Philippine E-Journals

Explosion Locator using Artificial Neural Network

Leah A. Alindayo | Marven E. Jabian

Discipline: Computer Engineering

 

Abstract:

The Explosion Locator using Artificial Neural Network (ANN) has emerged as an important research area because it can increase situational awareness in different scenarios. Network of sensors with artificial intelligence such as artificial neural network shows a promising approach towards efficient system response. In this study, network of sensors was carefully placed in a location to efficiently gather needed data. The data gathered from these three sensors were fed to ANN for training. A two-layer feed-forward back-propagation neural networks were designed to implement the functional relationship. A training algorithm based on Levenberg-Marquardt was used. During the training, important parameters such as number of epochs, network weights and biases, number of hidden neurons, number of vectors, number of inputs and training algorithm were varied. The training stopped at cross validation and test error increased for 30 iterations, which occurred at iteration 36. The fit is almost perfect for train, testing and cross-validation data over 0.9999 for the total response of accuracy through % error with the maximum of 0.067731 were achieved. The result showed that the network is trained. Lastly, it confirmed the superiority of feed-forward back-propagation with trainlm architecture with a low MSE values.