Experimental studies on the adaptive multi-mode vibration control of composite beams have been performed using piezoelectric actuator and neuro-controller. Generally, neuro-controllers need too much computational time, which blocks wide real-time applications of neuro-controllers. Therefore, in this paper an adaptive notch filter is proposed to separate vibration signals into modal vibration signals. Two neuro-controllers with fewer weights are connected to the corresponding modal signals so as to generate proper modal control forces. The performance of the proposed notch filter has been proven via simulations and experiments on with two composite beams that have different dynamic characteristics. The hardware for the adaptive vibration control experiment was prepared using DSP (digital signal processor) as digital controller. The vibration controls using adaptive notch filter and neuro-controllers have been performed with two specimens, model A and model B, which have variations in natural frequencies because of positions of tip masses. Significant vibration reduction has been observed in both cases. The vibration control results show that the present neuro-controller has good robustness with the system parameter variations.