A damage detection method for beam-type structures using ratio of natural frequency change and neural networks is proposed.With this method,ratio of natural frequency change is utilized as the damage index and neural networks are employed as the damage pattern discriminator.Unlike natural frequencies that are global dynamic properties of a structure and thus insensitive to damage location,the ratio of natural frequency change,as a function of sole variable of damage location,has the advantage of more sensitivity to damage location.The proof of the concept of the proposed method is verified by numerical simulations and its effectiveness is experimentally demonstrated on a cantilever beam with cracks.