Experimental studies on vibration control of a composite beam with a piezoelectric actuator and an extrinsic Fabry-Perot interferometer (EFPI) have been performed using a neural network controller. Because of their interferometric characteristics, EFPI sensors show non-linearity as dynamic amplitude increases. Within the linear range of EFPI, conventional control algorithms can be applied without serious difficulty. However, closed-loop control may make the target system unstable when sensor non-linearity gets high. Therefore, we examine the effects of the non-linearity of the sensor on the control stability and performance, and investigate any simple method applicable to the vibrations beyond the linear range. For this purpose, a neural controller is adopted and its performance is experimentally investigated. The neuro-controller showed good performance and adaptiveness to the sensor's non-linearity. Although the present neuro-controller is not a fundamental solution to vibration control of structural systems, it can be a simple practical choice for systems with sensor non-linearity.