Based on the deep foundation pit works of Wuhan Metro Line 2 Gold Garden station, orthogonal test method, finite difference method and BP neural network method are synthetically used to do the inverse analysis study of mechanical parameters of foundation pit. The mechanical parameters are substituted in the finite difference process of deep foundation pit to simulate and calculate. The deformation and bearing situation of building enclosure and soil in the construction process of foundation pit is acquired, which is returned to the design and construction management to provide reference for design modification, construction optimization and engineering prediction, to ensure the security and stability of foundation pit and to greatly reduce investment outlay.