Prediction of the Composite Required in the Design of a Toroidal Vessel Using an Artificial Neural Network

Abstract

Context: Within the design of toroidal vessels, minimizing the amount of material is very important for reducing production costs; Conventional methods used to minimize the amount of material are computationally time consuming each time a new container is tested. New techniques based on artificial intelligence require the prediction of the minimum amount of material required for the design of a container in the shortest possible time. Method: The prediction methodology is based on a linear regression model through an artificial neural network, which is implemented with the Keras model of Python; In its first phase, a data set created by a script with ANSYS APDL code is handled. Results: An artificial neural network model that learned to predict the minimum amount of material, adequate accuracy and loss of the model. Conclusions: A better performance of the model was obtained, the partition of the dataset into training and testing data, a level of precision was obtained that ensures the reliability of the machine learning model compared to the traditional ones.