When a wheat kernel is infected by insects, nutrients of the kernel are damaged, which reduces the commercial value of wheat and threatens its food safety. We suggest a new non-destructive method for revealing infected kernels, which is based on laser-induced ultrasonic. A detection process is simulated using a finite-element analysis. Firstly, we formulate a finite-element model for normal wheat and models for infected wheat (with the wormhole radii 0.4, 0.6, 0.8 and 1.0 mm), following from a classical thermal-stress theory. Secondly, a controlling equation is obtained, a process of ultrasonic excitation is simulated for all the models using a standard Newmark time-integration method, and the ultrasonic signals are obtained for those models. Finally, we derive the velocity of ultrasonic wave propagating in the kernel and analyze the relationships among the ultrasonic parameters and the structural parameters of the models. Our results show that the time interval between two adjacent ultrasonic echoes depends notably upon the internal kernel structure. In particular, the interval between the first and second echoes is longer and the interval between the second and third echoes is shorter with increasing wormhole radius. This can be used for finding out the infected kernels at early stages and determining the sizes of wormholes.