Fluidized Bed Fault Diagnosis Method Based On Acoustic Emission Research

Chemical industry is the pillar industry of national economy, the largest output inchemical production in synthetic polymer is polyethylene, its mode of production is dividedinto high pressure method, medium pressure method and low pressure method. In theproduction of low pressure has a solution method, slurry method, gas phase method, relatively,the gas phase method has obvious advantages in production, now is widely used. Method ofgas phase polyethylene production mainly relies on the gas-solid fluidized bed reactor. Thesystem is more complicated and its work environment is often has the characteristic such ashigh temperature, high pressure, high humidity. Easy to failure in the long run, if not found intime, will cause equipment damage and the risk of production, so the fluidized bed online faultdetection is very important.This paper mainly studies the acoustic emission technique is applied in fault diagnosis offluidized bed. In the view of chemical production of gas solid fluidized bed hood failure affectthe fluidization quality problem, application of acoustic emission location measurementtechnology, through the uniform distribution installed in the fluidized bed plate at the bottomof the acoustic emission sensor collection and distribution of solid particle impact in the gas-solid fluidized bed plate produced by the acoustic signal, then the signal for multi-scalewavelet packet decomposition, find out the characteristic frequency band energy and theenergy distribution difference; Put forward a kind of methods that can rapidly and accuratelydetect the cowl fault location and type of fault.And complete the experiment, verify thefeasibility of the method. In addition, the fluidized particles in the fluidized bed is influencedby factors such as quality of raw materials, catalysts which can produce local high temperaturethen form agglomeration. This paper applied acoustic emission sensors to collect the acousticsignal, analysis the data by empirical mode decomposition and energy information entropy toextract fault feature vector, and then to extract the feature vector input to the training of supportvector machine (SVM) classification, complete agglomerate fault.classification. Fluidized bed experimental platform in the laboratory to simulate the failure of hood andagglomerate, completed the signal acquisition and fault recognition algorithm of authentication,experiments show that acoustic emission technique for fluidized bed fault diagnosis has a goodeffect.