This thesis investigates two fundamental technological challenges that prevent water
utilities from deploying infrastructure monitoring apparatus with high spatial and temporal resolution: providing sufficient power for sensor nodes by increasing the power
output from a vibration-driven energy harvester based on piezoelectric transduction,
and the processing and storage of large volumes of data resulting from the increased
level of pressure and flow rate monitoring.
Piezoelectric energy harvesting from flow-induced vibrations within a water main
represents a potential source of power to supply a sensor node capable of taking high-
frequency measurements. A main factor limiting the amount of power from a piezoelectric device is the damping force that can be achieved. Electronic interface circuits
can modify this damping in order to increase the power output to a reasonable level. A unified analytical framework was developed to compare circuits able to do this in terms
of their power output. A new circuit is presented that out-performs existing circuits by
a factor of 2, which is verified experimentally.
The second problem concerns the management of large data sets arising from resolving challenges with the provision of power to sensor devices. The ability to process large
data volumes is limited by the throughput of storage devices. For scientists to execute
queries in a timely manner, query execution must be performant. The large volume of
data that must be gathered to extract information from historic trends mandates a scalable approach. A scalable, durable storage and query execution framework is presented
that is able to significantly improve the execution time of user-defined queries.
A prototype database was implemented and validated on a cluster of commodity servers using live data gathered from a London pumping station and transmission
mains. Benchmark results and reliability tests are included that demonstrate a significant improvement in performance over a traditional database architecture for a range of
frequently-used operations, with many queries returning results near-instantaneously.