Utilizing the Novel developed MLP Techniques to Survey Pile Subsidence via Optimization Algorithms

Document Type : Original Article


1 The King's School, Bujumbura BP1560, Burundi

2 Central Arizona College, Coolidge 85128, AZ, United States


The Pile settlement (PS) is one of the most essential issues in designing piles and its foundation type applied in real state. Over the variants in designing the pile penetrated in rock, the vertical settlement is of paramount importance to know. However, rigorous theoretical descriptions for soil-pile interactions are still ambiguous. In this regard, most research has tried to figure out the subsidence rate in piles after loading overtime via artificial intelligence methods. The Artificial Neural Network, as a widespread method, has absorbed attention to draw the actual picture of pile movement vertically during the loading period. This research aims to develop the Multilayer Perceptron coupled with the Novel Arithmetic Optimization Algorithm and Biogeography-Based Optimization) to find out the optimal number of hidden layers of neurons within MLP. The Klang Valley Mass Rapid Transit network built in Kuala Lumpur, Malaysia, was chosen to test the piles' settlement and earth properties algorithms. In the prediction process, the R2 value of MLP-AOA and MLP-BBO were obtained at 0.93 and 0.94, respectively. The measured range of piles movement was from 4.5 to 20 centimeters, which predicted settlements showed us an average one percent change compared to measured magnitudes.