Research On Pattern Recognition And Limit Analysis In Slopes Monitor

In this paper, a concept of state pattern of slopes together with a pattern recognition method with artificial neurual network is put forward for the analysis for the data from slope monitor system. The complexity of the factors in the stability of slopes makes the descripting of the stability of slopes very difficult. Suggested by the slide surface pattern in the analysis of stability of slopes, it is believed that anywhen the stable states of the slopes must belong to some state patterns. Any state patterns of slopes composed with a slide pattern and a distribution pattern of groundwater. In pratice, a set of slope state patterns must be built on the basis of thurough research on the essence of the slopes. The second work in this paper is to build a slope state pattern recognizor to analyse the monitor data of the slopes. The topology of the pattern recognizer is a multi-layer recognizor with BP network. The third work of this paper is introducing the slope limit analysis program EMU into the analysis of slope monitor data. The above slope pattern recognizor can bulid an EMU data file according to the stable state of the slope whom is indicated by the monitor data. After the data file is fed to the EMU program, a safety factor of the slope of any monitor time can be worked out and the safety state of the slope can" be easily judged. The fourth work is to build a curving and forcast model with artificial neurual network for the time series of the slope safety factor in monitor to predict the stable state of the monitored slope in the future or predict the failure time of the slope. At last, a comprehensive judging program is compiled to deal with all the monitor information to aid the decision-making of slopes' safety. The program is developed based on reasoning technique of expert system.As an example, the information and monitor data of Xiaolangdi Slopes have been researched in this paper.