Parametric Modeling and Detection of Ripple Fired Signals,
Abstract
We develop a parametric statistical approach to detecting and estimating the delay and duration parameters of a ripple-fired signal. Such signals are generated by mining blasts with roughly equally spaced charges. Detection of the echo structure induced by ripple firing can serve as a basis for discriminating between mining blasts and nuclear explosions or earthquakes which, in principle, will not contain such echo effects. The model assumes a noise corrupted signal that can be written as a convolution of source, path and instrument responses with a pulse sequence. The underlying responses are modeled by low order autoregressive moving average (ARMA) models that are consistent with conventional deterministic formulations of source theory and instrument response. The pulse sequence is modeled as a seasonal ARMA process, with the period corresponding to the firing delay and with a model order that is proportional to the signal duration. We use the cepstrum to suggest a range of values for the duration and delay parameters and then search all possible seasonal ARMA models within this range. The final model chosen is the minimizer of the corrected Akaike Information Theory Criterion (AICc). Limited simulations are given to show that both the duration and delay can be estimated effectively with the seasonal ARMA search when the delays are equal and that reasonable estimators are available when the delays are variable. The methodology is applied also to the P phase of a single Scandinavian mining explosion where we obtain reasonable estimators for delay and duration under the assumption of ripple-firing.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 14, 1995
- Accession Number
- ADP204512
Entities
People
- Allan D. Mcquarrie
- Robert H. Shumway
Organizations
- University of California, Davis