Void statistics as deterministic tools in avalanche prediction

Abstract

We analyze the statistics of voids or spaces in a sago grain pile to investigate the possibility of prediction of avalanche events. Sampled frames from a video of a sago granular mound formed by adding grains at a rate of ≈ 1/2 Hz are processed to detect void distributions in time. A time series plot of the normalized standard deviation σ/μ of the void sizes is examined, revealing specific frames corresponding to spikes in the plot. These spikes are due to the occurrence of large voids, indicating pile instabilities that precede avalanche events. The method detects both large and small spikes in the time series, corresponding to large and small avalanche events, respectively. The sharp increase in σ/μ and its subsequent decay in an avalanche event provides a small window for intervention, which in this case is in the order of ≈ 2 s, the driving period of the pile.