Abstract: The health assessment methods are studied for industrial robots, which are the most representative equipments in the field of intelligent manufacturing, to address the issues of their accuracy degradation and equipment failure. Firstly, the failure modes and their effects of the core components of industrial robots are analyzed, and the existing industrial robot health assessment methods are reviewed. Secondly, a health assessment framework of industrial robot based on cloud-edge collaboration and deep learning is proposed. At the edge layer, an anomaly detection method based on fleet clustering and peer-to-peer comparison is applied to detecting abnormal devices quickly. At the cloud layer, prognostics and health management with the artificial intelligence algorithms are used to perform deep health assessment on abnormal devices. Finally, the health assessment method of industrial robot based on deep learning is prospected.