The dispersive properties of Lamb waves have been widely employed for structural health monitoring and nondestructive testing. When a part of the structural wall is thinned, the variation of the frequency thickness product in that part leads to the variance in velocity of each Lamb wave mode and hence the time of flight (TOF) of that wave packet. In this article, a defect identification method based on multimodal Lamb waves is derived by utilizing the TOF difference response of Lamb waves to thickness thinning. The TOF difference of multimodal Lamb waves is a function of the length and the depth of defect. The average length and depth of the defect can be predicted with the known material parameters. The average length and depth of a defect can be evaluated in once measurement by two sensors. Numerical simulation and experiments showed this method can identify the average length and depth of a defect.