Variance Partition Analysis - Starting Point for Profile Optimization

摘要

Sixty years after the first publications on variance partition analysis (VPA) it is necessary to reconsider this topic again. The variability of a paper quality is split into 3 components, the variance in machine direction (MD), in cross machine direction (CD) and in a residual component (RES), which is not related to MD or CD. The break down helps to distinguish which methods of profile optimization are required to come to an improvement. The variance partition analysis needs to be exact, as only then the conclusions are built on solid ground. But, the actual situation is that at least 4 different algorithms for this VPA are currently used in paper industry: 1. 2 algorithms published in TAPPI publications. 2. 2 algorithms of large QCS suppliers These 4 algorithms may lead to largely different results. A mathematical analysis showed that even the Tappi calculation methods are not able to calculate the amplitudes of MD and CD variation precisely. Tappi T 585 [5] which is used for basis weight laboratory measurements gives almost correct results, but especially the TAPPI algorithm, which is recommended for on-line scanning systems (TIP 1101, [6]) leads with real data to up to 25% too high CD variances. In certain cases it pretends that CD improvement potential exists, but in reality it is not available. The different calculation methods lead in praxis to obscurity i.e. when assessing profile quality based on parent roll reports. The results may lead to wrong conclusions, especially in comparison to results obtained by lab measurements or by other QCS suppliers. These are the motivations to present a new calculation method for the variance partition analysis. It is an improved version of the formulas given by T585. Two practical examples of different paper grades show how the new algorithm performs, how it helps to assess the paper quality more accurate and how this leads to better decisions on optimization measures.