This paper addresses Gross Error Detection using uni-variate signal-based approaches
and an algorithm for the peak noise level determination in measured
signals. Gross Error Detection and Replacement (GEDR) may be carried out
as a pre-processing step for various model-based or statistical methods. More
specifically, this work presents developed algorithms and results using two
uni-variate, signal-based approaches regarding performance, parameterization,
commissioning, and on-line applicability. One approach is based on the Median
Absolute Deviation (MAD) whereas the other algorithm is based on wavelets.
In addition, an algorithm, which was developed for the parameterization of the
MAD algorithm, is also utilized to determine an initial variance (or peak noise
level) estimate of measured variables for other model-based or statistical methods.
The MAD algorithm uses a wavelet approach to set the variance of the noise in order to initialize the algorithm.