The existing methods of ship inspection and maintenance rely on manual operation and empirical judgment, which have problems of low efficiency and difficulty in precise prediction. This article adopts an intelligent ship inspection and maintenance system based on deep learning technology, aiming to improve detection precision and maintenance predictability, thereby optimizing ship operation efficiency. Firstly, the system utilizes the image identification technology and convolutional neural network (CNN) models to automatically identify defects such as cracks and corrosion on the surface of the ship bodies. Secondly, by analyzing sensor data through long short-term memory (LSTM) networks, potential equipment failures are predicted, and maintenance schedules can be optimized, thereby reducing unexpected shutdowns. Meanwhile, multi-source data fusion and ensemble learning strategies are adopted to further enhance the reliability and prediction precision of the model. The results show that the system performs significantly in multiple key areas: crack detection precision of 91.5% and recall rate of 94.0%; corrosion detection precision of 90.8% and recall rate of 89.5%. The research shows that the application of deep learning technology provides an innovative method for intelligent inspection and maintenance of ships, significantly improving operational efficiency and management effectiveness.