been addressed by recent technological improvements in wireless transmission standards. Yet in spite of the progress made by new transmission protocols and standards towards reducing radio power consumption, increasing bandwidth, and reducing transmission latency, they alone will not meet all of the needs of a full-scale wireless SHM system. In light of this realization the efforts of this paper are shifted from previous studies where time histories were streamed from the data acquisition site to a central location for processing. Now instead the focus is on decentralizing the SHM system by exploiting the on-board computational faculties of a Mote. Here a simple signal processing and spectrum curve-fitting technique is used to automatically extract dominant frequency components of an acceleration time history record. Identified frequencies are then implemented in a probabilistic neural network (PNN) and correlation-based damage localization procedures in such a fashion to simulate on-board implementation of this SHM methodology on a Mote platform. This paper will document the efficacy of successful damage detection and localization experiment performed with acceleration impulse response time histories acquired from off-the-shelf Motes and sensor hardware.