Spectral element method implementation on GPU for Lamb wave simulation

Abstract

Parallel implementation of the time domain spectral element method on GPU (Graphics Processing Unit) is presented. The proposed spectral element method implementation is based on sparse matrix storage of local shape function derivatives calculated at Gauss–Lobatto–Legendre points. The algorithm utilizes two basic operations: multiplication of sparse matrix by vector and element-by-element vectors multiplication. Parallel processing is performed on the degree of freedom level. The assembly of resultant force is done by the aid of a mesh coloring algorithm. The implementation enables considerable computation speedup as well as a simulation of complex structural health monitoring systems based on anomalies of propagating Lamb waves. Hence, the complexity of various models can be tested and compared in order to be as close to reality as possible by using modern computers. A comparative example of a composite laminate modeling by using homogenization of material properties in one layer of 3D brick spectral elements with composite in which each ply is simulated by separate layer of 3D brick spectral elements is described. Consequences of application of each technique are explained. Further analysis is performed for composite laminate with delamination. In each case piezoelectric transducer as well as glue layer between actuator and host structure is modeled.