The Research Into Incipient Fault Diagnosis Method On Hybrid System

With the development of information and technology,the accuracy and complexity of production process control system were increasing.The whole system will be invalid,paralysis and even causing unimaginable loss of life and property when a fault occurred in such system.Because there were many slow change in production process,such as aging equipment,pipeline blockage faults etc.The key to prevent and reduce the incidence and expansion of accidents were detecting and diagnosing such potential incipient fault.Compare with fault diagnosis method having significant fault feature,incipient fault diagnosis technology were little in present research.The paper proposed the global analytical redundancy relations generated residuals according to the characteristics of incipient fault,and using fuzzy logic system or EWMA filtering techniques and adaptive threshold evaluation method to evaluate the residuals and realize the incipient fault diagnosis.The main contents were as follows:Firstly,the background research status of incipient fault diagnosis were introduced.We analyzed the problems of incipient fault diagnosis usually encountered in practical engineering and proposed a incipient fault diagnosis framework based on the global analytical redundancy relations of hybrid bond graph model and using fuzzy logic evaluation or EWMA-adaptive threshold evaluation method.Secondly,the structure and working principle of three tank system and the definition and classification of incipient fault were introduced.It’s hybrid bond graph simulation model built in the software of 20-sim.The incipient faults of three tank system were enumerated,and system level output waveform were simulated under all kinds of incipient faults conditions.Then,we researched global analytical redundancy relations of hybrid bond graph incipient fault diagnosis method.The hybrid bond graph transform to hybrid diagnosis bond graph By adding virtual sensor and inverted causal method,and get the global analytical redundancy relations of hybrid bond graph to establish the mode-dependent incipient fault signature matrix.The detectability and isolability of each incipient faults were analyzed in current mode.The residuals were simulated in 20-sim by building hybrid bond graph model and hybrid bond graph model,and it used to decide whether the hybrid system was fault or not,and located incipient fault source by seeking in fault signature matrix.Finally,taking into account the impact of modeling errors and measurement noise wereunavoidable in the actual production process.A fuzzy logic residual evaluation system for incipient fault was proposed.But if the incipient fault affect the residual was less than modeling errors and noise impact,it’s signature would be hided.Therefore the exponentially weighted moving average(EWMA) filter technology and adaptive threshold evaluation method were proposed to appraise residual.The simulation results shows that the fuzzy logic,EWMA filter technology and adaptive threshold evaluation method enhanced the incipient fault detection and isolation based on the global analytical redundancy relations method.