DISCRIMINATION BETWEEN LOCAL MICROEARTHQUAKES AND QUARRY BLASTS BY MULTILAYER PERCEPTRONS AND KOHONEN MAPS

The results of the application of artificial neural nets (ANNs) to dis criminating microearthquakes from quarry and mining blasts in the West Bohemia earthquake swarm region are presented and discussed. Input ve ctors consisting of seven spectral and seven amplitude parameters, aut omatically extracted from local three-component digital broadband (0.6 to 60-Hz) velocigrams, have been employed for training of different A NN configurations. Multi-layer perceptrons (MLP) trained in supervised mode by different subsets of a representative set of 312 events have been used as discriminators, and unsupervised Kohonen self-organizing feature maps (SOFM) have been used as complementary reliability estima tors. The reason for comparative application of both techniques was to increase the reliability of the discrimination: complementary informa tion that a pattern has been recognized as a member of a conflict clus ter allows detecting problematic patterns that an MLP may not be able to classify correctly. The optimal MLP, trained by one randomly select ed half of the complete set of 312 input vectors and tested by the oth er half-set, and vice versa, correctly classified, on average, 99% of all events. The optimal SOFM correctly classified as problematic patte rns all events misinterpreted by the MLP, and about 20% of all events were classified by them as ambiguous cases. The obtained results evide nce that a relatively small number of spectral and amplitude parameter s of observed ground velocity may suffice for a reliable discriminatio n between local microearthquakes and quarry blasts by means of small n eural nets. The MLP/SOFM combination discussed in this article has att ained a discrimination reliability that allows it to be employed routi nely in observatory practice.