Wireless sensor networks (WSNs) have the potential to transform engineering infrastructure, manufacturing, and building controls by allowing condition monitoring, asset tracking, demand response, and other intelligent feedback systems. A wireless sensor node consists of a power supply, sensor(s), power conditioning circuitry, radio transmitter and/or receiver, and a micro controller. Such sensor nodes are used for collecting and communicating data regarding the state of a machine, system, or process. The increasing demand for better ways to power wireless devices and increase operation time on a single battery charge drives an interest in energy harvesting research.
Today, wireless sensor nodes are typically powered by a standard single-charge battery, which becomes depleted within a relatively short timeframe depending on the application. This introduces tremendous labor costs associated with battery replacement, especially when there are thousands of nodes in a network, the nodes are remotely located, or widely-distributed. Piezoelectric vibration energy harvesting presents a potential solution to the problems associated with too-short battery life and high maintenance requirements, especially in industrial environments where vibrations are ubiquitous.
Energy harvester designs typically use the harvester to trickle charge a rechargeable energy storage device rather than directly powering the electronics with the harvested energy. This allows a buffer between the energy harvester supply and the load where energy can be stored in a "tank". Therefore, the harvester does not need to produce the full required power at every instant to successfully power the node. In general, there are tens of microwatts of power available to be harvested from ambient vibrations using micro scale devices and tens of milliwatts available from ambient vibrations using meso scale devices. Given that the power requirements of wireless sensor nodes range from several microwatts to about one hundred milliwatts and are falling steadily as improvements are made, it is feasible to use energy harvesting to power WSNs.
This research begins by presenting the results of a thorough survey of ambient vibrations in the machine room of a large campus building, which found that ambient vibrations are low frequency, low amplitude, time varying, and multi-frequency. The modeling and design of fixed-frequency micro scale energy harvesters are then presented. The model is able to take into account rotational inertia of the harvester's proof mass and it accepts arbitrary measured acceleration input, calculating the energy harvester's voltage as an output.
The fabrication of the micro electromechanical system (MEMS) energy harvesters is discussed and results of the devices harvesting energy from ambient vibrations are presented. The harvesters had resonance frequencies ranging from 31 -- 232 Hz, which was the lowest reported in literature for a MEMS device, and produced 24 pW/g^2 -- 10 nW/g^2 of harvested power from ambient vibrations. A novel method for frequency modification of the released harvester devices using a dispenser printed mass is then presented, demonstrating a frequency shift of 20 Hz.
Optimization of the MEMS energy harvester connected to a resistive load is then presented, finding that the harvested power output can be increased to several microwatts with the optimized design as long as the driving frequency matches the harvester's resonance frequency. A framework is then presented to allow a similar optimization to be conducted with the harvester connected to a synchronously switched pre-bias circuit.
With the realization that the optimized energy harvester only produces usable amounts of power if the resonance frequency and driving frequency match, which is an unrealistic situation in the case of ambient vibrations which change over time and are not always known a priori, an adaptable-frequency energy harvester was designed. The adaptable-frequency harvester works by taking advantage of the coupling between a sliding mass and a beam. The derivation of the nonlinear coupled dynamic mathematical model representing the physical system is presented, as are the numerical and experimental results of the prototype device. Passive self-tuning was observed in this system and the mathematical model was found to successfully portray the physical behavior.
