基于人工神经网络方法的CRCP板顶拉应力影响因素敏感性分析

Analysis on Sensitivity of Influencing Factor of Tensile Stress on Top of CRCP Slab Based on Artificial Neural Network

为研究相对刚度半径、平均开裂间距、横向裂缝传荷能力及路肩传荷能力等因素,对配筋式连续水泥混凝土路面(Continuously Reinforced Concrete Pavement,简称CRCP)板顶产生的最大拉应力影响的敏感性,运用ISLAB2000有限元计算软件分别计算了不同输入组合情况下的应力的值,并使用1 260次ISLAB2000的计算的结果培训了人工神经网络(Artificial Neural Network,简称ANN)用来预测荷载应力值。通过正交设计的方法分析了各因素对CRCP板顶拉应力影响的敏感性。分析结果显示对表面拉应力的影响顺序依次为:路面板厚度,路肩传荷能力,横向裂缝传荷能力,开裂间距。通过绘制应力水平等高线云图进一步说明了路肩传荷能力对荷载产生的拉应力的影响比横向裂缝传荷能力更大。通过提高路肩位置的传荷能力,降低临界荷位应力水平,可减少CRCP发生冲断的可能性。

In order to investigate the effect sensititity of relative stiffness radius, average cracking spacing, load transfer ability of transverse cracking （ LTAc ） , and load transfer ability of road shoulder （LTAs ） on the maximum tensile stress on the top of continuously reinforced concrete pavement （ CRCP）, finite element software ISLAB2000 is employed to predict the stresses under various input combinations, and artificial neural network （ANN） is trained with the result of 1 260 runs of ISLAB2000 to predict the tensile stress under vehicle load. The effect sensitivity of different factors on the tensile stress on the top of CRCP is analysed with the method of orthogonal design. The analysis result shows that （ 1 ） the order of the effect of the factors is pavement slab thickness, LTAs, LTAc, and average cracking spacing. The stress contour is plotted to illustrate that LTAs is more influential than LTAc on the vehicle load induced tensile stress. By improving the LTAs and reduce the critical load induced stress level, the probability of the occurrence of punchout will be decreased.