| This paper is the continuation of “Development and Research of Predictive Protectors of Intelligent Motors”,the key scientific and technological project of Shanxi Province.The asynchronous motor,machinery drive equipment of which the motive power is electricity,is widely applied to drive various machinery and industrial equipment inindustrial and mining enterprises.Because of its heavy working load and bad operating environment,it is rather common to arise faults,and once faults occur,huge economic loss will be created on enterprises.Hence,with the purpose to find faults early,prevent the emergence and development of faults and avoid the occurrence of destructive and catastrophic accidents,it is very necessary to carry out the research on intelligent diagnosis and pre-warning of early-stage faults and eliminate the faults in the bud.Aiming to the complex fault mechanism of asynchronous motor,for example,there is no one-to-one mapping relationship between fault causes and fault characteristics.In fact,there are one to many mapping relationship in which the same fault can be shown as a number of features or many to one mapping relationship in which the different faults can be shown as the same characteristic.This indicates that the relations between the faults and their characteristics are so complex that it is difficult to determine the fault type according to single.This paper conducted systematic and deep analysis on the fault mechanism of the asynchronous motor,built the relational expression between electromagnetics and dynamics at fault time of the asynchronous motor by analyzing and researching the electromagnetic relation and dynamic characteristics at normal and fault time of the asynchronous motor and on the basis of the operating model of the asynchronous motor,explored various faults of broken rotor bars,static eccentricity,dynamic eccentricity,static&dynamic mixed eccentricity and rotor bearings and the frequency spectrum change law of such faults when many faults occur in the stator current and vibration signal simultaneously,provided feasible and effective methods and analysis basis for ascertaining the emergence and development of effectively-detected faults in theory.The non-stability of the asynchronous motor at fault time,the complexity of fault signal change as well as the weakness and fuzziness of early-stage physical signals make early-stage fault diagnosis and multi-faults separation more difficult.This paper adopted the method of multi physical information fusion.Wavelet packet analysis method with time domain local characteristic and multi resolution characteristic was adopted to conduct wavelet packet decomposition at 9 or 8 layers on the current signal of the motor and the vibration signal of multi detection points,realized fine division in the whole frequency range,made the characteristic frequency more obvious and frequency ranges decomposed able to reflect the energy change of the characteristic frequency at normal and fault time of the asynchronous motor clearly.This paper abstracted energy values of 16 characteristic frequency ranges related to faults as the characteristic information of motor state to provide basis for fault diagnosis of the motor.Regarding the dissatisfactory treatment effect problem resulted from the empiricism and randomness of the selection of wavelet primary functions when applying wavelets,this paper introduced noise power as the fitness function of wavelet primary functions according to the characters and property of the above-mentioned detection signals,proposed the method to select wavelet primary functions according to the two parameters-noise power and noise power tolerance,obtained the best wavelet primary function applicable to fault analysis of the asynchronous motor through the optimization and selection of wavelet primary functions,overcame the low treatment precision and poor effect problem resulted from the selection of wavelet primary functions according to experience or references in traditional way,made the method for motor signal treatment based on wavelet more effective,proved the effectiveness of this method through the inspection of practical operation data and provided a feasible method for relevant application.With the purpose to improve the fault identification rate and generalization ability of the neural network,the corrected fusion algorithm was designed in this paper to realize the optimization of various parameters RBF neural net used for motor state identification,adopt corrected immune algorithm to determine the central location and quantity of data in the hidden layer of neural network;Meanwhile,the fuzzy C means clustering algorithm was also adopted to further optimize RBF hidden layer.Especially,according to the characteristics of clustering data in the corrected immune algorithm,the new calculating algorithm of initial antibody crowd was proposed to improve the deleting algorithm of antibody by use of the selecting method of adaptive affinity threshold and the immune mechanism of antibody and the concentration regulation principle of antibody were added to the immune algorithm to increase the convergence rate of the algorithm.By running the sampling data,it has been proved that the method is effective to have resolved many problems in traditional neural network,such as lower identification ability,poor accuracy and bad generalization ability.Fifthly,regarding the fact that the electromagnetic relation and dynamic property of the asynchronous motor is difficult for actual simulation due to its large volume and heavy weight,this paper built electromagnetic and dynamic simulation model of the asynchronous motor at multi fault time on the basis of modern simulation platform.Ansoft finite element simulation software,P/roE 3D drawing software and MSC.ADAMS antibody dynamics analysis software were adopted to respectively build the finite element model of broken rotor bars of the asynchronous motor,the finite element model of static eccentricity,the finite element model of dynamic eccentricity,the finite element model of static&dynamic mixed eccentricity,the multi-body dynamics model of rotor supporting bearing ball injury,the multi-body dynamics model of rotor supporting bearing inner ring injury and the multi-body dynamics model of rotor supporting bearing outer ring injury,well simulated broken rotor bar fault,static eccentricity fault,dynamic eccentricity fault,static&dynamic mixed eccentricity fault,rotor supporting bearing ball injury fault,rotor supporting bearing inner ring injury fault and rotor supporting bearing outer ring injury fault,ect,conducted deep simulation analysis on these above-mentioned faults,the result verified the correctness of the theoretical analysis on one hand and provided further guidance for the following experimental analysis.With the purpose to connect the theory with the practice and verify the correctness of the theoretical and simulation analysis,this paper designed fault motors with broken rotor bars,static eccentricity,dynamic eccentricity,static&dynamic eccentricity,rotor supporting bearing ball injury,rotor supporting bearing inner ring injury,etc which were manufactured by specialty factories.Broken rotor bars is formed by the gaps existed in conducting bars resulted from the adding of insulated gaskets to rotor models;eccentricity is realized by the eccentricity of bearings,that is,on the basis of originally-matched bearings,make the diameter of the inner rings of eccentricity bearings bigger than rotation shafts and the diameter of outer rings slightly smaller than the inner diameter of the space where the outer rings are located,thus two air gaps will be generated both inside and outside bearings;bearing fault is realized by the damage of bearing-related positions.Hence,the case that other faults of practical fault motors will always occur when one fault occur will be avoided,e.g.the eccentricity will be caused when broken bars occur,etc.Through constantly replacing rotors,assembling eccentricity sleeves with different eccentricity distances and replacing various fault bearings in practical operating,we adopted one and many broken rotor bars,10%/20%/40% static eccentricity rate,20%/40% dynamic eccentricity rate and real-time data of bearing ball injury,bearing inner ring injury,bearing outer ring injury and other faults,accumulated precious operating data,further verified the correctness of the theoretical derivative result through analysis of experimental signals. |
PDF Full Text Request