Research on risk assessment method for tailings dam failure based on multi-source data fusion

Abstract

[Objective] The tailings dam failure accident, as a high-energy, low-frequency extreme event, poses a serious threat to the safety of people's lives and property in the downstream. To accurately assess the risk of dam failure, a risk assessment method based on monitoring data-level and decision-level fusion is proposed. [Methods] Based on the multi-source online monitoring data of the tailings dam and a four-level risk system framework, an adaptive weighted fusion algorithm(AWF) was used to achieve data-level fusion of monitoring projects such as displacement, minimum dry beach length, and infiltration line. To address the issue of conflicting evidence in D-S evidence theory, the cloud model was used to construct the basic probability assignment(BPA). The Wasserstein distance and an improved normalized projection method(iNP) were introduced to measure the degree of conflict in the evidence bodies and construct fusion weights, thereby achieving decision-level fusion of dam failure risk based on the improved D-S evidence theory. [Results] The case analysis result showed that the application of the proposed method for risk assessment of a tailings dam yielded consistent result with actual observations, both indicating “low risk”, with a support degree of 0.891 5 for the assessment result. [Conclusion] The result indicate that the risk assessment model based on multi-source data fusion has a higher support degree for the assessment result compared to existing typical improved method, offering more accurate and reliable risk assessment for tailings dams.

Keywords

multi-source data fusion / tailings dam / D-S evidence theory / risk assessment / conflicting evidence