Bearing Performance Degradation Assessment Based On Continuous Hidden Semi-Markov Model

With the change of people’s maintenance concepts, the traditional periodicalmaintenance and later maintenance are gradually converting into Condition BasedMaintenance (CMB).as the premise of establishing reasonable maintenance strategy,theperformance degradation assessment of equipment has been widely concerned. The substanceof equipment performance degradation assessment is to recognizethe current running state ofequipmentand to forecast the future state of the movements by analyzing the equipment data.There are two main difficulties:firstly, in the state monitoring of complex mechanical systemor key components in the field of industrial, in order to know the unit running conditiontimely and accurately, we often lay out a large number of sensors, sothe monitoring dataobtained ismultisource and high dimension. How to extract the useful information which canreflect the running status of the equipment and remove the redundant information from thesecomplex monitoring data are the first problem. Secondly, for complex mechanical systemfaults, the accumulation of data under the whole process of abnormal state changes is usuallydifficult, which requires that we can forecast the monitoring state only based on the effectivelearning of normal operation data.Aiming at the above two problems, the multi-sensor feature fusion method based onTangent Space Alignment (LTSA) is studied in this paper, andthe correspondingimprovement of Continuous hidden semi-Markov model(CHSMM) in practical applicationshas been made to assess bearing performance degradation. Finally the effects of the featurefusion and model are verified through the simulation experiments, simulated bearing outerring fault experiments and bearing fatigue life experiments.The main contents of this paper include:(1) The extraction process of bearing performance degradation characteristics isdiscussed, including the extraction of single-sensor frequency band energy feature and thefusion of LTSA with multi-sensor characteristics. What exactly involved are the selection ofband’s number and determination problem of parameters in LTSA.(2) Focuses on the form applied to equipment performance degradation assessment of theextended form of HMM-Continuous Hidden Semi-Markov Model (CHSMM). Animprovement has been made against the practical application of CHSMM.The initializationof the model and parameters selection methods are discussed in this paper. The faultrecognition framework of different degradation degree is set up. (3) The effectiveness of this method is verified using experiments on different degrees offailure in bearing outer ring and fatigue life experiments. The model’soutput results of fatiguelife experiment are analyzed and compared with common time domain index, which reflectsthe sensitivity to the early failure of the model.