This thesis presents a theoretical and experimental study of self-tuning vibration control. Feedback design is often based on the assumption of time-invariance, which means that the controller has constant coefficients. Self-tuning control takes into account process changes in the response of the system under control by incorporating an adjusting mechanism which monitors the system, compares its status with the required one and adjusts the coefficients of the controller. In this thesis a self-tuning process is analysed for active and semi-active control of broadband vibration based on the maximisation of the power absorbed by the controller. The absorbed power can be locally estimated without using extra sensors to monitor the global response of the system under control. This is particularly advantageous in applications where many actuators are required, in which case each actuator and the collocated sensor can be treated as an independent self-tuneable unit. A theoretical analysis of vibration control using this approach is presented for lumped parameter systems and also for distributed systems, such as beam and panels. Different tuning strategies are compared in terms of the reduction of the global response of the system under control. An algorithm is then discussed that tunes the feedback gains of independent control units to maximise their individual absorbed powers. Experimental studies are then presented of a selftuning control system with two decentralised control units using velocity error signals and electromagnetic reactive actuators installed on an aluminium panel. In the second part of the thesis the analysis is extended to the use of inertial actuators. In this case the implementation of the self-tuning control based on the maximisation of the power absorbed is investigated using simulations of velocity feedback control and shunted inertial actuators.