|
Linear prediction of hourly aggregated AE and tremors energy emitted from a longwall and its perform |
|
|
|
|
Linear prediction of hourly aggregated AE and tremors energy emitted from a longwall and its performance in practice Author: J. Kornowski
Possibilities and achievable results of some variants of the linear prediction method have been studied with four long time series ofhourly, logarithmic, total (AE + tremors) energy as observed with AE and seismic networks at two mining longwalls. Energies of AE and seismic events from the same space region and time interval (of one hour) have been aggregated resulting in evident1y autocorrelated, predictable (to a degree) time series. Two of the serie s have been observed at the not very hazardous longwall and the two others - at a very hazardous one. It is argued that, contrary to difficulties encountered while predicting parameters (eg. time, place and magnitude) of seismic events, prediction of time series of total (AE + tremors) energy emitted from the observed longwall at constant time intervals is well-defined, simple to predict in the probabilistic sense and can be useful in practice. The linear prediction method allows to predict at any discrete time moment t sub i, the mean value and variance of the energy which will be emitted at the observation region during the nearest time unit (e.g. during the nearest hour). Since the prediction error distribution can be (for logarithmic energy data) approximated with the normal distribution, confidence intervals for prediction and probabilities of any ("safety" or "alarm") threshold cxceedance can be easily estimated allowing for the formai hazard assessment. Up to (approximately) ten-fold reduction of prediction error variance and threefold reduction of confidence intervals for prediction have been obtained, comparing to prediction which takes benefit of tremors only (i.e. neglecting AE) data.
|
2011