Compressed Sensing in Vibration Monitoring Wireless Sensor Network

Abstract

Over 70% of the bridges in Costa Rica have critical elements in the structure. Wireless sensor networks have been used for structural monitoring because of their short installation time and low economic cost due to the lack of wiring. 

In this research is presented a wireless sensor network which uses a compressed sensing algorithm for vibration monitoring on bridges. The network design is proposed expecting data integrity and energy harvesting. 

The algorithm performs downsampling by coder-decoder pairs. The selected pair is conditional encoder and predictive decoder because this combination has elements in common that can shared to get better estimates in few steps. 

Two coder-decoder pairs variants are presented: variable-fixed and fixed-variable. The first one proposes a constant factor compression during each sampling period while the second presents variable compression that depends of the signal behavior over time. 

After an experimental test using Waspmotes the fixed-variable variant has a 56.58% reduction of power consumption by introducing a maximum error ± 0.00195g and compress in 52.44% the amount of samples. This algorithm increased the network energy autonomy from 17 hours to 26.5 hours. Through mathematical analysis, the variable-fixed technique reduces in 74.81% the power consumption in sensing nodes transmissions and decrease in 90% the number of samples.