神经网络预测爆炸密实引起的地表沉降

爆炸密实法用于地基处理已有很长的历史,用此方法处理时所引起的地表沉降是衡量其处理效果的一个重要指标,许多学者提出了引起地表沉降量的统计预测模型。本文尝试建立了预测爆炸密实引起地表沉降的神经网络模型,由于用来训练网络的原始样本太少,难以训练出一个具有高精度的预测模型,为此采用了遗传算法与神经网络相结合的方法,得到了一个只有小样本输入却能得到高精度输出的神经网络模型。对检验样本的预测结果表明,所建立的神经网络模型具有很强的预测能力,其预测精度比已有的统计预测模型高。

PREDICTION OF SETTLEMENT OF THE EARTH'S SURFACE CAUSED BY EXPLOSIVE COMPACTION USING NEURAL NETWORKS

Explosive compaction has been used for ground improvement for many years. Surface settlement achieved by explosive compaction is one of important results to evaluate the strength, and therefore, many schol- ars proposed some statistical model of predicting surface settlement. In this paper,a predictive ANN model based on the ANN and GA was developed. Due to the data being not enough for training a precise model, GA was used to generate data for training ANN. Through this way, more precise model was acquired. Comparing the predicted settlements found by the ANN model with the values predicted by traditional statistical model indicated that the ANN model was more powerful and more accurate.