机构地区: 东南大学
出 处: 《公路交通科技》 2006年第5期7-10,共4页
摘 要: 以人工神经网络法为主,研究了多因素条件下的沉降预测问题,同时与浅岗法和S型曲线法这2种在近几年推广应用的预测方法进行了对比。结果表明,3种方法预测的最终沉降大体相近,它们之间的区别在于人工神经网络法预测的沉降较大(同时更接近实测值);S型曲线法较小;浅岗法居中。由于神经网络是用实测数据直接建模,少了人为干扰因素,并且偏大的数值对工程来说是偏于安全的,所以选用人工神经网络预测沉降比较适宜。 The artifical neural networks (ANN) are studied as a main method for settlement prediction, compared with two other methods of Asaoka method and sigmoid curve method used in recent years. The study results indicate that the predicted final settlements with the three methods are consistent on the whole, and that there are differences among them with greater value for ANN (closer to actual settlement); smaller for sigmoid curve method; and Asaoka method in between.Since ANN builds the model directly from survey data and reduces the man-made effect, furthermore, the bigger prediction value is more safety in project, it is feasible to predict settlement with ANN.
关 键 词: 软土地基 沉降预测 人工神经网络 浅岗法 曲线
领 域: [交通运输工程]