Study On Condition Monitoring And Performance Evaluating System Of Liquid Film Seals

Liquid film seals are gradually used in petro-chemical industry because of the excellent property of zero leakage of medium.However,small solid particles in barrier fluid usually block grooves on seal end face,so the hydrodynamic action is weakened,which shrinks the pumping rate and increases the frictional power.Meanwhile,the performance is terribly influenced and even results in a failure.If the abnormal states are detected timely and countermeasures are taken early,bad situations such as leakage,barrier liquid polluting and even a shutdown would be avoided.On all these considerations,a spiral groove down-streaming liquid film seal was chosen as the subject and the condition monitoring and performance evaluating program was proposed based on BP neural network model and fuzzy logic inference.The film thickness and pressure of barrier liquid were selected as the directly monitoring valuables,while pumping rate and frictional torque were selected as the evaluated valuables.Influences that shallowed grooves made on the performance of liquid film seal were analyzed theoretically.Dynamic model of the seal with a flexibly-mounted rotor was established.The film thickness equation,axis and angular dynamic equations sets and Reynolds equation with squeezing item which was discretized by finite difference method were solved simultaneously based on a MATLAB program.Results indicated that in the transition stage,the response curves of torque varied because of the changed groove depth,which led to the phase lag of misalignment response curve.In steady response stage,the lower the goove depth was,the smaller the amplitude of misalignment response curve was,which resulted from the limitation to the precession amplitude caused by the thinner film thickness.Response curves of pumping rate were consistent with those of film thickness,while response curves of friction torque were opposite to those of film thickness,and the varison of groove depth had a strong impact on parameters.The h-h_g,Q-h_g and M-h_g equations are fitted by solving performance parameters in different groove depth and pressure.750 sets of performance parameters,which were devided into training samples and testing samples,were built by the method of experimental measurement combined with theory fitting.BP neural network was designed and trained based on MATLAB Neural Network Toolbox.The best node number of hidden layer was 9.BP neural network maps the evaluated valuables when the directly monitoring valuables are inputted.Based on MATLAB Fuzzy Inference System and method of expertise,the performance evaluating program was built,and the results of the liquid film seal were gained by inputing pumping rate and frictional torque.Based on MATLAB Graphical User Interfaces,the liquid film seal performance evaluating software was developed by combining functions of data acquisition,BP neural network maping and fuzzy logic inferring.The software was verified in laboratory,and the results showed that the software provided correct evaluation.Furthermore,a thickness-measurable liquid film seal product was designed to cooperating with the evaluating software.