Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Long Term Behavior Monitoring of An Arch Dam Using Adaptive Neuro-Fuzzy Model Technique

Download as ppt, pdf, or txt
Download as ppt, pdf, or txt
You are on page 1of 1

2nd International Conference on Long Term Behaviour of Dams 12th-13th October 2009-Graz, Austria

Long Term Behavior Monitoring of an Arch Dam Using Adaptive Neuro-Fuzzy Model Technique
Uur AHN1 and Seyfullah DEMRKAYA2 1stanbul Technical University, Department of Mathematical Engineering, Istanbul, Turkey 2Yildiz Technical University, School of Vocational Studies, Istanbul-Turkey E-mail: demirkay@yildiz.edu.tr T-S (Takagi-Sugeno) type Fuzzy System
The structural health and reliable functioning of a dam requires good understanding of causative factors and the mechanism of deformations. This can be achieved only through proper monitoring and analysis of investigated dam. Because of the strongly non-linear, high degree of uncertainty and time-varying characteristics of the structural behavior, none of a wide variety of approaches has been proposed for deformation prediction can be considered as a single superior model. Recently, ANN (Artificial Neural Networks) have been accepted as a potentially useful tool for modeling complex non-linear systems and widely used for displacement of an arch dam forecasting on dam bodies. Besides to ANN, Fuzzy Inference System (FIS) to describe complicated systems has become very popular and been successfully used in various engineering problems. For that reason, we propose the use of the ANFIS method to self-organize network structure and to adapt parameters of the fuzzy inference system to forecast the displacements s crest. This method is data-driven; it deduces the model directly from the data. Also, it is based on a supervised learning algorithm to optimize the parameters of a FIS.

Model Construction
It is possible to estimate the displacements from a given water level and temperature values. Gaussian-type MFs were used, so as to run T-S fuzzy inference system in this study. The 2555 data set (daily, 7 years) are divided into three independent subsets: the training, verification, and the testing subsets. The training subset includes 1095 data (3 years); the verification subset has 1095 data (3 years); while the testing subset has remaining 365 data (1 year). One of the most important tasks in developing a forecasting model is the selection of the input variables which determines the model. The input variables are consisted of: WL= Water level of the reservoir, UT, and DT are the values of the thermometer embedded in upstream and downstream face and MT is in the middle of the dam, respectively. AT is the air temperature near vicinity of the dam. The measured horizontal displacement values are showed with P. The MFs for input variables is shown in Fig:1. All of the input MFs has four subsets and named suitable as QL: Quite Low, L: Low, M: Medium and H: High. In this study, our aim is to demonstrate that the ANFIS has the ability to deal with expert knowledge and enhance the model performance. In the beginning, we developed two ANFIS models named of Model_1 and Model_2 based on the number of MF for displacement forecasting. They have 4 and 6 MF. Lastly, the Model_3 is being obtained from divided the data set to year based models and has 4 MF. The training, verification and testing data sets had been selected randomly. The rules related to the proposed model of the displacement forecasting can be given as follows in the Rule Base:

It also shows that the forecasting Model_3 results in a much lower value of the RMSE, MAE and higher value of the CC than Model_1 and 2. These results might also suggests that the ANFIS has a great ability to learn from input-output patterns, which represent the water level and temperatures are lumped effects on displacements of dams crest. Overall, the performance of three ANFIS models is very good. The results demonstrate that the ANFIS can be successfully applied to establish the forecasting models that could provide accurate and reliable daily horizontal displacement prediction of the Schlegeis Arch Dams crest.
Table 1: Evaluation the Performance of the models with training and testing data.

Z ( x)

w z
l 1 M

l l

w
l 1

Figure 2: Forecasts of Crest Displacements for Training Data.

Data Base

Rule Base

Fuzzification

Defuzzification

The advantage of this fuzzy logic system (T-S) is that it provides a compact system and, therefore, parameter estimation methods such as neuro-fuzzy algorithms or neuro-adaptive learning techniques can be developed to estimate the parameters. These techniques provide a method for the fuzzy modeling procedure to learn information about a data set, in order to compute the membership function parameters that best allow the associated fuzzy inference system to track the given input/output data. In this study, the well known adaptive algorithm called ANFIS with the aid of Matlab Fuzzy Logic Toolbox is used.

Figure 3: Forecasts of Crest Displacements for Testing Data.

Fig. 2 and Fig. 3 nicely show that models performances are accurate and Model_3 is consistently superior to the others in training (training+verification) and testing phases. To get a brief picture of the general performance of the constructed model, we also provide the pendulum records of the crest displacements versus water levels for training and testing data, in Fig. 4 and Fig. 5, respectively.

Output Data

Artificial Neural Network


Difficult to use prior rule knowledge Learning from scratch Black box Complicated learning algorithms Difficult to extract knowledge

Input Data

Fuzzy Inference System

Data Used In the Analysis


We used the data from the Theme C of the 6th ICOLD Benchmark Workshop on Numerical Analysis of Dams (2001) was dedicated to the interpretation and a subsequent prediction of the crest displacements of Schlegeis arch dam. The observed radial crest displacements of dam are analyzed using the time histories of water level and concrete temperatures as input parameters. The data for this study are the water level, the air temperature and the concrete temperatures at 6 points-one value per day. The response value which has to be interpreted is the radial crest displacement of the central cross section. This crest displacement is measured by pendulums. All of these data are related to the period 1992 to 1998.

Fuzzy Inference System


Prior rule-base can be incorporated Cannot learn (linguistic knowledge) Interpretable (if-then rules) Simple interpretation and implementation Knowledge must be available

Figure 1: The Membership Functions for the Input Variables.

Rule Base
Figure 4: Crest Displacements versus water level for training data. Figure 5: Crest Displacements versus water levels for testing data.

Neuro-Fuzzy Modeling
The neuro-fuzzy modeling refers to the way of applying various learning techniques developed in the neural network literature to fuzzy modeling or to a Fuzzy Inference System (FIS). The basic structure of a FIS consists of three conceptual components: a rule-base, which contains a selection of fuzzy rules; a data-base which defines the membership functions (MF) used in the fuzzy rules; and a reasoning mechanism, which performs the inference procedure upon the rules to derive an output. FIS implements a nonlinear mapping from its input space to the output space. This mapping is accomplished by a number of fuzzy if-then rules. The parameters of the if-then rules (antecedents or premises in fuzzy modeling) define a fuzzy region of the input space, and the output parameters (also consequents in fuzzy modeling) specify the corresponding output. Hence, the efficiency of the FIS depends on the estimated parameters. In the presented study, the concept of the adaptive network, which is a generalization of the common back-propagation neural network, is employed to tackle the parameter identification problem in a FIS. This procedure of developing a FIS using the framework of adaptive neural networks is called an Adaptive Neuro Fuzzy Inference system (ANFIS).

Conclusion

Evaluation of the Model Performance


Schlegeis Dam (Austria)

Because of there are no fixed rules to develop an ANFIS, we followed a way based on trials and found that the data set 1992, 1996 and 1997 are used for training whilst that 1993, 1994 and 1995 are used for verification. Lastly, 1998 data set is used for testing. This ANFIS models are compared based on their performance in training and testing sets. The results are summarized in Table 1. It appears that the ANFIS models are accurate and consistent in different subsets, where all the values of the Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) are smaller and all Correlation Coefficients (CC) are also very close to unity.

IF WL
Acknowledgements
We would like to thank to Prof. Dr. E. BAUER, Prof. Dr. G. GENZ and Mr. F. PERNER and also their valuable working friends for providing data. In addition, the authors are indebted to the reviewers and auditions of the LTBD_09 conference for their valuable comments and suggestions.

The ANFIS models have been provided accurate and reliable displacement forecasting, where the correlation coefficients (CC) are very close unity. Constructed ANFIS models, through the subtractive fuzzy clustering, can efficiently deal with vast and complex inputoutput patterns, and has a great ability to learn and build up a neuro-fuzzy inference system for prediction, and the forecasting results provide a useful guidance for deformation analysis studies. Investigations should be carried out using the ANFIS in combination with results from structural and/or statistical analysis.

Selected References

http://nw-ialad.uibk.ac.at/Wp3/Tg4/Se2/Files/6_BW_C.pdf Demirkaya, S., Sahin, U. (2008). ANFIS approach for modeling horizontal displacements of an arch dam, In: Topcu, I. B. et al (eds), Proc. of The Symposium for Modern Methods in Science, Eskisehir, Turkey, 2008, pp. 345-356 (in Turkish). Jang, J.S.R.(1993). ANFIS: Adaptive-Network-based Fuzzy Inference Systems, IEEE Transaction Systems, Man, and Cybernetics, vol. 23, pp. 665-685, May 1993.

You might also like