Advancements in digital technologies have allowed for the development of increasingly complex active noise and vibration control solutions, with a wide range of applications. Such control systems are commonly designed using linear digital filters which cannot fully capture the dynamics of nonlinear systems. To overcome this, it has previously been shown that replacing linear controllers with Neural Networks (NNs)can improve control performance in the presence of nonlinearities in both the system plant and primary path. Inferring the outputs of NN controllers can, however, be computationally expensive, limiting the practicality and real-time performance of such control systems. It has been demonstrated previously that, with nonlinearity in the primary path, the range of nonlinearity over which a single NN controller can achieve significant levels of control is limited by the size of the NN. In this paper, a method of dynamically switching between multiple smaller NN controllers is presented. In a simple time-discrete simulation, the performance and computational cost of this approach is compared to that of individual fixed NN controllers, demonstrating improved computational efficiency and performance over a range of nonlinearity as the system disturbance changes over time.

This content is only available via PDF.