【摘要】 为提升岩石单轴抗压强度的预测效率与精度，提出一种基于卷积神经网络(CNN)与长短期记忆网络(LSTM)相结合的预测模型，并采用牛顿-拉夫逊优化算法(NRBO)对模型关键参数进行优化。通过广泛收集包括点载荷强度、施密特锤回弹数、孔隙率及纵波波速的岩石物理参数试验数据，经预处理后保留381组有效样本，并进行可视化分析与异常值处理。采用CNN进行特征提取，LSTM捕捉时序依赖，NRBO优化学习率、正则化系数及隐藏层节点数等超参数。试验结果表明，NRBO-CNN-LSTM模型在测试集上表现最优，其决定系数达到0.98,均方根误差为6.27,平均绝对误差为4.82,显著优于其他对比模型，显示出更强的拟合能力与泛化性能。该优化模型能够有效融合岩石多源物理特征，实现高精度、高稳定性的单轴抗压强度预测，可为工程实际提供可靠的技术参考。更多还原

【Abstract】 In order to improve the prediction efficiency and accuracy of uniaxial compressive strength of rocks, a combined model based on convolutional neural network(CNN) and long short-term memory(LSTM) network was proposed and optimized using Newton-Raphson-based optimizer(NRBO). Extensive collection of rock physical parameter experimental data, including point load strength, Schmidt hammer rebound number, porosity, and longitudinal wave velocity, was conducted. After preprocessing, 381 valid samples were retained and visualized for analysis and outlier handling. CNN was used for feature extraction, LSTM captured temporal dependencies, and NRBO optimized hyperparameters such as learning rate, regularization coefficient, and number of hidden layer nodes. The experimental results show that the NRBO-CNN-LSTM model performs the best on the test set, with a coefficient of determination of 0.98, a root mean square error of 6.27, and an average absolute error of 4.82, significantly better than other comparison models, demonstrating stronger fitting ability and generalization performance. This optimization model can effectively integrate multiple physical characteristics of rocks, achieve high-precision and high stability prediction of uniaxial compressive strength, and provide reliable technical reference for engineering practice.更多还原