Návrh metod detekce varovných stavů pro geotechnické monitorovací systémy na bázi Internetu věcí

Abstract

The dissertation focuses on designing methods to detect warning states in geotechnical retaining fences using IoT-based monitoring systems. The introduction discusses the current state of geotechnical monitoring and retaining fence condition monitoring and includes an analysis of data classification possibilities from accelerometric sensors. The main part of the dissertation focuses on monitoring retaining fence conditions using accelerometric sensors and wireless sensor networks. A comprehensive monitoring system was designed, implemented, and tested, consisting of three sub-parts: developing a wireless sensor module for monitoring fence conditions, designing a data collection and transmission method, and creating a server part. A custom sensor module was also designed for mounting on retaining fences using IQRF® wireless technology. The collected data was analysed, processed, and evaluated using designed procedures and methods, including machine learning methods. Two methods for classifying measured data using convolutional neural networks were proposed, and the resulting data on retaining fence conditions were visualized in the Grafana system. The system was tested in two existent locations and a physical model of a retaining fence, proving its functionality. A unique system for geotechnical monitoring of retaining fences was created using wireless IoT technologies and machine learning methods for event detection.