1 Research on the life and reliability of rolling bearings
1.1 Development and status quo of bearing life and reliability testing machines
Bearing life refers to the total number of revolutions or working hours of a rolling element or raceway of a bearing before a fatigue peeling occurs. Bearing life testing is inseparable from bearing life testing machines. The development of bearing life testing machines has also witnessed the development of bearing life research. The test research on bearing life in my country is mainly led by two scientific research institutes, Luoyang Bearing Research Institute and Hangzhou Bearing Test and Research Center (United Nations Aid), supplemented by the life and reliability test bases of other related companies, to jointly undertake the bearing life of my country’s bearing industry , Reliability and performance test research work. At present, the design, R&D, and production of bearing life testing machines in my country have been completely independent, and some technical concepts have reached the international leading level. However, compared with SKF, Schaeffler, Timken, NT, etc. A large foreign bearing company started very late. In the early 20th century, the development of China's bearing industry mainly relied on the technical support of the former Soviet Union's big brother. The life test of the bearing was mainly carried out on the basis of the ZS bearing life testing machine, and the evaluation quality of this testing machine has long been eliminated from the service performance of the bearing. Development requirements; and the "F&M 5" new rolling bearing fatigue life testing machine introduced from the United States by the Hangzhou Bearing Test and Research Center (HBRC) through the United Nations aid project is not only expensive and technologically monopolized, but also uses a pneumatic high-voltage power source and 60Hz electrical frequency. Not suitable for China's national conditions. Therefore, it is imperative for bearing life intensification testing machines to realize independent production. In the 1990s, Hangzhou Bearing Test and Research Center independently developed ABLT-1 automatic control rolling bearing fatigue life enhancement on the basis of foreign advanced life testing machines. The testing machine opened up new markets and new prospects for the domestic life testing machine, and it had reached the international advanced level at that time. With the birth of the ABLT-1 rolling bearing fatigue life intensified testing machine, the development of domestic bearing life testing machines has sprung up, but most of them are derived or improved on the basis of ABLT-1.
1.2 Research on prediction of bearing life and reliability test
During the operation of the bearing during the whole life cycle, it is likely to be affected by factors such as high temperature, poor lubrication, improper assembly, foreign matter intrusion, etc., resulting in damage to the bearing and failure of the bearing. Because the bearing life is very discrete, a batch of bearings with the same structure, the same material, the same heat treatment, and the same processing method under the same working conditions have a difference of dozens of times or more between the maximum life and the minimum life. Traditional mathematical statistics show the bearing life. The test data approximately conforms to the Weibull distribution or the lognormal distribution, but it is still difficult to predict in actual working conditions. Therefore, the effective processing of bearing life test data is particularly important, and domestic and foreign research institutions are also actively carrying out relevant research on bearing life test data. Saxena et al. used the power spectral density parameter as the performance degradation index of rolling bearing to predict the remaining service life of the bearing. The density parameter can diagnose the location and extent of the fault. Xiao Ting et al. used kurtosis and multi-domain feature sets as trend prediction indicators, which not only effectively reflect the running state of the bearing, but also predict the performance degradation trend of the bearing. Banjevic et al. used the proportional hazard model to predict the reliability function and remaining life of the equipment, and used the covariate at a certain moment as a benchmark to predict the remaining life. Based on previous studies, Kacpnynski proposed a prediction model combining monitoring data with material parameters, and used this model to predict the life of rolling bearings. Kimotho et al. proposed a hybrid differential evolution particle swarm optimization (DE-PSO) optimization algorithm to optimize the kernel function and penalty parameter prediction method of support vector machine, which improved the classification accuracy of support vector machine and the accuracy of remaining life prediction, and adopted NASA standard bearing failure data was verified. Orsagh et al. used the Yu-Harris model to predict the initial time of fatigue spalling failure of rolling bearings, and used the Kotzalas-Harris model to predict the failure time of rolling bearings. Panigrahi proposed a diffusion particle swarm optimization algorithm (DPSO) to solve the problem of maximum likelihood function estimation in the research of bearing performance degradation, and achieved good prediction results.
The current life model based on statistics still occupies a leading position in bearing life prediction. However, experiments and engineering applications have found that the life calculated by the statistical life model is usually conservative and the life span of the bearing is large. Therefore, how to study the mechanism of bearing performance degradation to improve Bearing life models are a major issue. The life prediction method based on condition monitoring has become a hot area of bearing life prediction research with the development of new information technology and artificial intelligence. With the help of big data, artificial intelligence information and other technologies, dynamic signals reflecting bearing service performance can be obtained, signal characteristic parameters characterizing bearing performance degradation, and the mapping relationship between signal characteristic parameters and remaining life can be established, so as to realize the prediction of remaining life. However, there is a lack of appropriate characteristic parameters to measure the evolutionary law of bearing performance gradual decline during operation. Compared with traditional life prediction models, artificial intelligence methods such as neural networks are not clear enough in physical meaning, and the parameters have greater influence factors. How to conduct in-depth research on the difficult points is very important to the bearing life prediction technology.
2 Research on the big data health monitoring of rolling bearings
Bearing health monitoring is based on the existing known data to predict changes in the future operating status of the bearing within a certain period of time, in order to accurately and quickly obtain fault development information. Carrying out condition monitoring and health monitoring of bearings can grasp the law of bearing degradation process, prevent larger failures from occurring, and prevent problems before they occur.
Caesarendra et al. used correlation vector machine regression algorithm and logistic regression combination method to evaluate the degree of bearing degradation and predicted failure time. Yu et al. used the local protection projection method to extract the characteristics of bearing operation, and used Gaussian mixture model and statistical indicators to evaluate the health of the bearing. The study showed that the effect of the feature extraction was significantly better than the component analysis method. Li Xiuwen et al. used frequency domain morphological filtering to reduce the noise of low-speed rolling bearing acoustic emission signals, and compared simulation with actual bearing signals. The results showed that this method has good results. Rojas et al. proposed a SVM-based fault diagnosis method for rolling bearings. The time domain characteristics of rolling bearing vibration signals were identified by SVM. Jeong et al. used the methods of discrete wavelet transform and spectral kurtosis analysis to obtain the characteristic frequencies of the faults in each part of the rolling bearing, and completed the inner ring-outer ring, inner ring-rolling element, outer ring-rolling body, inner ring-outer Diagnosis of ring-rolling element compound fault form.
Some of the indicators for bearing running status monitoring can be achieved by collecting physical parameters, such as temperature, vibration, noise amplitude, etc., and some require researchers to advance the data through signal processing methods, such as the degree of temperature change. , Vibration intensity, sound pressure intensity, etc. Through the characterization of these indicators, the health status of the bearing can be evaluated, the prediction and early warning of the running status of the bearing can be carried out, and the engineering staff can be guided to take corresponding protective measures to avoid the deterioration of the health status of the bearing. However, for bearing health monitoring and analysis at this stage, a unified platform has not been established to integrate common features in bearing failures, and it is still not very good in the mining of failure causes, high-response prediction methods, and collaborative management of multiple bearings. perfect.