A deep learning model for predicting greenhouse film degra dation

Abstract

The behavior of LDPE films has been studied experimentally, as temperatures; UV rays and precipitation are accumulated in these films, which are responsible for the deterioration of greenhouses. Multilayer films have been subjected to the aging process in different ways, whether it is artificial over a specific period of time (such as multilayer Agro-film) or as a result of external factors. Based on the cost of these experiments in terms of time and the complexity of the factors to consider, we aim in this work to use computer simulations as a tool to study the aging process in these films. Deep learning-based AI has proven useful in a wide range of applications and research fields, as it can produce high-level graphical representations of large amounts of raw data, unlike traditional machine learning algorithms. Thus, it is an excellent solver for a wide range of linear problems and also shows high efficiency in solving nonlinear problems. The purpose of this study is to propose a methodology for designing a deep learning model for nonlinear analysis, and subsequently to develop a CNN model capable of dealing with this problem. Then it will be applied to solve real-world problems such as life prediction and mechanical performance degradation of multilayer polyethylene films in greenhouses. To carry out this work, we created a deep learning hybrid model that is able to meet this challenge, and the training and validation phases of the model were successful, indicating the feasibility of using the D1 CNN model for nonlinear analysis