With the continuous development of mechanical equipment towards highspeed,largescale and intelligent,fault prognostics and health management (PHM) has been a hot issue in industrial field.The working environment of mechanical equipment is poor,the working conditions are complex,and multiple systems are coupled with each other,so the condition monitoring information presents the typical big data 4V characteristics of “huge volume,multisource heterogeneous,fast generation and sparse value” during longterm service.This makes the health management of mechanical equipment in the context of “big data” presents “three highs” features:(1) highcapacity big data storage capability is required; (2) highefficiency realtime processing capability is required; (3) highstrength heterogeneous adaptation is required.There is a need for a data processing method that can adaptively extract fault features from massive data and perform diagnostic and RUL prediction.As a further development of machine learning,deep learning theory has achieved great success in fields of image processing,speech recognition,and natural language processing with powerful modeling and data processing capabilities.Applying deep learning theory to solve the problem of equipment fault prognostics and health management (PHM) in the era of “big data” has become a hot research point.Starting with the basic process of deep learning in equipment fault prognostics and health management applications,the main types and characteristics of neural network input features applied in PHM were summarized.The four types of neural network models and corresponding model training algorithms in PHM applications were summarized,and the research progress of deep learning in PHM applications at home and abroad was reviewed.The future development directions of deep learning in PHM application were pointed out.