The Application of Deep Learning Method in the Management of Environmental Risks of Construction Projects with the View of Passive Defense

The construction industry in Iran is considered one of the leading sectors in the economy and enjoys significant popularity among the people. This industry has also been a top choice for investment and has shown high returns based on published reports. Additionally, environmental risks associated with this industry have been of utmost importance, especially in recent years, for relevant authorities. This article focuses on environmental risk management in civil projects using the highly successful FMEA method and employing machine learning techniques. The main innovation of this research lies in simultaneously utilizing a novel machine learning method and risk management with the help of the FMEA method. The hybrid approach incorporates both descriptive and numerical modeling. Over 10 major projects in the metropolitan area of Tehran have been evaluated in this study, identifying more than 20 environmental risks. The prerequisite modeling using deep learning with FMEA was prepared, followed by modeling using a combination of artificial neural network and the MVO optimization algorithm. This study conducted extensive preprocessing to select the type of neural network and optimization algorithm. The research results demonstrate the effectiveness of the combined method employed. Four different types of neural networks and two optimization algorithms have been utilized based on specific evaluation criteria, with the most successful type being introduced. Finally, sensitivity analysis for identified environmental risks on the successful modeling has been performed, and the most significant risks have been identified using relative and absolute sensitivity methods.