Deep machine learning in bridge structures durability analysis

Abstract

According to Eurocode 0 structural durability is next to ultimate and serviceability one of the basic criteria in the structural design process. This article discusses the subject of concrete cracks observation in bridge structures, as one of the factors determining their durability. The durability of bridge structures is important due to both social, economic aspects and also the defense aspects of countries. Cracking of the reinforced concrete structures is a natural effect in concrete. The aim in the design and construction of structures is not to prevent the formation of cracks, but to limit their width to acceptable values. At the same time, there is a need for structure tests that allow for non-contact, fast measurements and algorithms that allow for efficient analysis of large amounts of measurement data. Deep machine learning algorithms can be used here. They can be used to analyse data which are acquired by means of photogrammetric methods (especially helpful during construction to inventory concealed works). Moreover, they can also be applied to standard data acquisition methods, consisting in photographing objects damage during works acceptance or periodic inspections. This paper discusses the application of deep machine learning to assess the condition of bridge structures based on photographs of object damage. The use of this method makes it possible to observe the rate and extent of damage development. Consequently, this method makes it possible to predict the development of damage in time and space in order to prevent failures and take structures out of service.