| Rolling bearing is one of the key element in industry.Spalling is an important failure form of rolling bearing.Usually,crack core is formed on the weak point under the contact surface by alternating stress on the surface of ball and race,then pitting and spalling will be formed when the crack propagates to material surface.Acquiring the performance of rolling bearing in time that can relate to the reasonable plan of maintenance and its spare parts directly to avoid economic loss,major accidents even casualties when suffering an emergency situation.It can take advantage of the whole rolling bearing life in a reasonable way.The potential of it will be digged sufficiently that the waste will be avoided.Thus,it is helpful to improve the compony economic benefit when mastering the fatigue evolution process of rolling bearing.It is also helpful for the research of the damage mechanism of rolling bearing.In the meantime,an exact fatigue evolution data is the important data sources and precision preserve for the calculation results of performance degradation assessment and life prediction.But the dispersion of rolling bearing life is obvious,it will not acqurie the fatigue information from a single rolling bearing well by the methods of probability theory,fracture mechanics,damage modeling,etc.Classical condition monitoring and fault diagnosis technique of rolling bearing based on vibration technique,etc.can acquire the state after surface damage only,and still has a certain limitation for the early fatigue detection.Material will generate acoustic emission(AE)technique when a deformation or crack propagation appears by loading on it.In the early stage of 1950s,the pioneering research of German scholar Kaiser improved the birth of acoustic emission and its development.Then it became a powerful tool to acquire the early fatigue damage information of rolling bearing regularly.Acoustic emission technique can detect active defect under material surface effectively and reliably.Nature failure AE data of rolling bearing can acquire the fatigue evolution process in a more correct way.But the time cost is much higher,and the transfer path is complex in current fatigue testing program that lead to a serious signal attenuation.It is hard to avoid noise mixed in AE signal,and normal means can not process it well.Especially during the early damage stage,weak signal will be hiden easily.The current method of acquiring the fatigue process of rolling bearing usually adopt trend analysis with 3-5 traditional AE indexes.There are some insufficients in them,such as deciding the fixed threshold voltage artificially that will easily introduce the subjective error.It will not reflect the fatigue state of rolling bearing comprehensively just with a few features.The research on the relationship of feature and damage is not studied deeply enough.Calculation cost will be increased with more features.The redundancy and uncorrelation in the features will interfere the acquiring of fatigue process information.Besides,the sensitivity in each feature is different,evolution information disperses in these features in an uneven way,and it needs a certain professional knowledge or experience.It is inconvenient to judge the evolution process by analyzing the features one by one.It will take more time and be a heavy workload.Thus,it is very important to extract the fatigue damage evolution characteristic accurately and effectively.The present research is supported by the National Natural Science Foundation of China(No.51465022,51265018,etc.).The dissertation is around solving the problems in extracting fatigue evolution information on the life-cycle stage of rolling bearing efficiently based on acoustic emission technology.The life-cycle stage is made up of the stages from no damage stage until spalling failure stage.The research route of combination of theoretical research and experimental verification is adopted.An evolution information extraction frame of rolling bearing is established preliminarily.The main research contents are given as follows:(1)Reviewing the research statuds of the researches of rolling bearing fatigue,AE theory and its monitoring technique,de-noising technique,feature evaluation technique and feature extraction technique combine with rolling bearing mornitering theory and engineer require.Then the research contents and technological route are decided.(2)Insufficients traditional AE monitoring indexes are analyzed,and a new kind of AE monitoring index with floating threshold voltage and mean algorithm is proposed.The multi-indexes trend analysises are carried out.The experiment results show that the new AE index could acquire fatigue evolution information of ball bearing better.Then a few indexes,which are more sensitive to fatigue,are extracted.A general evolution law of ball bearing is summarized and the corresponding conclusions are obtained.A new rolling bearing fatigue test rig is designed as a test and examination platform for the subsequent theory researches.(3)For solving the problem of noise can be mixed easily in AE signals,researches on noise sources and its characteristic in AE signals are further studied.Then the normal process methods are given.A wavelet packet de-noising algorithm based on secondary correlation weighted threshold is proposed.The algorithm considers the similarity between the noise in AE signal autocorrelation form and the introduced noise autocorrelation form.Then,a simulation and a measured signal of trust ball bearing experiment researches are carried out.Results show that the proposed algorithm can restrain noise well in the acquisition AE signal.SNR can be improved steadily.The impact is obvious in the de-noised AE signal.(4)The insufficients of current distance extimation method are analyzed.For solving the problem of the irrelevant and the redundancy possible in multi-parameters that can disturb the effective damage information recognition of rolling bearing.According to the traditional distance estimation algorithm,the dissertation presents a novel method based on position compensation coefficient,where the accuracy and stability of object recognition has been improved.Experiments show that the proposed algorithm can acquire damage sensitive feature set well with a high damage recognition accuracy.It has a better stability and a higher classification precision compare with other methods.(5)Based on the research of the algorithms proposed above,a fatigue evolution information extraction algorithm with kernel entropy component analysis(KECA)method which is based on improved particle swarm optimization(PSO)is proposed with the object that the least features cover the most effective information.Then,a secondary feature fusion algorithm is proposed according to the entropy calculation law.It can fuse the fusion features further.The method can extract evolution information of rolling bearing more effectively.The results of the test AE signal analysis show that the main kernel entropy score of improved KECA algorithm can recognize the rolling bearing fatigue evolution process effectively.The secondary fusion algorithm can converge the fatigue evolution information greatly from each feature.The fatigue process of rolling bearing can be described conveniently and effectively by just adopting a single secondary fusion feature. |
