Automated detection of alkali-silica reaction in concrete using linear array ultrasound data

Abstract

This paper documents the development of signal processing and machine learning techniques for the detection of Alkali-silica reaction (ASR). ASR is a chemical reaction in either concrete or mortar between hydroxyl ions of the alkalis from hydraulic cement, and certain siliceous minerals present in some aggregates. The reaction product, an alkali-silica gel, is hygroscopic having a tendency to absorb water and swell, which under certain circumstances, leads to abnormal expansion and cracking of the concrete. This phenomenon affects the durability and performance of concrete cause significant loss of mechanical properties. Developing reliable methods and tools that can evaluate the degree of the ASR damage in existing structures, so that informed decisions can be made toward mitigating ASR progression and damage, is important to the long-term operation of nuclear power plants especially if licenses are extended beyond 60 years. The paper examines the differences in the time-domain and frequency-domain signals of healthy and ASR-damaged specimens. More precisely, we explore the use of the Fast Fourier Transform to observe unique features of ASR damaged specimens and an automated method based on Neural Networks to determine the extent of ASR damage in laboratory concrete specimens.