Enhanced fatigue crack growth rate prediction in alloy steels using particle swarm optimized neural network

Abstract

In the manufacturing sector, fatigue crack growth (FCG) poses a critical challenge to the structural integrity and safety of components, with significant implications for human safety and economic impact. The relationship between stress intensity factor range (ΔK) and FCG rate (da/dN) is often nonlinear, even within the Paris region, influenced by factors like stress ratio (R-ratio), threshold values of ΔK (ΔK_th) and da/dN (da/dN_th), critical stress intensity factor (K_c), specimen geometry, mechanical properties, and alloy compositions. These complexities render traditional empirical methods inadequate for accurate FCG rate predictions. This study introduces a Particle Swarm Optimized Neural Network (PSONN) model, trained and tested across a range of alloy steels, including 316, 316 L, 316 L(N), AISI 301, AISI 302, 304, St 980, Q345qc, St-4340, and Fe 430D. The PSONN model outperforms traditional methods by delivering superior accuracy and reducing error in FCG rate prediction, highlighting its potential for improved safety and reliability in design.