Dynamic support method based on probability classification and device for automatically designing support scheme

The invention provides a dynamic support method of probability classification, which comprises the following steps: on-site indexes of tunnel construction are collected and probability distribution function thereof is obtained, and preset support grade is selected; indicators and support level samples are used for training differential evolution-limit learning machine machine learning model, and the implicit function of support grade is obtained by nonlinear fitting of samples; the trained DE-ELM machine learning model is used to calculate the probability statistic combination Monte Carlo sampling of the tunnel construction site index, and the support grade judgement of reliability is obtained; if the judgement result of the support grade is consistent with the preset grade, the preset support grade is used for construction; otherwise, the support grade is determined according to the displacement of surrounding rock. This method solves the problem that the randomness and nonlinearity of the automatic design o