Hybrid metaheuristic optimization algorithm for prediction of fatigue life performance of fiber-reinforced concrete

Abstract

This paper deals with an integrated and multi-scale hybrid intelligence framework for optimizing fiber orientation, improving crack control, and predicting the fatigue life of FRC. At the core of such model construction is a hybrid evolutionary-physics informed neural network (HE-PINN), which employs evolutionary optimization and physics Informed learning to refine fiber orientation on the basis of outcome material properties and cyclic loading parameters in process, thus ensuring physical prediction of fiber-matrix interaction under stress redistribution. Along with this construction, the adaptive fractal dimension based metaheuristic is introduced to understand and monitor real-time image-based crack evolution in real-time; this will use fractal geometry to control the microcrack growth toward localization while dissipating energy within concrete-satisfied limits. the stress wave propagation-based adaptive swarm optimization (SWP-ASO) constructs a fusion of wavelet-transformed stress features that come from finite element simulations for load redistribution optimization function of fiber placements. It thus advances a data-driven generative designing scheme eliminating the human bias and training itself to learn high-performance reinforcement layouts through the generative adversarial network for fiber network optimization (GAN-FNO). Finally, an adaptive Bayesian-Gaussian process regression (AB-GPR) module provides real-time fatigue life prediction with uncertainty quantification and an adaptive-learning process. This combined architecture therefore provides an improvement of between 30 and 40% in fatigue life, an increase of up to 50% in energy absorption, and a reduction of up to 35% in the rates of crack propagation, presenting a considerable advancement in predictive and prescriptive modeling of smart FRC designs.