Fatigue crack growth evaluation is very important for the damage tolerant design of machines or structures. It has been widely recognized that fatigue crack growth rate can be well expressed in terms of the effective stress intensity factor range $ΔK_eff$, based on the crack closure concept. In estimating $ΔK_eff$, the most important and difficult part is to determine consistently the crack opening load $K_op$.
In this study, a neural network approach proposed by previous researchers is modified to determine more precisely the crack opening level from differential displacement signal curve. Various $K_op$ determination methods are evaluated and compared in a quantitative manner by using new criteria proposed here. The modified neural network approach is found to be very promising for $K_op$ determination.
In addition, it is observed that the 2/π and 2/π0 methods proposed by Paris and Donald for estimation of $ΔK_eff$, provide better results, compared with conventional $ΔK_eff$ estimation.
