滨海金矿涌水危险评价及涌水量混沌预测研究

针对海底金矿涌水危险评价过程中的不确定性及随机性问题，选择工程地质及水文地质中涉及的10个重要因素构建评价指标体系，建立起涌水危险评价的云模型。基于评价结果，对危险区域的涌水量时间序列进行相空间重构，通过G-P算法及自相关法获得了重构相空间参数；分析了涌水量变化的相点距离演变规律，建立了涌水安全预警机制。结合混沌相空间重构，建立了涌水量预测的RBF神经网络模型。研究表明：涌水危险性的云模型评价结果准确可靠；相空间重构揭示了系统的混沌特性，最邻近相点演化将涌水量的内在细微变化特征放大，为涌水安全预警机制的建立提供了依据；混沌RBF神经网络能够实现涌水量的短期精确预测，为井下安全开采提供了技术保障。

Research on Water Inrush Risk Assessment and Water Inflow Chaotic Prediction of Coastal Gold Mine

Mine water inrush is one of a major problem in the process of mining，and it is also a key technical problem in seabed mining.At present，the study on the potential dangers of water gushing and inrush in the mining of submarine metal deposits in China is still in its infancy，and there is no mature experience to draw lessons from.Therefore，this paper has made relevant research on the risk assessment and safety early warning system of seawater intrusion in seabed metal deposits.In the first，in view of the uncertainty and randomness in the risk assessment process of submerged gold mine water，an evaluation index system is established by selecting 10 important factors in geological and hydrological conditions.The factors are RQD value，joint spacing，rock permeability coefficient，structural plane condition，in-situ stress，water inrush volume，seawater proportion，brine proportion，atmospheric precipitation ratio and absolute value of seepage temperature difference.In the second，based on the evaluation results，the phase space of the time series of water inflow in dangerous areas is reconstructed，and meanwhile，the parameters of reconstructed phase space are obtained by G-P algorithm and autocorrelation method.What’s more，the evolution law of phase distance of water inflow is also analyzed，and then the early warning mechanism of water inflow safety is established.Combining with the reconstruction of chaotic phase space，the RBF neural network model for forecasting water inflow is established.Finally，after 6 000 training iterations，the model meets the training requirements，which is used to compare the simulated value with the actual value.It is found that the predicted result of water inflow within 13 days is basically consistent with the actual value，with the maximum error of 1.48%.After more than 13 days，the prediction effect becomes worse and worse.Therefore，the chaotic neural network model can be used to make a more accurate prediction of water inflow in a short time.The final research shows that the evaluation results of cloud model for water inrush risk are accurate and reliable.On the one hand，the phase space reconstruction reveals the chaotic characteristics of the system，and the evolution of the nearest phase points enlarges the intrinsic subtle variation characteristics of water inrush volume，which provides a basis for the establishment of early warning mechanism for water inrush safety.On the other hand，the chaotic RBF neural network can achieve short-term accurate prediction of water inrush volume and provide technical guarantee for underground safe mining barrier.