Reliability Evaluation Method Research Based On Condition Monitoring For Rotating Parts

Considering modern production equipment have being more and more integrated, intelligent, sophisticated, automated in recent years, it gains more importance to obtain state information and reliability of equipment in time and in accuracy in order to attain safe sustainable and stable processing system. Condition monitoring is an important means to obtain state information of operation equipment in time, and it is also an important step in equipment reliability assessment and life prediction. Milling cutter and rolling bearing are common rotating parts using in industrial manufacturing, thus their life and reliability have close impact on operation sustainability and stability of whole processing system and manufacturing equipment. In this paper milling cutter and rolling bearing were taken as example and we proposed a method to supervise state of rotating parts and analyze state data, besides states feature extraction of rotating part was analyzed, as well as reliability prediction. The main contents are as follows:(1) First, in order to compare several kinds of tool condition monitoring methods, foreign cutter life test data are analyzed and found that acoustic emission and cutting force are better than other methods. Wavelet packet analysis method is used to extract the relevant characteristics of acoustic emission signals which reduce the noise. In order to assess the reliability of specific components in the process of running, a reliability evaluation method based on Logistic regression model is proposed. The feature parameters which meet the dynamic characteristics of deterioration are extracted form historical monitoring data. Combined with tool state, a reliability estimation model based on logistic regression is set up. This method can be used online assessment equipment reliability.(2) Reliability prediction based condition monitoring is an important direction of the reliability of technical studies, with practical engineering significance. With reliability prediction for single rotating parts, a reliability prediction method based on state space model is proposed. During the manufacturing process, vibration signals or acoustic emission signals are measured online by experiment. The wavelet packet energy features are extracted from monitoring signals. Evident characteristic parameters are selected as indicators. A moving average filter was applied to the time series feature to further improve the signal-to-noise ratio of status characteristic indexes. A state space model is set up and applied to predict the probability distribution of characteristic index, then, reliability was calculated. Combined with tool and bearing test data, the accuracy and validity of the method is verified.(3) For the reliability prediction of rotating parts, in the final part of this article, a real-time reliability prediction method based on RVM (relevant vector machine) and sum of two exponential models is proposed. Using RVM sparse learning modeling method to train the current condition monitoring indicators, then, regression forecast of characteristic parameters and estimate the noise variance. The sum of two exponential models is fitted to predict the characteristics of indicators, which used to predict the probability distribution of characteristic index and calculating the reliability of rotating parts.