Click for next page ( 33

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement

Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 32
32 performance. However, the two models that included all nine Thresholds of the strand samples both predicted that three of these strand Thresholds for two of these QC tests (Contact Angle Mea- samples would be judged to be acceptable, in contrast to the surement after Lime Dip and Change in Corrosion Potential) six of nine that would be judged acceptable based on the pull- have been developed. This was done based on prediction in- out test itself. Although these multiple-predictor regression tervals for the regression calculated from the available data, models do appear to be more effective than the individual QC a minimum criterion on the mortar pull-out stress adopted tests, the strongly conservative nature of the conclusions by NASPA, and a selected confidence level. The available regarding acceptable performance is related to the large pre- data, consisting of the mortar pull-out results and QC test re- diction intervals. sults for the included strand sources, were not sufficient to There are a number of possible reasons, as follows, that the allow threshold determination for the other two methods prediction intervals are not smaller: with the same constraints. The thresholds that were possible were calculated in a The QC test methods are inadequate--It is possible conservative manner to ensure adequate bond performance. that the QC methods do not measure a property of the However, the 90% confidence prediction interval thresholds strand that is sufficiently strongly linked to bond per- on the change in corrosion potential and contact angle test formance. It is also possible that the QC tests measure would suggest that of the nine samples (two of which came only a part of what determines bond and that other fac- from the source) included in the program, only two and three tors exist that are equally or more important. This may of the nine samples would be judged to be acceptable based mean that, while an individual QC test result is not suf- on these test methods, respectively. Although any conservative ficient to determine strand bond performance by itself, approach for predicting a response based on an empirically the QC result must be combined with the result from developed relationship should be expected to underestimate another test. that response, this is in contrast to the six of nine that would be The QC test methods or the mortar pull-out test method judged acceptable based on the pull-out test itself. The inabil- were susceptible to large scatter--All four of the recom- ity to develop thresholds for two QC test methods and the mended test methods produced regression models that strongly conservative nature of the thresholds that were predicted average results for the range of QC test results developed has resulted from the large prediction intervals obtained that spanned the mortar pull-out threshold. calculated for these relationships. However, the difference between the predicted average result Regression with multiple predictors also has been performed and the lower bound on the prediction interval is strongly to determine if results of selected QC methods could be influenced by the scatter about the best-fit line. Recall that combined to better predict bond. The following three com- the line fit shown on the regression plots is the average binations showed the best correlation, based on the adjusted result (i.e., half of the pull-out test results will be above this coefficient of determination (R2 adj.): line and half will be below). A large amount of variation in the data used in the regression analysis will lead to a large Weight LOI & Contact Angle Measurement after Lime prediction interval. Therefore, despite the fact that the Dip & Change in Corrosion Potential, method may test a property strongly linked to bond, if that Contact Angle Measurement after Lime Dip & Change in property is difficult to measure with good precision, defi- Corrosion Potential, and nition of threshold may be difficult. It should also be noted Contact Angle Measurement after Lime Dip & Organic that this large scatter may be the result of significant local Residue Extraction (for stearate based residues). variations in the bond properties of the strand. It has been suggested by other researchers that the concentration of The adjusted coefficients of determination for each of these lubricant residue is highly variable even within a single combinations were higher than the coefficients of determi- spool and that significant differences may exist in strand nation for the single-predictor regression models. separated by as little as 20 ft. Although some attempt was Thresholds for multiple-predictor regressions can not made to track the proximity of strand samples used in each be determined using the same procedure used for single- of the various test methods in the Screening Round, this predictor regressions. Instead, the lower bound on the pre- was not possible in the Correlation Round, and any such diction interval must be calculated for each combination of variation has become inseparably combined with variations test results. One of the models (for the combination of Contact in the test methods. Angle Measurement after Lime Dip & Organic Residue The sampled sources were too closely grouped in terms of Extraction) for stearate-based residues accurately predicted bond performance--In any regression analysis, greater