Mechanical systems play a vital role in industry and society, but they inevitably suffer from faults and accidents, leading to both economic losses and life-threatening consequences in the production process. Prognostics and Health Management (PHM) offers an effective way to enhance the safety and reliability of mechanical systems. Over the past few decades, PHM technologies have evolved through three phases: reactive maintenance, preventative maintenance, and predictive maintenance. With the rapid development of cutting-edge information technologies, such as the Internet of Things (IoT) and artificial intelligence (AI), the increasing volume of data from mechanical systems now contains abundant health information, driving a new trend of intelligent maintenance in the big data era. This chapter introduces a series of intelligent fault diagnosis and prognosis methods, which are key components of PHM in mechanical systems. First, an unsupervised learning-based intelligent diagnosis model is developed to automatically extract fault features from the monitoring big data. This model links the learned features to the health states of mechanical systems, informing users about the fault’s location, type, and severity. Second, deep transfer learning is employed to construct the intelligent diagnosis model when there is insufficient data for training. After fault diagnosis, a data-model-fusion stochastic process model is established to predict the remaining useful life (RUL) of mechanical systems using the 3 \(\sigma \) interval and particle filtering algorithms. Experimental and industrial case studies are conducted to demonstrate the effectiveness of these methods.