REPLY TO HILSDORF AND MULLER'S DISCUSSION OF "COMMENTS ON THE USE OF ROSS ' HYPERBOLA AND RECENT COMPARISONS OF VARIOUS PRACTICAL CREEP PREDICTION MODELS"*

Hilsdorf and MUller's detailed discussion is deeply appreciated. They raise several interesting points which call for further analysis. The discussers claim that a plot of t/C versus ~ (Eq. 3) is preferable to the plot of I/C versus i/ ~ (Eq. 2). This is not true, for three reasons: i) As one can verify by numerical examples, both plots yield essentially the same values of a and b. 2) It is not true that determination of the "final " creep value from the plot of I/C versus I/ ~ gives too little weight to the long time creep data and too much weight to the short-time creep data. The opposite appears to be true. The y-intercept (i.e., point_i/ ~ ÷ 0) is very close (in the horizontal direction) to the points for high t (Fig. 4) and is, therefore, influenced by an error, e I, at points for large t (small l/T) much more than by an error, e~, at points for small ~ (large i/t) which lie far from the intercept (see Fig. 4). 3) The plot of E/C versus E (Eq. 3) may be misleading since it gives a