The main objective of this study is to simulate the fatigue limits properties of a high-strength unidirectional carbon/epoxy ply. The focus is on reconstructing the fatigue characteristics of a carbon/epoxy elementary ply using advanced approaches based on Physics-Informed Neural Networks. In this context, artificial neural networks were developed to represent experimental data while incorporating heat transfert models that describe the self-heating phenomenon in laminated composite materials under cyclic loading. Classical fatigue tests were used to construct Wöhler curves, which represent the relationship between the number of cycles to failure and the applied load amplitude. These curves provide a comprehensive view of the performance of laminated composite materials under different loading regimes. In parallel, the fatigue limits determined from self-heating tests were compared with those obtained from classical tests. The results show a strong correlation between the two methods, validating the self-heating-based approach as a fast and effective alternative for evaluating the durability of composites. Identifying the fatigue limits and parameters of a PINN involves training the neural network on a dataset to represent all experimental data within the identification domain. In the case of PINNs associated with experimental data, an inverse analysis is performed to determine the neural network and the parameters corresponding to the solution of the differential equation or partial differential equation derived from physics. In our case, this is the heat equation and, consequently, the conduction properties of the laminated specimen. After optimizing the model parameters and those of the neural network, the network will reproduce the experimental data and provide approximations outside the identified points. Unlike traditional numerical methods such as finite element analysis, the domain is not meshed. These neural networks, known as Physics-Informed Neural Networks (PINNs), stand out for their ability to integrate physical laws into the learning process, thereby ensuring greater accuracy and enhanced robustness of predictions. Various composite laminates were tested to better understand the influence of laminated architecture on fatigue behavior and fatigue limits. In conclusion, the self-heating method is a valuable tool for engineers and researchers seeking to quickly assess the fatigue performance of materials while saving time and resources. Moreover, the use of physics-informed neural networks represents a significant advancement in the field of modeling and predicting fatigue behavior, enabling a reduction in the number of experimental tests required and accelerating the design and validation processes for composite structures.