پیش‌بینی فاصله بین کومه‌های هلالی چندترازه با استفاده از شبکه‌های عصبی مصنوعی

کومه‌های هلالی، عوارض موجی شکل منظمی هستند که معمولاً در وجه ساحل، قابل رویتند. در این مطالعه به پیش‌بینی فواصل بین کومه‌های هلالی چند سطحی با استفاده از شبکه عصبی مصنوعی پرداخته شد. نتایج اصلی این تحقیق حاکی از مطابقت عالی بین نتایج پیش‌بینی فواصل بین کومه‌ها توسط مدل‌ بهینه‌ شبکه عصبی با مشاهدات میدانی است. به طوریکه نتایج حاکی از آن است که این مدل‌ ابزار موثری را برای پیش‌بینی سریع و دقیق فواصل کومه‌ها در ترازهای مختلف منطقه شستشو به خصوص در تراز بالایی وجه ساحل فراهم می‌کند .نتایج همچنین نشان داد که دقت بدست آمده از نتایج شبکه عصبی با اختلاف جزیی به ترتیب در مدل‌های پس‌انتشار المان، پس‌انتشار پیشرو و پس‌انتشار آبشاری پیشرو، بیشترین به کمترین مقدار است. نتایج دیگر این تحقیق حاکی از آن است توابع، بسته به نوع مدل انتخاب شده دقت متفاوتی در نتایج پیش‌بینی شده، در مراحل مختلف آزمون، ارزیابی و آموزش نشان می‌دهند.

prediction of the multi level beach cusp spacing using artificial neural networks

beach cusps are rhythmic wave shaped features usually observable on the beach face. these features in the swash zone are so variable in terms of space and time due to wave attacks and tides. in this study, an artificial neural network was used to fully understand the behavior of multi level beach cusps on the beach face. a neural network is a soft computing method for solving problems as an intelligent system that can learn, remember, and create relationships between different data. in this research, the parameters related to the beach cusps were recorded as the input of the neural network model, including the cusp amplitude, cusp elevation, and cusp depth in the lower, middle, and upper levels of the beach, as well as the cusps spacing as the output of the model. to achieve the goal of this research to predict the cusps spacing, the performance of three back propagation neural network models was investigated in different functions and neurons. then, relevant statistical criteria were calculated and compared at each stage. back propagation learning is an iterative search process that adjusts the weights from the output layer to the input layer in each run until no further improvement in the error value is found. the main results of this research indicate an excellent agreement between the results of the neural network model and the recorded values of the cusps spacing in the field observations. the comparison of the scatter plots related to the values of the predicted spacings of beach cusps against the values of the observed spacings in different parts of the swash zone on the beach face indicates that the accuracy of the results of the neural network in the upper part of the beach face is the highest. these results are perfectly consistent with other researchers’ results who introduced the lower part affected by tides. the evaluation of the graphs related to the trend of statistical criteria changes in different neural network models in the whole simulation indicates that the accuracy of neural network results is the highest to the lowest in elman back propagation (elman bp), feed forward back propagation, and cfbp, respectively. scatter plots related to the predicted spacings of cusps by the cascade forward back propagation (cfbp) model (lm optimum function) against other parameters of beach cusps show that there is a strong correlation between the cusps spacing with all the morphological parameters of the beach cusps system, especially the cusp elevation on the beach face. another result of this research using a comparison of the radar charts related to different statistical criteria in the different stages indicates that the functions provide various accuracy in the predicted results depending on the type of selected models in the different stages of the testing, evaluation, and training.