The kernel density estimation (KDE) method is used to solve the damage detection problem based on the spectral element method (SEM), and the probability density functions (PDFs) of damage locations can be obtained. A three-dimensional SEM including electromechanical coupling is developed to simulate the propagation of Lamb waves in both healthy and damaged aluminum plates. The time-of-flight (TOF) of response signals in each actuator-sensor path is calculated by a continuous wavelet transform (CWT) to measure the speeds of Lamb waves. Rayleigh-Lamb equations are analyzed by a numerical calculation method, and the theoretical speeds of Lamb waves in aluminum plates can be obtained, which are compared with the ones obtained by SEM to demonstrate SEM′s accuracy. On the basis of elliptical position technology, the KDE is introduced as a probabilistic approach for damage detection in conditions of environmental uncertainty. The PDFs of damage locations are discussed under three levels of noise that are added into the response signals, and the final probability distributions are given. The results demonstrate that the SEM-based approach with the application of KDE is capable of efficiently identifying damage locations, with a maximum error of about 5%.