Research On Tailing Pond Risk Prediction Algorithm Based On Support Vector Machine

With the rapid development of industrial economy,the society's demand for mine resources is gradually increasing.However,more than 90% of the tailings enterprise using tailing pond to stockpiling tailing residue owing to the low rate of ore extraction and the low construction cost of ore mining tailings,which gradually formed a situation that the number of tailings is large and the number of safety personnel is not enough.In recent years,the mass casualties caused by the collapse of the tailings occurred frequently,causing serious harm to public property and personal safety.Therefore,it has great significance for us to put forward a scientific and precise,fast and reliable prediction method for the tailings safety risk assessment to meet government and enterprise's the needs of regulation.Most of the existing models for assessing tailings pond's risk are lack of spatial index evaluation,and it has not been adequately excavated in the data collected from the tailings monitoring system.Aiming at the problems above,this paper proposes a method based on improved support vector machine to predict the safety risk of tailings pond.The main work and achievements are as follows:(1)This paper analyses the existing tailings pond risk assessment method,summarizes the shortages of these methods using different theories and methods.This paper establishes a new tailings risk evaluation model by adding the index spatial position information aiming to improve the tailings risk assessment on spatial information.(2)On the basis of the tailings project structure,this paper analyses the main causes of the tailings' accidents.According to the characteristics of monitoring indicators such as dry beach length,the water level,dam displacement,underground displacement,etc.,this paper selects the key monitoring indicators clearly in the risk assessment through the feature extraction.(3)In this paper,the ensemble learning theory is applied to the support vector machine regression.Based on the stacking algorithm,we establish an improved support vector machine risk grade prediction algorithm,and carry out the details of simulation experiments.The results show that the improved support vector machine has the superiority in prediction accuracy and training speed.(4)We apply a novel meta-heuristic algorithm based on the improved cuckoo search algorithm to figure out the optimal parameters for support vector machine so as to get the global optimal solution.Besides,we will acquire a risk assessment SVM model with optimal parameters.In order to prove the validity of our algorithm,we perform some experiments to compare our algorithm with other algorithms,such as the particle swarm algorithm and the genetic algorithm.The result shows that our algorithm has high prediction accuracy and fast convergent rate.The results showed in this paper verify that the improved SVM is a viable risk prediction method for the tailing ponds.It can precisely reflect the change trend of the tailing ponds' risk grade.This paper provides a fresh thought and a new theoretical method for ensuring the safety of the tailings.