Improving Non - Destructive Test Results Using Artificial Neural Networks

In the construction industry, non-destructive testing (NDT) methods are gaining more popularity for their ability to examine the in-situ component properties without damaging the structure. One of the most common NDTs for measuring the concrete compressive strength on site is the Rebound Hammer Test. Using the rebound value obtained from the test hammer, the concrete compressive strength can be estimated using the conversion chart provided by the instrument manufacturer. Despite for its convenience, rebound hammer test estimations have an average of over 20% mean absolute percentage error when comparing to the compressive strength obtained by destructive tests. In light of this, this research proposes an alternative approach to obtain the concrete compressive strength using the rebound value from the test hammer. That is, by applying the Artificial Neural Networks (ANNs) to develop a prediction model for concrete compressive strength estimation. Data collected from 838 lab Rebound Hammer tests are collected to train and validate the ANNs model. The ANNs model prediction results have successfully reduced the average mean absolute percentage error to 7.27%. It is recommended that Artificial Neural Networks can be applied to improve non-destructive test (rebound hammer test) results.