Reliability prediction of electronic components based on physical of failure with manufacturing parameters fluctuations

Abstract

Reliability prediction based on the physics of failure (PoF) methodology involves examining the physical variables that impact the performance parameters of electronic components, developing mathematical models to describe the evolution of these parameters, and predicting the components' reliable operational lifespan. However, the current PoF model does not account for the influence of manufacturing parameters, such as material properties and structural characteristics, which limits the accuracy of reliability predictions. Therefore, establishing a PoF model that incorporates manufacturing parameters is a critical challenge in enhancing reliability prediction accuracy. To overcome this limitation, an improved PoF model incorporating manufacturing parameters is proposed in this study. The study examines how the manufacturing parameters influence the PoF model, then develops an adapted PoF model that incorporates these factors for improved predictive accuracy. Then, a parameter estimation method based on Long Short-Term Memory (LSTM) is proposed, with the Hybrid Bat Algorithm (HBA) employed to adaptively optimize the network's parameters. Finally, the effectiveness of the proposed method is demonstrated through a case study on an electromagnetic relay. Compared to the actual lifespan, the reliability prediction model incorporating manufacturing parameters accurately estimates the relay's lifetime, achieving deviation of only 5.8 % at 100,000 cycles, thereby verifying the model is feasibility and effectiveness.