An Effective Hybrid Approach for Processing Deformation Monitoring Data

Abstract

This paper describes the procedure of a hybrid approach based on grey model and artificial neural network (GM&ANN) to analysis and forecast of deformation data. The GM&ANN is formulated into three steps:(1)according to the monotonously increasing characteristics and the nonlinear characteristics of deformation time series, total deformation can be divided into tendency part and stochastic part.(2) use GM(1,1)to fit the trend of the data and obtain the residual series, on this basis by using artificial neural network to fit the stochastic part (residual series) .Then the forecasting value of deformation is obtained by adding the calculated predictive displacement value of each sub-stack. (3) validate the model. The results of experiments show that this hybrid has higher performances not only on model fitting but also on forecasting and therefore can be applied to deformation data processing.

Study on the Applications of Neural Networks for Processing Deformation Monitoring Data

“…BP algorithm which is an iterative gradient descent algorithm is a simple way to train multilayer feed forward neural networks. In a BP network, there are typically three layers including input layer, output layer and hidden layer [2][3][4][5]. Hidden layer can have one layer and can also be several layers.…”

“…Instead they let data speak for themselves and have the capability to identify the under-lying functional relationship among the data. But most importantly, the ANN can tolerating the presence of chaotic components and thus is better than most methods [2][3][4][5]. In this paper, three prediction models including the BP, RBF and GRNN neural network prediction models were established for deformation data processing.…”

Grey Forecasting Model Refining in Deformation Prediction Based on Semi-Parametric Regression

“…The analysis and prediction results of deformation data is a main basis for decision making, and its quality can directly influence the effect of the whole monitoring work. Hence, it is necessary to research the deformation data processing theory [1][2][3][4].…”

“…As for short data sequences from which little information are provided, certain methods including statistic forecast method are difficult to be conducted [1,2]. To overcome this problem, GM(1,1) model has been widely employed in deformation monitoring and achieved satisfactory results [1][2][3][4],but one problem is that the grey model is constructed by exponential function, as discussed in many literatures, the curve-fitting effects in random time series is not good enough [3][4][5].…”

“…In order to increase the accuracy of GM(1,1) model, numerous scholars suggested various methods for improvement, such as correcting the residuals, improvements of grey derivative, optimization of initial condition for time response function, and constructing of background value [1][2][3][4][5][6][7] etc. All the above approaches may make GM (1, 1) model more adaptive and precise in practical applications.…”