In the preliminary studies which have been described in the first report, it is recognized that reliability analysis is a useful tool in order to estimate a reduction of safety due to the hot straining embritticment (HSE). As the main purpose in the report is to realize the unfavorable effect of HSE, the model adopted has only two surface flaws, the one detected and the other undetected.
There are, however, many flaws in actual welded joints, and adequate inspections and following repairs of detected flaws are repeated many times until flaws are not detected.
Expanding and modifying the method mentioned above, this paper presents analytical procedures to evaluate reliability of actual welded joints as a function of joint length under some simplicities. Applying the weakest link model, probabilities of failure for no repair, repair with or without HSE are calculated. It is pointed out that detected flaws should not be repaired if those sizes are smaller than a certain allowable value which depends on the statistics of critical COD, undetected flaws, thermal effects and so on.
In addition, Monte Carlo simulations which generate actual circumstances of welded joints as precisely as possible, are also performed in order to verify the proposed method, and it is shown that the analytical method is effective from the results of numerical examples.