Comparing the precision of two analytical methods for determining boron in glass-a statistical approach using Gage R&R

摘要

Analysing the composition of glass is important in industry and academia. It is important to have quantitative knowledge of the analysis method's precision. This allows the use of statistical methods to calculate the number of samples that should be measured to be sure of drawing a conclusion with an acceptably low risk of being incorrect. Calculating the appropriate number of samples in advance avoids the time and costs associated with either unnecessary or insufficient sampling. In the present study a statistical approach, Gage repeatability and reproducibility (Gage R&R), has been applied to two techniques for measuring boron in glass: induction-coupled plasma optical emissions spectroscopy (ICP-OES) and electron probe microscopic analysis (EPMA); to quantify and compare their precision, and identify the different sources of measurement system variation. Information has also been obtained on the measurement systems' accuracy. The precision of both methods was found to be acceptable for typical glass analysis situations, although the precision of the EPMA method was significantly lower than the ICP-OES method (95% confidence). The number of replicate measurements needed in order to reliably detect a difference of 0-5% B2O3 is practical for most situations: three replicates for the ICP-OES method and six replicates for the EPMA method. Measurement system variation in the ICP-OES technique was found to be primarily due to repeatability rather than reproducibility. The EPMA method was found to have bias and linearity issues. It was concluded that the ICP-OES method has acceptable precision, and the EPMA method has been shown to be a viable alternative provided steps are taken to correct the bias and linearity. The information obtained on the precision of both measurement systems allows better data driven decision-making through the use of appropriate sample sizes.