基于PCA-LM-BP神经网络的岩石可钻性预测研究

Research on rock drill ability prediction based on PCA-LM-BP neural network

摘要:

预测岩石的可钻性等级能够为钻探工程项目的开展提供有效帮助，根据岩石的可钻性等级选择合理的工艺、方法、技术为项目提供技术支撑。本文考虑岩石在地下空间中受复杂环境因素影响，从地球物理勘探数据、岩石的力学性质和物理性质中选择5种影响岩石可钻性的等级因素，用主成分分析法（PCA）解释每种影响因素之间的相关性及贡献率，消除5种影响因素之间的相关性，选择相关性低的3个主成分代替数据样本进行预测评价。编写LM-BP算法，合理设置预测模型参数值，以主成分分析后的数据样本作为基础，建立岩石可钻性等级预测模型，对预测结果与室内实验法的实测结果进行分析对比，经分析得知，PCA-LM-BP预测模型在岩石可钻性等级预测中，具有预测精准度高、预测时间短的特点，可被应用于钻探工程中的岩石可钻性分析。

Abstract:

Predicting the drill ability level of rocks can provide effective assistance for the development of drilling engineering projects， and selecting reasonable processes， methods， and technologies based on the drillability level of rocks can provide technical support and assistance for the project. In this paper， considering the impact of complex environmental factors on rocks in underground space， five factors affecting rock drill ability grade are selected from geophysical exploration data， mechanical properties and physical properties of rocks. Principal component analysis （PCA） is used to explain the correlation and contribution rate between each influencing factor， and the correlation between the five factors is eliminated. Three principal components with low correlation were selected to replace the data samples for prediction evaluation. The LM-BP algorithm was compiled， reasonably set the parameter values of the prediction model， and based on the data samples after principal component analysis， establish a rock drill ability level prediction model. Analyze and compare the prediction results with the measured results of indoor experimental methods. It was found through analysis that the PCA-LM-BP prediction model has the characteristics of high prediction accuracy and short prediction time in rock drill ability level prediction， it can be applied to rock drill ability analysis in drilling engineering.