| 杨青松.基于概率神经网络的智能找矿方法—以四川雅江县木绒锂矿为例[J].地质与勘探,2023,59(5):985-999 |
| 基于概率神经网络的智能找矿方法—以四川雅江县木绒锂矿为例 |
| Intelligent prospecting method based on probabilistic neural network: Taking the Murong lithium deposit in Yajiang County of Sichuan Province as an example |
| 投稿时间：2022-12-06 修订日期：2023-07-19 |
| DOI：10.12134/j.dzykt.2023.05.006 |
| 中文关键词: 地电化学 稀有金属 木绒锂矿 概率神经网络 智能找矿预测 雅江县 四川省 |
| 英文关键词: geoelectrochemistry, rare metals Murong lithium deposit, probabilistic neural network, intelligent prospecting prediction, Yajiang County, Sichuan Province |
| 基金项目:国家自然科学基金（编号：42203067）、广西科技计划青年基金项目（编号：2021JJB150149）和广西高校中青年
教师科研基础能力提升项目（编号：2022KY0261）联合资助 |
|
| 摘要点击次数: 787 |
| 全文下载次数: 349 |
| 中文摘要: |
| 概率神经网络是一种分类准确率高、泛用性强、可以包容一定数量错误样本的人工神经网络，极其适合勘查地球化学找矿中的预测找矿靶区。本文以四川雅江县木绒锂矿为例，运用概率神经网络搭建智能找矿模型，以已知区的Li元素及与其相关性强的Rb-Cs-Al-Fe元素作为训练指标，对模型进行训练，经过多次训练后将Spread值确定为0.08，使模型在训练集和测试集的准确率均大于80%，实现非线性的指标与成矿潜力的对应，得到本矿区的PNN模型，然后对预测区的样本数据进行预测，成功圈定了1处靶区。为检验靶区准确性，以Li、Rb、Cs元素数据累计频率的80%作为异常下限，圈出的异常区域与靶区位置基本重叠。对预测区进行了实地查证工作，发现两条红柱石带，其中一条与靶区位置吻合，表明该神经网络模型准确性高，可用于矿产勘查的预测研究。 |
| 英文摘要: |
| Probabilistic neural network is an artificial neural network that can accommodate a certain number of erroneous samples, has high classification accuracy and is highly generalizable, and is extremely suitable for predicting target areas in exploration geochemical prospecting. Taking the Murong lithium mine in Yajiang County of Sichuan Province as an example, this work used probabilistic neural network to build an intelligent prospecting model. Taking Li elements in known areas and strongly correlated Rb-Cs-Al-Fe elements as training indicators, trained the model, and determine the Spread value as 0.08 after several training sessions. The accuracy of the model in both training and test sets is greater than 80%, which has realized the correspondance nonlinear indicator with mineralization potential. A PNN model was built, the sample data of the prediction area was predicted, and a target area was successfully delineated. In order to check whether the target area is correct, 80% of the cumulative frequency of Li, Rb and Cs elemental data was used as the lower limit of anomaly, and the delineated anomaly area basically overlapped with the target arean. Field verification work was carried out on the predicted area, and two andalusite bands were discovered, one of which matched the location of the target area, indicating that the neural network model is highly accurate and can be used in mineral prediction. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
