The concept of predictive maintenance has gained popularity in the field of intelligent and sustainable manufacturing because of its potential to increase equipment uptime, decrease unplanned downtime, and better leverage available resources. In order to improve the efficacy of predictive maintenance in intelligent manufacturing systems, this paper presents an improved XGBoost algorithm linked with big data analytics. First, in the proposed framework, massive datasets are gathered and preprocessed from sensors, IoT devices, and other sources in the industrial setting. Then, the meta-algorithm XGBoost is used to improve the efficiency of subpar learners, allowing for reliable failure and deterioration prediction in machinery. Adjusting hyperparameters like the number of iterations and the learning rate is part of the algorithmic optimization process to strike a good balance between model accuracy and computational efficiency. The proposed model gains the accuracy level of 0.972 value, Precision level of 0.977 value, Recall level of 0.972 value and F1-score level of 0.974 value. Optimized XGBoost-enabled predictive maintenance framework offers a scalable, efficient, and intelligent method for handling the complexity of big data analytics in the context of manufacturing. The framework provides a viable path forward for enhancing predictive maintenance tactics, contributing to intelligent manufacturing practices, and supporting sustainability in industrial operations by capitalizing on XGBoost's capabilities and applying optimization methodologies.