The invention relates to an on-line real-time failure monitoring and diagnosing system device for industrial processing of residual oil. The device has the functions of data acquisition, data filtering, sensor efficiency analysis, alarm management (process monitoring, pre-alarm and alarm counting), equipment monitoring, expert knowledge management, intelligent self-explanation of failure scenario, visualization display and the like, and can provide the daily management functions of shifting of duty, operation log, working calendar, process chart and the like. An inference engine subsystem is a core part, comprises a process monitoring engine and an equipment monitoring engine, is in charge of most inference work, and mainly comprises a signed directed graph (SDG), principal element analysis, fuzzy logic, mathematical analytical model and expert rule base. A process monitoring module uses a hybrid algorithm engine with various fused inference algorithms to obtain compatible paths of failure propagation, and removes false compatible paths by the filtering effect of expert rules to obtain the final diagnostic result.