The traditional faulting distress detection method is a manual detection method, which is inefficient, less reliable and dangerous to use. To overcome these shortcomings, a new automatic detection method based on statistics values of vertical acceleration was proposed. Firstly, noise reduction was applied to acceleration signals by 5 layers Daubechies wavelet, acceleration sequence was put forward from the acceleration-gray map and a new joint location method based on the joint periodicity in the acceleration sequence was obtained. Secondly, three statistic parameters were adopted to analyze the acceleration variation near the joint, which included the position distribution of peak acceleration values (gm) as well as the accumulative mean of acceleration values (ge) and the accumulative absolute mean of acceleration values(ga) at the same position on each slab from the acceleration-gray map. Test results show that the position distribution of peak acceleration values(gm) on slabs is similar to random distribution, peak acceleration values have no relation with faulting distress. The position of peak values and the amplitude near each joint on curves of ge and ga have weak relation with faulting distress. The increment near each joint on the curve of ga has fair relation with faulting distress. Based on the analysis, a new acceleration statistic parameter (af) was put forward to estimate the faulting value, which had good relation with faulting distress.