(646d) Health Monitoring of Multi-Scale Systems Using an Optimal Multi-Rate Wireless Sensor Network

Systems with multiple time scales where the fastest time scale can evolve in seconds or minutes whereas the slowest time scale can evolve over weeks or months are ubiquitous. Health monitoring of such systems is challenging since the traditional estimator-based approaches become computationally intractable. On the other hand, sensors that are embedded within the structures or are integral part of the equipment items are being increasingly investigated for health monitoring of critical equipment items. These sensors are often wireless. Measurements from wireless sensor networks can be noisy and suffer from communication constraints, packet dropouts, and synchronization errors. Due to the limitations in the communication bandwidth, only limited packets can be transmitted at any instant resulting in a multi-rate sensor network. In addition, random packet dropouts caused by unavoidable transmission error are unavoidable in wireless sensor networks. Furthermore, measurements can be out of sequence. There is hardly any work in the existing literature on process monitoring by using multi-rate measurements from a wireless sensor network that suffers from packet dropouts and synchronization errors.

In this work, an optimal sensor network is synthesized for a multi-scale system with widely varying time-scales considering that the measurements are available from a multi-rate wireless sensor network. It should be noted that in the existing literature, optimal sensor network synthesis for multi-rate systems has been mainly reported for linear systems.¹ In this work, algorithms are developed for optimal sensor placement and health monitoring of a multi-scale, nonlinear, differential-algebraic-equation system. The algorithm is tested on a smart refractory brick with embedded sensors. It is observed that the developed algorithms result in superior estimate of relevant variables for health monitoring that evolve over widely varying time scales from minutes to hours.

References

(1) Kadu, S. C.; Bhushan, M.; Gudi, R. Optimal sensor network design for multirate systems. J Process Control, 2008, 18, 594-609.