This paper describes a damage identification method for structures mainly subject to bending load, considering the applicability in a decentralized computing environment of wireless sensor networks. A modal flexibility-based damage indices that have simple and intuitive physical interpretation is presented. Since the proposed damage indices are computed using angular mode shapes of the lowest several modes, MEMS gyroscopes are adopted as the sensing device. The damage evaluation algorithm is then modified to a decentralized form, which is to be implemented as the local computation of mode separation at each sensor unit and the global computation of the damage index at the central monitoring station. Experiments using a bolted beam are conducted to show the applicability of the proposed algorithm to the detection of loose bolt failures.
In this paper, we design a sensor network system with a host computer and a sensor terminal, which has local data processing by using off-the-shelf hardware. If all data, which are obtained by many sensors, are transmitted to the host computer and data processing is centralized at the host computer, the task of the host computer becomes large. Therefore, we consider that tasks of the host computer are decreased by making each sensor unit to share data processing locally and transmit results to the host computer.
The sensor terminal consists of an accelerometer, an analog-to-digital (A/D) converter and the T-Engine (M32104) has capacity of data acquisition, processing and transmission in real time. We use the wavelet transform as an algorithm of data processing. The host computer can receive the result of the wavelet transform by sending the request to the sensor terminal. However, The host computer and the sensor terminal are connected by LAN cables. Next version of the prototype has wireless LAN.
We demonstrate the performance of the sensor network system by applying this system to a structural model and make the sensor terminal execute locally the wavelet transform.