Abstract: The 5# adit of Xianglushan tunnel under construction of Central Yunnan Water Diversion Project features complex geological conditions as it crosses active fault zones. Severe water-mud bursting disasters and prominent stability problems of surrounding rock hinder the construction progress and project safety. According to field monitoring and geophysical exploration data, we determined the mechanical parameters of surrounding rock at emergency rescue section by using inversion analysis based on neural network and genetic algorithm; on this basis, we simulated the whole process of construction, excavation, and support, and analysed the surrounding rock stability. Results manifested that the surrounding rock mass of emergency rescue tunnel section of adit 5# was in an overall stable state. Except that the deformation of surrounding rock on the right side of the tunnel section K0+501-513 was relatively large, the deformation of other parts was less than 15 cm in general; the depth of plastic zones was within the range of 2-5 m; and the stress of support structures was at a normal level. The research findings would guide the safe and rapid construction of subsequent tunnel sections of 5# adit or tunnels with similar geological conditions.

Key words: inversion of rock mass parameters, neural network, genetic algorithm, stability analysis, Xianglushan Tunnel