Solving Geoseismic Problems with Soft Computing

ABSTRACT: Soft Computing, SC, is an association of computing methodologies that includes, as its most important members, fuzzy logic, neuro computing, evolutionary computing and probabilistic computing. These tools are a great match for geoseismic applications that require the analysis of uncertain and imprecise information and where an incomplete understanding of the phenomena further compound the problem of generating models used to explain past behaviours or predict future ones. We outline the advantages of applying SC techniques (neural networks, genetic algorithms and regression trees) and in particular the synergy derived from the use of hybrid SC systems (fuzzy systems tuned by neural networks and neurogenetic models). Some successful geoseismic SC applications (the estimation of liquefaction induced lateral spread, the modelling of Mexico City ground motions, the generation of synthetic seismic signals, the prediction of peak ground accelerations from earthquake-subduction mechanisms, and the spatial variation of soil -geometrical, mechanical and physicalproperties) are described. They represent the basis for producing the technology and human resources necessary to transition from current geoseismic design methods to more scientifically-based procedures.