Fault Diagnosis Of Agglomeration In Gas-Solid Fluidized Bed Based On Electrostatic Signals

Fluidized bed reactors are widely used in petroleum, chemical, biochemical, environmental, pharmaceutical and other fields. Agglomeration/sheeting is one of the most common problems in fluidized bed reactor. Large agglomeration would block distribution plate or discharge system, leading to implosion and emergency shutdown, which affects the safe and stable operation of the reactor seriously. Since the hydrodynamic behavior in industrial fluidized bed is highly complex, thus forming a complicated system with nonlinear and multi-scale characteristics, fault detection and diagnosis in multiphase flow reactors is rather difficult. Traditional detection methods (such as temperature detection, pressure detection, etc.) cannot meet the needs of modern industrial production. Therefore, researches regarding agglomeration issue in the fluidized bed and developing fast and accurate fault detection technology are not only challenging, but also of great theoretical importance and industrial application value.Based on the online measurement of electrostatic detection, temperature and pressure pulsation technique, combining modern fault detection and diagnosis technology, this thesis has an in-depth study of the early warning of agglomeration in fluidized bed. Moreover, by establishing a high-temperature agglomeration experimental device, the evolution of the agglomeration and the interaction between agglomeration and electrostatic potential are investigated in industrial as well as cold model experiments; the agglomeration model under high level electrostatic potential is proposed. Furthermore, early warning of agglomeration based on the electrostatic signal is realized. These research findings are of great importance for the fault detection of agglomeration and the stable production of polyethylene unit. The main contents in this paper are as follows.1. By examining the variance of a large number of process parameters during agglomeration process, the correlation degree between each process parameter and agglomeration is analyzed. By extracting the most relevant process parameter, the time sequence of signal fluctuations during agglomeration is revealed. The study finds that wall temperature shows the highest correlation coefficient with agglomeration (0.85), followed by electrostatic potential signal (0.71). The time sequence of signal fluctuations in agglomeration process is as follows:electrostatic potential fluctuation â†'temperature fluctuationâ†'cesium source signal fluctuation. This sequence reveals that when agglomerations are formed, the electrostatic potential reacts first, followed by wall temperature due to wall sheeting, and lastly cesium source detector when the agglomerations are too large.2. Based on fault detection and diagnosis technology, nonlinear fluctuations of wall temperature are investigated theoretically and the nonlinear characteristics are extracted and the agglomeration detection method (E detection method) is established. This study finds that the high-frequency components of the intrinsic mode functions (IMF), extracted from empirical mode decomposition results of wall temperature signal, increase significantly when wall sheets are formed. The energy moving average value, Em, is found to be effective in indicating initial sheeting. By importing Em into principal component analysis, the statistical T2 and upper control limit (UCL) are achieved, and the wall sheeting detection method is proposed:when T2 is less than UCL, no sheets are formed; when T2 is greater than or equal to UCL, sheets are formed. The industrial application results show that, compared with traditional wall temperature fluctuation discriminant or cesium source detection method, E detection method can achieve early warning of sheeting of about two hours in advance, and can greatly reduce false positive rate and false negative rate.3. Through simulating agglomeration process by heating polyethylene particles using hot air in the cold model device, the variation rules of temperature, electrostatic potential and electrostatic current, pressure signals are analyzed. The interaction mechanism of key process parameters (electrostatic potential or other parameters) and agglomeration is revealed and the effect regularities of fluidized gas velocity and particle diameter on particle agglomeration are acquired. This study finds that the mass fraction of initial particles decreases, and the mass fraction of particle agglomeration first increases and then decreases. Meanwhile, the bed level declines gradually and the whole bed pressure drop decreases before increasing, and finally declines. The effect of temperature on electrostatic level is ignorable. The charge-to-mass ratio and electrostatic potential is negative before heating, and first increases before decreasing to steady value as the fluidization time lasts. After heating and agglomeration, charge-to-mass ratio increases again and electrostatic potential first increases and then decreases, which shows a "V" shape variance. As the gas velocity increases, all of the maximum value, mean value and standard deviation of electrostatic current and electrostatic potential show an upward tendency during agglomeration process, and the time from heating to agglomeration decreases. As the average particle diameter increases, the maximum value, mean value and standard deviation of electrostatic current first decreases and then increases slightly. The polyethylene particle with an average diameter of 517 μm shows the highest electrostatic potential and forms agglomeration firstly.4. Based on the analysis of variation regularity of electrostatic signal during industrial agglomeration process, the agglomeration model in strong electrostatic field is proposed, and the agglomeration detection method based on the IMF instantaneous amplitude of electrostatic signal is developed. This study finds that electrostatic level decreases because of wall sheeting during industrial agglomeration process, and electrostatic potential presents V shape variance during agglomeration process. The agglomeration detection method based on the IMF instantaneous amplitude of electrostatic current signal cannot achieve early warning of agglomeration, while that of electrostatic potential signal shows great results since it can achieve early warning of agglomeration of 10～15 min in advance compared with pressure drop method in cold model experiment, and 1.2～3.5 h in advance compared with E detection method in industrial device. The result reveals that IMF instantaneous amplitude of electrostatic potential signal can achieve early warning of agglomeration and shows great universality.