Bearing Remaining Useful Life Prediction Based on a Nonlinear Wiener Process Model
Table 1
Parameters update and RUL prediction process.
Step 1: Monitor the bearing’s health state using the method described in Section 3.1. When the degradation starts, define the beginning time as t0.
Step 2: At time t0, define the initial distribution of z0. Then sample particles from the initial distribution and initialize particle weights with =1/N.
Step 3: At any inspection time tk (k≥1), sample particles from the importance PDF defined by the state space model (19).
Step 4: Update particle weights with the newly obtained HI using equation (8).
Step 5: Resample from according to .
Step 6: Update the HI and the model parameters at time tk with equation (10).
Step 7: Calculate the future states recursively by equation (22) for each particle, producing N degradation paths. For each degradation path, its failure time is defined as the time when the predicted HI exceeds the FT, and is the RUL of this degradation path. Consequently, the failure time PDF and the RUL PDF can be obtained from all the values of failure time and RUL , respectively.
Step 8: For the next inspection interval, let k = k + 1, then turn to Step 3 and repeat Step 3-7 until the estimated HI .
