GPU Acceleration for Triaxial Induction Logging Tool Responses in Layered Uniaixal Formation

摘要

Recently, a new formulation for calculating triaxial induction logging tool responses in layered uniaxial (also called transverse isotropic) formation is developed. To obtain the tool responses in deviated wells, there are five independent Som-merfeld integrals need to be carried out, and the fast Hankel transform (FHT) is employed to do this with high efficiency. For horizontal wells, the integrals become divergent, which can be circumvented by using singularity extraction. For vertical wells, the FHT will fail, and the double-exponential (DE) rule is implemented to deal with this special situation. The computation of tool responses when the logging tool is drilling at a arbitrary dip angle in layered formation is further accelerated by exploiting the parallel programming on graphics processing unit (GPU). The responses of a large amount of logging points are calculated in parallel on GPU where one thread calculates the response of one logging point. Compared with CPU-based algorithm, the GPU-accelerated algorithm can achieve a speedup up to 100 times when the number of logging points is large.