The life of asphalt concrete (AC) is significantly influenced by its resistance to fatigue. Traditional methods of fatigue life are time consuming and require a significant number of resources, while the previous applications of artificial neural networks (Ann) have not given a detailed description of their architectures, nor used extensive data files. Our study uses Ann to predict the fatigue life of AC, focusing on the level of transformation, binder content, gap and working with a data set of 245 test bodies. To increase the prediction capacity of Ann, we have optimized hyperparameters and used a medium quadratic logarithmic error as a loss function. Our results show the improved accuracy of the prediction of fatigue lifetime, with a special importance attributed to the content of the binder. This approach shifts the possibilities and transparency of Ann application when predicting fatigue life AC.
Keywords: asphalt concrete, fatigue life, binder content, gap, machine learning
The Fatigue Life of Asphalt Concrete (AC) is significantly influenced by its fatigue resistance. Traditional Methods for Fatigue Life Assessment Are Time-Consuming and Resource-Intensive and Previous Applications of Artificial Neuron Networks (Ann) Have Not Provided and Detailed Description of Their Architectures or Used Large DataSets. Ur study uses ann to the pre -fatigue life of ac, focusing on the level of strain, binder content, void content, and working with and dataset of 245 test specimens. This Incime The Predictive Ability of the Ann, We Optimized the HyperParameters and Used the Root Mean Square Logarithmic Error as Loss Function. OR Results Show Improved Accuracy of Fatigue Life Prediction, with special importance being attribted to the binder content. This approach advanceces the possibilities and transparency of the appllication of ann in the pre -preparation of the Fatigue Life of AC.
Keywords: Asphalt Concrete, Fatigue Life, Binder Content, Voids, Machine Learning
