| Popis: |
Abstract The quality of microseismic location data from a volume of interest is dependent on the three dimensional geometry of the monitoring array. When the volume of interest contains an arbitrary shaped cavity not all the sensors will detect all the seismic activity. The performance of the array can be modelled so its configuration can be optimised. A genetic algorithm is used to steer the search for the optimal array of sensors. Various fitness functions were tested in the genetic algorithm and the minimax criterion was found to be the most satisfactory. 1 INTRODUCTION The delineation and characterisation of microseismic or acoustic emission (AE) activity has provided useful information in many geological and geotechnical settings. The spatial location of the seismic activity recorded is amongst the primary data provided, therefore the accuracy of these locations is important. The usefulness of the location data and subsequent interpretations obtained are functions of many variables. One of the most important of these variables is the geometry of the spatial positions of the seismic monitoring array1. The location accuracy of the recorded seismicity, allowing for certain simplifying assumptions, can be modelled before the actual network is installed2. This means that it should be possible to optimise the geometry of the network with respect to the region where the expected seismic activity will take place. The optimisation process which is described in this paper consists of finding a configuration of seismic sensors which gives the most accurate locations for a set of hypothetical microseismic events. These hypothetical microseismic events are located where the actual microseismic activity is either expected to be greatest or is of the most critical interest. The network optimisation problem is essentially a search problem, a search to find the optimum combination of sensors from a larger population of candidate sensor positions. For traditional seismic networks placed on the surface of the earth, where the problem is essentially two dimensional, Monte Carlo search techniques have been used to optimise the network configurations3. Two added complications occur when the region which is to be monitored contains a cavern or void through which no seismic energy can travel. Firstly, not all the sensors will be able to detect seismic energy from all the areas of seismic activity. Some of the sensors will be screened by the cavern. Hence only sub-sets of sensors will detect any given microseismic event. Secondly, the seismic network placed around such a structure may well be three dimensional. This means that the process of network optimisation becomes more complex and so a search strategy more sophisticated than a simple Monte Carlo approach is need. The monitoring situation is shown schematically in Figure 1. As the figure indicates only a sub-set of the sensors will be able to detect seismic signals from each microseismic event. The situation becomes more severe the nearer the microseismic events are to the cavern. |
