A Hybrid Intelligent Model Combining SOM & CV-MLP to Predict Blast-Induced Ground Vibration

Research areas:

• Volume 3,Issue 4, Jan. 2015

Year:

2015

Type of Publication:

Article

Keywords:

Vibration, Prediction, Hybrid Intelligent Model, Complex Valued Multi-Layer Perception, Self Organizing Map

Authors:

Amin Zadeh Shirazi; Seyyed Javad Seyyed Mahdavi

Journal:

IJAIM

Volume:

3

Number:

4

Pages:

137-145

Month:

Jan.-Feb.

ISSN:

2320-5121

Abstract:

Blast-induced ground vibrations motivated the researchers to focus on controlling and predicting the vibration rate. This paper suggests a hybrid model based on unsupervised and supervised learning techniques i.e. Self- Organizing Map (SOM) and Complex-Valued Multilayer Perceptron Neural Network (CV-MLP) for predicting blastinduced ground vibration. The dataset used in this paper consists of 102 samples with 8 features. The proposed model is used for the first time and can be categorized in two sections. In the first section, considering the input samples fluctuations and nonlinear behavior of them, self-organizing map technique has been used to cluster the samples with the most similarity. Then, in the second part, the samples allocated to each cluster have been applied to complex-valued multilayer perceptron neural network and dealt with to predict the vibration rate. Three criteria of coefficient of determination (R2), Root Mean Square Error (RMSE), and Mean Absolute Error (MAE) were adopted to evaluate the proposed hybrid method and their values were obtained as much as 0.946, 2.183 and 1.587, respectively.

Full text: IJAIM-386_final.pdf