Study On The Fault Diagnosis Of Rolling Bearings In Dryer Section Of Paper Machine

To be the infrastructure industry of the country, papermaking industryoccupies an important position in our national economy. Paper has close relationto almost all aspects of our national life, such as economy, culture, life, nationaldefense, and so on. It has become the basic raw material of industrial anddomestic. Paper machines are moving to the direction of large, broad andhigh-speed along with the social progress and technological development.However, the development process is often influenced by some restricting factors,for example, one of the main bottlenecks of restricting the machine speedincreased is the dryer section. Meanwhile, as the most critical and most easilydamaged parts of the paper machine, the rolling bearings are directly related tothe operation of the entire production line. Therefore, once the bearings failure,the production line has to be stopped to replace the fault bearings. If thereplacement is not timely, to be light, it may result in huge economic losses, inserious cases, it will produce a significant or even catastrophic casualties andprofound social impact. The dryer section is the largest part of the paper machinefor rolling bearing application. Therefore, in order to effectively improve thepaper machine speed, as well as ensuring safe and reliable operation of thehigh-speed paper machines, the bearing fault state detection and diagnosis of thedryer section is quite necessary and has far-reaching significance.Firstly, the generation mechanism of the typical rolling bearing faults wasintroduced, and the problems of rolling bearing fault diagnosis have been studied,the main research work can be summarized as the follows.(1) The fault diagnosis scheme based on wavelet analysis and BP neuralnetwork was proposed. Considering the characteristic of effective for themulti-resolution of wavelet packet transform and the ability of different modesclassification of neural network, a rolling bearing fault intelligent diagnosismodel was established: Firstly, pre-processed the vibration data utilizing wavelet packet analysis which serviced as a preprocessor of the BP network, and thefrequency domain energy eigenvectors were established; then constructed asuitable BP neural network model according to the need; and finally utilizedthese energy eigenvectors as the input of the neural network to realize the modeidentification of the rolling bearing failure.(2) Collection of vibration data of rolling bearings under several operatingconditions was completed through the designed experimental platforms. Thesevibration data was pre-processed using the MATLAB wavelet analysis function,and the energy eigenvectors of each set of data were extracted. Then using theMATLAB neural network function constructed fault diagnosis model based onBP neural network, and applied it to the rolling bearing fault diagnosis.Simulation results show that the combination of wavelet analysis and BP neuralnetwork for fault diagnosis, not only is able to detect the presence of the rollingbearing fault, but also can accurately identify the type of bearing failure.(3) The diagnosis results were compared among the ordinary BP networkalgorithm, the adaptive modifying learning-rate algorithm and the improvedLevenberg-Marquardt BP neutral network algorithm, the results show that theimproved Levenberg-Marquardt algorithm can not only improve the convergencespeed, but also of a relatively high fault diagnosis accuracy rate. However, thereis still shortcoming of convergence unstable.