Application of Artificial Neural Networks as Design Tool for Hot Mix Asphalt

Reducing lab work to save time and costs has become one of the challenges which need to be taken on by researchers, for this reason, and a lot of studies have been performed to achieve this aim. Many approaches were developed for developing these models such as statistical, mathematical, mechanistic and so on. In the last decades, an advanced approach simulating the human neural system has opened the door for researchers to do what other approaches have failed with. One of these advanced approaches is called artificial neural networks (ANN) as one of so-called artificial intelligence (AI). The purpose of this study is to introduce ANN as a design tool for hot mix asphalt (HMA) based on actual data for job mix formula; achieving these points will contribute to reducing the time, cost, and data in the design of hot mixes asphalt. Data of 252 mixes were collected from different highway agencies in Iraq, all these mixes were designed according to the Marshall method. The data of 170 mixes were used for developing the ANN model, while the data of 82 mixes were discarded due to the shortage of information. Ten parameters (penetration, kinematic viscosity, surface area of aggregate, mechanical abrasion, and binder content) were extracted from the data of 170 mixes to be used as input and outputs variables for training the ANN. The results showed the visibility of ANN as a tool for designing the HMA as shown from the high correlation and quality analysis.