The application of Machine Learning (ML) in materials science has expanded rapidly, it offers powerful tools for property prediction, materials discovery, and process optimization. Despite this, several challenges continue to limit complete integration of ML into materials workflows. This chapter aims to examine the key ML methods used in materials science, including supervised, unsupervised, reinforcement learning, and hybrid physics-informed models. It outlines their application across domains such as structural health monitoring, alloy design, and simulation-augmented materials development. Focus is given to recent challenges such as data scarcity, multi-modal integration, model interpret-ability, and reproducibility. Integration with experimental methods and physical models persists to be a complex task due to the debatable nature of materials behavior and inconsistencies in data sources. Through thorough research and analysis of emerging techniques, this chapter highlights both limitations and practical advances. Addressing these issues through explainable AI, federated learning, and benchmark development is expected to improve the reliability and scalability of ML applications in materials science.