Abstract:Rockburst possibility prediction is an important activity in many underground opening design and construction as well as mining production.  Insufficient knowledge，lack of characterizing information and noisy data restrain the rock mechanics engineers as well as mining engineers from achieving optimal prediction results.  In this paper the authors present a novel approach to predict probable rock bursts in underground openings.  The approach is based on learning and adaptive recognition of neural networks and allows input infomation to be incomplete，vague qualitative and noisy.  The predicion task is carried out by two neural network subsystems in cascade.  First a neural network is used to predict intensity and location of probable rock bursts.Next, another neural network uses this predicition and other geological features to identify the practical measures for prevention and mitigation of rock bursts.  The experimental results on 1 0 cases show that a rockburst prediction accuracy of 1 00%was reached with constructed two neural network subsystems.

Key words: rockburst   prediction   neural network