In a wide range of industries, machine condition monitoring is one of the most cost effective ways to minimise maintenance efforts and machine downtime. To implement such a system, wireless solutions have increasingly become an attractive proposition due to the ease of installation and minimal infrastructure alternation. However, currently most wireless sensors in the world are powered by a finite battery source. The dependence of batteries not only requires frequent maintenance, but also has adverse environmental consequences associated with battery disposal. These reasons render massive deployment of wireless sensors in the industry problematic. With the advances in semiconductors, power consumption of wireless sensors has been continuously decreasing. It is an inevitable trend for self-powered wireless sensors to emerge and become the norm for machine and environmental monitoring. In this research, vibration is chosen to be the energy source to enable self-powered wireless sensors due to its ubiquitousness in machinery and industrial environments. As a result of relying on resonance, the biggest challenge for vibration-based energy harvesters is their narrow bandwidth. Even a small deviation of the vibration frequency can dramatically reduce the power output. The primary goal of this research is to address this problem. In particular, Piezoelectric generators are identified to be the most suitable technology. In this work, extensive theoretical and experimental studies are conducted in single mass and multi-modal harvesters, and in resonance tuning harvesters by modulus and impedance matching as well as by mechanical actuation. Mathematical modelling plays a significant role in energy harvester designs. A dynamic model that generalises the single degree of freedom models and the continuum models is derived and validated by experiments. The model serves as the building block for the whole research, and it is further refined for the investigation of modulus and impedance matching. In the study of multi-modal harvesters, a continuum model for double-mass piezoelectric cantilever beams is derived and experimentally validated. To study the feasibility of resonance tuning by mechanical means, prototypes were built and performance evaluated. This document details the theoretical basis, concepts and experimental results that extend the current knowledge in the field of energy harvesting. This research work, being highly industrially focused, is believed to be a very significant step forward to a commercial energy harvester that works for a wide range of vibration frequencies.