Seite - 905 - in Book of Full Papers - Symposium Hydro Engineering

1. INTRODUCTION Dam safety assessment has received much attention since the end of the last century. The safety assessment of a gravity dam requires a wide range of information that is acquired from monitoring systems. Usually, there are many instruments equipped in the dam and its surroundings for monitoring the water level, temperature, deformation and other aspects. Now, with the progress of monitoring technology, the current measurement technology has advantages of high precision, good stability and strong sensitivity. But the response of dam structural behavior is the result of multi-factor synergies. So it’s necessary to extract the main factors, which influence the dam performance, and in the meantime analyze their development trend. The positive analysis models are a fundamental component of dam safety systems. They provide an estimate of dam response faced with a given load combination. Then the calculated value can be compared with the actual measurements to draw conclusions about dam safety. The statistical models based on monitoring data have been used for decades for this purpose since 1955. In particular, the hydrostatic- season-time models are fully implemented in engineering practice. In recent years, powerful tools such as neural networks are used by some scholars to analyze the observed data for interpreting the complex systems. But the multicollinearity issue among the components will influence the generalization ability and prediction accuracy of the model. In this paper, a monitoring model based on principal component analysis (PCA) and radial basis function (RBF) neural network is put forward to analyze the displacement trend of the concrete dam. The principal components of the displacement monitoring data of the dam is extracted and reconstructed by PCA. On the basis, the method of the RBF neural network is used to predict the displacement trend of dam body. 2. PCA-RBF NEURAL NETWORK During the operation period, usually, there’s a large amount of dam monitoring data accumulated, including displacement, temperature, water level and so on. The large amount monitoring data provides a good basis to analyze the dam behavior and is very important in the diagnosis. But massive data can also be a problem for analysis. It’s necessary to separate the useful information from observation and find out the main factors that affect the dam performance. The PCA is used to reduce the dimensionality of m components including hydraulic components, temperature components and irreversible components. Then the 905