RESULTS

Figure 1 shows replicate values from each participant for SRMs 2686, with the certificate values and 2σ limit indicated by a solid line and dashed line, respectively. Most participants' data fell within the 2σ ranges for each of the clinkers, indicating that there was no significant bias among test results. Microabsorption corrections improve estimates of ferrite and periclase, and the lack of a correction resulted in a bias for four participants in the estimation of ferrite. The use of these corrections needs further investigation.

Based upon these data and calculations, a provisional precision and bias statement for ASTM may be developed using the individual phase values pooled across the three clinkers (Table 1). Precision and bias levels are all expressed as percentage points by mass relative to the total clinker. These results will be balloted in the Compositional Analysis Subcommittee (C01.23) and a summary of the findings will be incorporated in an appendix of the test method describing the Rietveld approach for clinker analysis. Comparison with results from earlier round robins [5, 6] indicates an improvement in both within- and between-laboratory repeatability and reproducibility (Fig. 2). This is most likely the result of improved intensity estimates from the method's whole-pattern approach, through the ability to refine lattice parameters and peak shapes. The earlier round robin data were mostly based upon peak area measurements on duplicate scan data. A cement round robin is currently underway to assess the precision and bias for a cement analysis and the influence of microabsorption corrections. Results will be incorporated in a revised XRD standard test.

Figure 2. Multi-laboratory standard deviations (s_R) for alite, belite, aluminate, and ferrite from this round robin (ASTM '03), from Moore [5], and Aldridge [6] shows improvement with the application of Rietveld pattern fitting approach.