نوع مقاله : مقاله پژوهشی
نویسندگان
1 استادیار، گروه علوم و فنون دریایی، دانشگاه آزاد اسلامی، واحد جویبار، ایران
2 استادیار گروه کامپیوتر، دانشگاه آزاد اسلامی، واحد جویبار
چکیده
کلیدواژهها
عنوان مقاله [English]
نویسندگان [English]
Beach cusps are rhythmic wave-shaped features that can usually be seen on the beach face. In this study, the morphological evolution observed in a multi-level cusp system was investigated using artificial neural network. The main results of this research indicate an excellent agreement between the prediction results of the cusp spacing by the neural network model with field observations. So that the results indicate that this model provides an effective tool for quick and accurate prediction of the cusp spacing at different levels of the swash zone, especially at the upper level of the beach face. The results also showed that the most accurate and reliable predictions It is achieved by the Cascade-forward back propagation (CFBP) with 30 neurons, and the lowest performance in predictions by the Elman back propagation (EBP) model with 5 neurons. Mean squared error (MSE) and correlation coefficient (R) values for training and testing in CFBP model with 30 neurons obtained respectively equal to 0.358, 0.376, 0.997 and 0.997, while the values obtained for the same criteria in EBP model with 5 neurons are equal to 0.477, 0.430, 0.996 and 0.989 Calculated. Other results of this research indicate that the Levinberg Marquardt function (lm) for different neural network models shows the most accuracy in different stages of testing, evaluation and training.
Beach cusps are rhythmic wave-shaped features that can usually be seen on the beach face. In this study, the morphological evolution observed in a multi-level cusp system was investigated using artificial neural network. The main results of this research indicate an excellent agreement between the prediction results of the cusp spacing by the neural network model with field observations. So that the results indicate that this model provides an effective tool for quick and accurate prediction of the cusp spacing at different levels of the swash zone, especially at the upper level of the beach face. The results also showed that the most accurate and reliable predictions It is achieved by the Cascade-forward back propagation (CFBP) with 30 neurons, and the lowest performance in predictions by the Elman back propagation (EBP) model with 5 neurons. Mean squared error (MSE) and correlation coefficient (R) values for training and testing in CFBP model with 30 neurons obtained respectively equal to 0.358, 0.376, 0.997 and 0.997, while the values obtained for the same criteria in EBP model with 5 neurons are equal to 0.477, 0.430, 0.996 and 0.989 Calculated. Other results of this research indicate that the Levinberg Marquardt function (lm) for different neural network models shows the most accuracy in different stages of testing, evaluation and training.
کلیدواژهها [English]
)