Application of parallel processing to probabilistic fracture mechanics analysis of gas turbine disks

Abstract

A parallel processing methodology is developed and applied to improve the efficiency of probabilistic fracture mechanics analyses of gas turbine disks subject to metallurgical defects. A parallel processing spatial decomposition approach using a network of workstations and personal computers is described whereby each computer analyzes a region of the disk. The individual analyses are then combined to obtain the probability of fracture of the total disk. Practical implementation issues of job scheduling and optimum file size are addressed. Numerical applications are presented that demonstrate the methodology. This capability can significantly facilitate efficient evaluations of gas turbine rotor designs.