Research On The Prediction Model Of Blasting Fragmentation Distribution On Open Pit Bench And Engineering Application

Blasting fragmentation distribution is a important index to measure the blasting quality, rational blasting fragmentation of mine is significant for later production to improve efficiency and reduce the production cost. Control the rock fragmentation has been the difficult technical problems in the process of open-pit mine production, because of the size of the blasting fragmentation is influenced by many factors, and it is difficult to use mathematical formula to describe, often based on experience to design blasting parameters in the actual mine production, but the blasting effect is difficult to meet the production demand.Neural network is an new technology in recent decades, it has excellent performance in pattern recognition, signal processing, automatic control, and auxiliary decision-making, neural network has strong nonlinear ability, self learning ability,adaptive capacity, generalization, and fault tolerance ability, so that it is valuable to solve the complex rock blasting process, which influence factors on blasting effect is difficult to use mathematical description. Combining Yichun tantalum niobium mine’s production practice, in view of the high blasting chunk rate, the high powder ore rate and the large volume that exist in the industrial production,,from the perspective of the influence factors of rock blasting fragmentation, analysis the main influence factors, such as pore diameter, hole deep, chassis resistance line, hole distance,row spacing, plugging length and the unit explosive consumption etc. Then introduce the neural network theory,use blasting parameters as network input layer and blasting fragmentation distribution as network output layer, then set up a three layer neural network prediction model, select12main influencing factors as input layer node, four different distributions of rock fragmentation as output layer node, and through try the network,when the hidden layers node is16, the network model’s prediction error is minimality.After then put the test sample data into the network prediction model to test network model’s prediction ability, through compare the output value and actual value, it was found that network can satisfy the precision requirement, the model has certain guiding significance and reference value for the mine blasting practice Finally, through optimize blasting parameters, set a blasting test, the prediction model works well, and has a good applied potential.