Designing an intelligent model to optimize the safety risk of the takeoff flight using BIM-LSTM

This article presents a new model for optimizing the safety risk of take-off, as the most important and dangerous flight process, using a combination of BI algorithm and recurrent neural network LSTM. The goal is to train an effective neural network with past data records of air accidents to predict safety risk parameters. For this purpose, 17 safety features, such as weather conditions, aircraft configuration and preparation, flight information and air traffic were obtained. The data related to 2019 to 2020 was selected after performing exploration, summarization, cleaning, normalization operations with 28813 data records. Due to the dependence of flight data on their previous inputs and the need for a kind of memory, training was performed by deep learning algorithm (LSTM) in Python environment. After learning, the learning error was about 6 percent and the mean square error was about 116/0. It shows that the error percentage is negligible and the proposed model has high validity. Also, this model solved the problem of exploration and cleaning of bulk flight data by having advanced tools such as ETL, metadata and real-time monitoring and was able to predict the most important safety risk factor (speed V1) with high accuracy. This pattern helps the flight service in controlling the important parameters of safety risk, such as the speed of aircraft taking off from the runway, controlling the safe take-off speed and most importantly controlling the loss of flight with a reliable strategy.