Auto-commissioning of Acoustic Control of IM Drive using Bayesian Optimization

Recently proposed model predictive acoustic control of an AC drive relies on availability of the transfer function from the drive current to acoustic pressure. Since such model is typically not available, we propose to consider its parameters to be unknown and use methods of Bayesian optimization to find them. Specifically, we define a acoustic control performance criterion and optimize it in an additional outer loop. A microphone is necessary in this operation to provide feedback for tuning of the parameters used in model predictive drive control algorithm. However, the microphone is removed after the tuning. Since single evaluation of the performance criteria is time consuming, we choose the Bayesian Optimization that is able to find optimal tuning with very few evaluations.