Research on fatigue reliability assessment of engine cylinder head based on neural network

Abstract

Fatigue reliability (FR) evaluation is crucial for extending the service life and improving the reliability of automotive engines. This study proposed a general framework for assessing high cycle FR based on the back propagation neural network (BPNN). Using a diesel engine cylinder head (CH) as an example, the stochastic finite element analysis was implemented by considering the uncertainties of materials and loads. Subsequently, the stochastic finite element calculation process was replaced with the BPNN model to expedite fatigue life prediction. Ultimately, the reliability assessment of the CH was conducted, taking into account the influence of the load sequence. The research findings indicated that the reliable life t0.95 of the CH considering the load sequence was approximately 75 h, representing a 7.1% increase compared to the linear cumulative damage. During the service life of the CH, the failure probability sensitivity factor for the gas force load exceeded 0.9, offering an important reference for optimizing FR.