论文部分内容阅读
为有效辨识金属矿山采空区危险性等级,搜集国内50组典型的金属矿山采空区失稳资料,建立遗传BP神经网络辨识模型。该辨识模型利用遗传算法全局搜索的优点,确定BP网络初始权值和阈值,改善了纯BP模型选择初值时的随机性。对辨识模型进行训练和检验,得出模型对训练样本和检验样本的平均辨识误差分别为1.14%和1.54%。通过计算辨识模型训练后输入节点的权重值,确定对采空区危险性影响最大的3个因素依次为矿柱面积比、岩石质量指标(RQD)值和矿柱宽高比。将该辨识模型应用于某金属矿山采空区实例,所得辨识结果与实际情况相符。
In order to effectively identify the risk level of the metal mine goaf, metal instability data of 50 typical metal mines in China were collected and a genetic BP neural network identification model was established. The identification model uses the advantages of global search of genetic algorithm to determine the initial weights and thresholds of BP network and improves the randomness of initial selection of pure BP model. The recognition model is trained and tested, and the average recognition errors of the model and the test sample are 1.14% and 1.54% respectively. By calculating the weights of the input nodes after the model was trained and identified, the three factors that have the greatest impact on the hazard of goaf are the pillar area ratio, rock mass index (RQD) and pillar aspect ratio. The identification model is applied to an example of a metal mine goaf, the resulting identification results in line with the actual situation.