Neural net-based robust controller design for brushless DC motor drives

In this paper, a nonlinear neuro-controller is developed for controlling the speed of brushless dc motors operating in a high-performance drives environment. The control inputs and the identification parameters of the system are adjusted simultaneously in real time using a system composed of three-hidden-layer dynamic neural networks while the system is in operation. The control architecture adapts and generalizes its learning to a wide range of operating conditions and provides the necessary abstraction when measurements are contaminated with noise. The problem of persistently spanning excitation faced with the use of an on-line neuro-controller is addressed. In particular, the ability of the neuro-controller to "remember" previously-trained reference tracks when confronted with an input excitation that is markedly different from what it was trained with is investigated. The intent is to capture the nonlinear dynamics of a brushless de motor over any arbitrary time interval in its range of operation. The sensitivity of real time neuro-controllers to random changes in the load torque also is investigated and very promising results are observed.