Purpose: Pores are the primary cause of fatigue failure in laser-directed energy deposition (L-DED) titanium alloys, which are largely determined by their location, size and shape. It is crucial for promoting the application of L-DED titanium alloys and ensuring their safety that establishing a fatigue life prediction method induced by pores, resulting in a proposed fatigue life prediction framework for L-DED Ti-6Al-4V based on a physics-informed neural network (PINN) algorithm.Design/methodology/approach: In this study, a novel fatigue life prediction framework for L-DED Ti-6Al-4V based on a PINN algorithm was proposed. The influence patterns of various fatigue-sensitive parameters were revealed. The paper also included validation and analysis of the method, such as hyperparameter analysis of the PINN, efficacy analysis driven by physical information and comparative analysis of different methods.Findings: The proposed method demonstrated high accuracy, with a correlation coefficient of 0.99 with experimental life. The coefficient of determination was 0.95 and the mean squared error was 0.06.Originality/value: The results indicate that the proposed fatigue life prediction framework was of strong generalization capability and robustness.

Purpose: The purpose of this review is to evaluate the transformative potential of machine learning (ML) in predicting fatigue behavior. The review seeks to highlight current progress, practical applications, and emerging challenges, offering a roadmap for leveraging ML to improve reliability, accuracy, and efficiency in fatigue-related engineering analyses. Design/methodology/approach: This review systematically examines the application of machine learning (ML) techniques to fatigue behavior prediction, encompassing material properties, fatigue life, and crack growth. It employs a comprehensive survey of recent advancements, focusing on data preprocessing, regression methods, deep learning architectures, and hybrid approaches that integrate physics-based models with ML. Multiscale modeling and ensemble techniques are also analyzed for their potential to enhance prediction accuracy and reliability. Findings: The review reveals that ML techniques significantly enhance fatigue behavior prediction by addressing complex, multiscale, and nonlinear characteristics of materials. Case studies demonstrate successful ML applications in aerospace and civil engineering, underscoring its practical value. However, challenges such as limited data quality, model interpretability, and computational scalability persist, necessitating further innovation and interdisciplinary collaboration to fully realize ML’s potential. Originality/value: This review provides a comprehensive and up-to-date analysis of machine learning (ML) applications in fatigue behavior prediction, addressing material properties, fatigue life, and crack growth. It bridges the gap between traditional fatigue analysis methods and data-driven approaches, emphasizing the integration of physics-based insights with ML for enhanced accuracy and reliability.