Abstract: The premise and foundation of the scientific construction is to know the circle time and formation characteristics of the frozen wall under the different depths and lithologic strata in the process of vertical shaft freezing construction. In response to the existing objective problems of the vertical shaft freeze design, the reasonable input parameters as well as output parameters have been determined by using the neural network system theory. Based on the learning and training, the neural network prediction model of the vertical shaft freezing construction information has been set up. And then, the forming time and characteristic parameters such as inside and outside extended range, the average propagation velocity, effective thickness, the temperature of the wall, diameter range of the frozen rock-soil, average temperature have been predicted. Finally, the measured data and predicted data have been analyzed contrastively. The results show that, the measured results are consistent well with the predicted results. The prediction model has the advantages of the high prediction accuracy, wide applicability, which provides a theoretic foundation for the scientific design of the construction method and the supporting schemes during the course of the vertical shaft construction.

Key words: vertical shaft construction, frozen wall, neural network, engineering prediction