研究目的:静荷载试验是基桩检测中最直观有效的检测方法,其试验数据的科学有效性对试验结果的分析判定至关重要。本文旨在通过并联灰色神经网络模型,探讨静荷载试验中相似地质条件下基桩沉降量的有效预测和数据修补问题,并给出计算方法。
研究结论:通过对桩周围土体的灰色关联分析,确定了区域内桩周围土体的关联度;借助线性加权的方法对灰色模型和BP神经网络进行并联整合,实现对单一模型的降噪优化;运用并联灰色神经网络,对相似土层区域范围内单桩静荷载试验数据进行有效预测,并进行误差比对。结果表明:该方法可综合考虑多方因素,对试验过程中缺失数据的修复、已知沉降量的拟合、未来沉降量的预测和关联区域内基桩沉降量参考值的确定具有实用价值。
Research purposes: The static load test is the most visual and effective approach in all foundation pile tests and the scientific validity of the tested data is essential to analysis of test result.By using parallel gray neural network,this paper discusses the effective prediction of foundation pile settlement and data remedy in static tests under the similar geological conditions,and offers the calculation methods for them.
Research conclusions: The correlation degree of the soil mound pile is decided by making gray correlation analysis. With the aid of the linear—weighted method,the gray model is integrated with BP neural network to optimize the noisy—reduction for sole model.By adopting parallel gray neural network,the effective static load test data of sole pile in the area of the similar soil layer are predicted,and the errors are compared.The results indicate that overall factors could be considered in this approach,and the decisions of missing data remedy in the test,fitting of known settlement,prediction of future settlement and reference value of foundation pile settlement have practical values in correlation region.