Abstract This paper addresses the performance analysis of an optimization algorithm from the field of evolutionary computation, namely an Evolution Strategy, in back analysis to evaluate the geomechanical parameters of the formation surrounding an underground structure. The algorithm is first tested using a synthetic case of a tunnel excavation. In this case, different scenarios are considered through a parametric study. Then the algorithm is tested using real data from the excavation of an underground structure built in the North of Portugal using a 3D model. The results show that the Evolution Strategy algorithm is robust in the identification of geomechanical parameters related to underground engineering.
1. Introduction In back analysis, field measurements are used together with models to calibrate input parameters (geomechanical, stress state, etc…) matching, under a defined tolerance, predicted with observed measures. Normally, an iterative procedure is needed to find the best set of parameters through the minimization of an error function that measures the difference between real and computed quantities. For the minimization task optimization algorithms are used and this is a main issue to obtain the best set of parameters.
In geotechnics, two main types of algorithms have been used in back analysis: algorithms from the field of classical optimization theory such as the Simplex, the Newton-Raphson or gradient methods; and Evolutionary optimization algorithms like Genetic Algorithms (GA), Evolution Strategies (ES), Simulated Annealing, etc. (Moreira et al., 2013).
Classical algorithms present a satisfactory performance in smooth-shaped error functions, with a clearly defined and unique minimum (Miranda, 2007). However, the error function topology is normally complex and the uniqueness of the solution cannot be guaranteed since many local minima may occur. Hence, this kind of algorithms are most of the times limited to simpler models and a reduced number of parameters to identify (Moreira et al., 2013).