Use of recurrent neural networks to predict vibration load indicators

The development of intelligent systems for monitoring vibration impacts from transport on infrastructure facilities requires new approaches to analyzing time series obtained using field measurements. An actual direction is the prediction of vibration characteristics using recurrent neural networks, in particular LSTM (Long Short-Term Memory) networks. One of the advantages of the LSTM network is to overcome the problem of gradient attenuation which allows it to efficiently process long sequences of data and store information for a long time. The purpose of the article is to develop an approach to predicting vibration characteristics of an approach to predicting vibration data based on the architecture of recurrent neural networks belonging to the LSTM type, as well as to analyze the possibilities of using the LSTM module as part of a software package that provides assessment and prediction of vibration impacts on urban infrastructure. The article describes the principles of functioning of LSTM models in the framework of time series analysis. The stages of time series preprocessing including data filtering, normalization, and segmentation have been considered. The architecture of a neural network model based on STM has been proposed and a machine learning technique for predicting peak vibration values has been considered. The components of the LSTM module have been described and the analysis of the possibilities of integrating its model into the Vibcalc software package designed for modeling and evaluating the vibrational effects of traffic flows on urban infrastructure facilities has been performed. The use of systems for visualizing the results of forecasting and automatic generation of warnings when the permissible levels of vibration acceleration are exceeded has been proposed. The results can be used in practice to solve problems of forecasting vibration levels, to develop and improve intelligent monitoring systems and decision support systems for the design and construction of transport facilities.