Performance Of Optimization Algorithms For Deriving Material Data From Bench Scale Tests

In this work the performance of optimization algorithms for inferring material parameters for fire modeling from bench scale tests are compared to each other.

The well known Shuffled Complex Evolution algorithm (SCE) is compared to Artificial Bee Colony algorithm (ABC) and Fitness Scaled Chaotic Artificial Bee Colony algorithm (FSCABC). First, these algorithms are tested with synthetic data, where all the properties are certain in advance. After that, the algorithms are tested with real data gained from bench scale tests, namely thermogravimetric analysis (TGA) and mass loss calorimeter (MLC). Fire Dynamics Simulator (FDS) with its implemented pyrolysis model is used to carry out the simulations in an automated optimization framework on a high performance computing cluster in parallel. The achieved results show which of the compared optimization strategies perform better than SCE related to efficiency and accuracy.