Jones, Christopher Charles Rawlinson (2002) A study of novel computing methods for solving large electromagnetic hazards problems. Doctoral thesis, University of Central Lancashire.
PDF (Thesis document)
- Submitted Version
Restricted to Repository staff only Available under License Creative Commons Attribution Non-commercial Share Alike. 259MB |
Abstract
The aim of this work is to explore means to improve the speed of the computational electromagnetics (CEM) processing in use for aircraft design and certification work by a factor of 1000 or so. The investigation addresses particularly the set of problems described as electromagnetic hazards comprising lightning, EMC and the illumination of an aircraft by external radio sources or HIRF (high intensity radiated fields). These are areas which are very much aspects of the engineering of the aircraft where the requirement for accuracy of simulations is of the order of 6dB as build and test repeatability cannot achieve better than this. Computer simulations of these interactions at the outset of this work were often taking 10 days and more on the largest parallel computers then available in the UK (Cray T3D - 40 GFLOPS nominal peak). Such run times made any form of optimisation impossibly lengthy.
While the future offered the certain prospect of more powerful computers, the simulations had to become more comprehensive in their representation of materials and features, geometry of the object, and particularly the representation of wires and cables had to improve radically, and tum around times for analysis had to be improved for design assessment as well as to make design optimisation by trade-off studies feasible. All of these could easily consume all the advantage that the new computers would give.
The investigation has centred around techniques that might be applied via alteration to the most widely used and usable numerical methods in CEM applied to the electromagnetic hazards, and to techniques that might be applied to the manner of their use. In one case, the investigation has explored a particular possibility for minimising the duration of computation and extrapolating the resulting data to the longest time-scales required.
Future improvements in the capabilities of radiating boundary conditions to mimic the effect of an infinite boundary at close range will further improve the benefits already established in this work, but this is not yet realisable. However, it has been established that a combination of techniques with some processes devised through this work can and does deliver the performance improvement sought. It has further been shown that the issues such as object resonance that could have incurred significant error and distrust of computational results can be satisfactorily overcome within the required accuracy.
Four papers have been published arsing from this work. Some of these techniques are now in use in routine analyses contributing to BAE SYSTEMS programmes. Plans are in place to incorporate all of the successful techniques and processes.
Repository Staff Only: item control page