Single-node SOR Method for statistical 3D thermal analysis

摘要

As the first study on Statistical 3D thermal analysis,we propose an efficient method for fewer over-hot spots. The new method,called SN-SOR (single-node SOR),is based on a novel localized relaxed iterative approach and it can perform statistical analysis on one over-hot spot at a time. Experiments show that SN-SOR is 590X faster than Monte-Carlo method with small errors and is 29X faster than general global SOR(successive over relaxation) method in statistical analysis of 450 over-hot spots. As a result,SN-SOR only take average 0.1481 second to statistically analyze one over-hot spot for a large test case of 1.3M nodes under spatially correlated energy and thermal conductance disturbances.