Influencing factor assessment model of loess slope stable based on neural network

To objectively analyze the factors of slope stability, as well as the coupling effects between such factors, this paper has collected 476 slope case reports from the loess deposition area of the Ordos Basin. On the basis of sufficient data, finite difference numerical models were established for these slope engineering cases, and the stability coefficients of each slope were analyzed and calculated. And then, using the strength parameters of loess and the geometric data of slopes as the input set, and the stability coefficient obtained through finite difference calculation as the output set, deep neural networks (DNN) are then employed to train these sets, resulting in a proposed method for establishing an influencing factors assessment model. Finally, the analysis and optimization of the loess slope design schemes and strategies were proposed in the loess deposition area of the Ordos Basin. It can provide reasonable references to understand the current loess engineering issues and risks, and the future engineering planning.