It　is　still　challenging　to　accurately　predict　the　remaining　useful　life　(RUL)　of　bearings　with　fluctuating　degradation　processes.　To　address　this　issue,　this　article　proposes　a　novel　robust　dual　unscented　particle　filter　(DUPF)　method　for　RUL　prediction.　First,　a　dual-stream　unscented　particle　filter　model　is　constructed　to　leverage　the　hidden　degradation　information　at　different　time　scales　with　different　prediction　models,　which　enhances　model＇s　capability　to　track　various　fluctuating　degradation　trends.　Second,　a　comprehensive　fusion　strategy　is　designed　to　adaptively　optimize　the　weights　of　double　streams,　in　which　the　maximum　failure　probability　of　dynamic　Bayesian　(DB)　is　quantitatively　evaluated　to　improve　the　reliability　of　the　prediction　results.　The　proposed　method　is　tested　using　two　datasets　and　compared　with　several　state-of-the-art　methods.　The　results　show　that　the　proposed　method　can　improve　prediction　accuracy　and　is　robust　to　fluctuations　in　degradation　processes.