A Hybrid Process Data Denoising Method Based on EEMD and Piecewise Curve Fitting

To improve the fault detection efficiency in chemical process monitoring, process data preprocess aiming at filtering noise and eliminating gross errors is valid and effective. In view of the features of chemical process data, a novel hybrid preprocessing method is presented Based on the EEMD denoising and Piecewise Curve Fitting. In this method, a denoising scheme Based on EEMD method is used to remove white noise from the signal. The first order and second order derivative sequences are obtained Based on piecewise fitting of the sampling data of variable signals. The smoothness and continuity of the boundary is guaranteed through weighting the over-lapping data. Compared with traditional filtering, this EEMD and piecewise curve fitting Based filtering does not need to define the coefficients of filter, so it is fully data-driven and adaptive. The simulation and experimental results demonstrate effectiveness of the proposed method.