Research On Fault Diagnosis Of Key Components Of Reducer Based On Variational Mode Decomposition And SVM

As a heavy-duty equipment in the three smelting production processes of iron making,steel making and steel rolling,the reducer has its internal gears due to its high temperature and dusty environment,frequent start-stop,acceleration and deceleration and variable working load.Key components such as bearings are easily damaged and cause failures,which will inevitably lead to the suspension of the gearbox or even the entire production line,giving rise to financial losses,production safety accidents.Therefore,it is of great significance to adopt effective monitoring methods for key components such as gears and bearings of the reducer and pre-judgment of early failures to ensure the safe and efficient operation of the iron making-steel-steel rolling production process.1)Whether it is possible to extract the effective information of the vibration signals of key components such as gears and rolling bearings of the reducer in the iron-steel-rolling smelting process is the difficulty and key point of its condition monitoring,and it control the correctness of subsequent state identification and diagnosis.About the subject,based on the Variational Mode Decomposition(VMD)theory,combined with energy entropy,permutation entropy and support vector machine,the details are listed below:1)Application research on vibration signal feature extraction of vibration signal of key components of reducer based on variational mode decomposition.For the traditional Ensemble Empirical Mode Decomposition(EEMD),it is easy to fall into modal aliasing,end effect,etc.The VMD is used to effectively extract the vibration signal characteristics of the rolling bearing and gear of the key components of the reducer.2)Research on State Identification Model of Key Components of Reducer Based on SVM.For the complex,variable and strong background noise of the reducer,the fault characteristic frequency of the key components is easily submerged,resulting in low accuracy of subsequent model recognition.In this regard,the acquired vibration signal is decomposed by VMD,and the energy entropy and permutation entropy of each component are extracted to construct a high-quality feature vector;Secondly,the above part of thefeature vector set is used as the training sample,and the state recognition model of the two key components of the rolling bearing and gear of the metallurgical reducer is constructed by the learning of SVM algorithm.Finally,the vibration data of the laboratory and the vibration data measured by the large reducer of Baotou Steel Wire Factory were used for verification.The results show that the proposed method can find,analyze and diagnose the location and type of faults,indicating its effectiveness and practicability.