Application of design of experiment techniques to measurement procedures. An example of optimisation applied to the digital measurement of partial discharges

In the last decade, multistage digital measuring systems of partial discharges (PDs) have been introduced, allowing one to support the diagnostic of defects (sites of PDs) in power electric components. They transfer information about defects from full data sets of PD patterns, obtained from a Phase Resolved Partial Discharge Analyser (PRPDA) to reduced data set by implementing pattern recognition techniques. The latter data sets are then classified versus similar reference data sets. The validity of the above diagnostic requires that the measuring process, which is influenced by several factors, is optimised. The three main settings of the PRPDA are among such factors of influence, but so far, as a simple mathematical model of this measuring process is not available, it has not been possible to quantitatively assess their "weight" on the validity of above diagnostic. This work presents a successful "Design Of Experiment" (DOE) approach to solve the latter problem. The DOE analysis of the results of 81 PD tests performed on a simple physical model of an insulation system quantified the weight and the interaction of the three factors and allowed one to derive criteria for selecting the "optimal" values of such factors and the "optimal" composition of the reduced data sets.