Research On Risk Assessment Of Tailings Dam Based On Possibility Theory

Tailings dam is a dangerous source in the mineral resources. In recent years, frequentaccidents about tailings dam have occurred, which lead to serious losses to the people and ourcountry. Therefore, it is urgent to strengthen safety monitoring and make risk assessment fortailings dam. However, due to the imprecision of the sensor itself and the influence of variousexternal factors, the monitoring information has the characteristics of randomness andfuzziness. Thus there exist difficulties in extracting effective information features, whichaffects the results of risk assessment directly. So it is necessary to analyze the features ofmonitoring information and extract instability features reflecting dam running status and setup risk assessment model.In this paper, on the background of monitoring system of upstream tailings dam and bymeans of possibility theory, instability features were extracted and a risk assessment model oftailings dam based on the instability feature was built. Firstly, possibility distribution functionwas introduced for the uncertainty description of the monitoring information, becauseprobability density function and membership function can only deal with a kind of uncertainty.Secondly, based on analyzing possibility distribution function, digital features were extractedand the sensitivity of digital features reflecting the dam running status was researched.Instability features were extracted by SVD (singular value decomposition) and neural networkmethod, and a case study proved its validity. Finally, the possibility and consequences of riskevents were quantified by analyzing the similarity of instability features and a tailing dam riskassessment model was built.This paper carried on risk assessment of tailing dam through analyzing instabilityfeatures. In the assessment process, the uncertainty was reduced by measuring the uncertaintyof monitoring information. Then two risk variables were quantized. This method not onlyreduced the subjectivity of original assessment method, but also described the transition statewhich can not be represented with original method.