Machine Learning (ML) Model for Blasting and Seismic Event Classification on Chilean Copper Mines

In this paper, the support vector machine model (SVM) was used to classify the earthquakes and blasting from the files recorded by an open pit copper mine seismic network. The database cleaning of false positive records (vibrations records not from blasting) must be supervised. This implies spending several hours of highly skilled engineers doing this routine task. Changing to an unsupervised classification method, based on AI, has shown an extremely high overall accuracy (as high as 0.97 over the validation dataset), allowing to perform the database filtering on real time while the vibration events are recorded by the seismographs. 1 INTRODUCTION In open pit mining, it is a very common practice to implement a seismographers network into the pit to monitor 24/7 any vibration on the ground that exceed a certain limit and especially to record the vibrational event generated by blasting. These records are extremely important to understand and model the vibrational behavior of the rock mass and guarantees the operation safeness for workers and equipment. Due to the heterogeneously of the rock mass and the pit it self, the existence of unstable areas in the pit are a extremely likely scenario. It must be watched carefully and the vibrations generated by the blasting reported on real time to avoid risks. To model the rock mass vibrational response, the "tendency behavior" is widely used that means to model the Peak Particle Velocity (PPV) as function of "Scalar distance", the distance from the sensor to the closest blasthole, scaled by the explosive load of the blasthole. On the other hand, the database housekeeping is a highly time consuming process that must be performed by a well trained engineer and the main cause of that time expending is the matching between the blasting information given by the mining company (date, time and blast pattern position) to guarantee that all registered database is in fact blasting induced vibrations. As above mentioned, it is quite important to differentiate the vibration from blasting or by any other source, in particular seismic events.