摘要
Based on the recent huge demand in artificial intelligence and big data applications, recent research has increasingly focused on improving the efficacy and reliability of data center design. In this study, mechanical estimation is utilized to assist the engineering development of MLCC placement in PCB layout design. Enhancing the lifetime and reliability of power supplies also plays a critical role. One persistent challenge in power supply, is the mismatch in thermal expansion coefficients (CTE) between different materials, particularly under high temperature operating environments. Moreover, plastic deformation associated with creep progressively accumulates, eventually causing mechanical issues such as solder cracking or delamination, which in turn may accelerate electrical degradation or even result in complete breakdowns. To solve these challenges, this study proposes an integrated framework that combines physics based reliability modeling with machine learning approaches. In study also aim to develop a practical framework that integrates Coffin–Manson lifetime modeling through nonlinear finite element analysis (FEA) with multilayer perceptron neural networks, in order to achieve a more accurate assessment of reliability and to explore novel approaches for extending product life. Solder materials are considered using the Anand model. After validation of the FEA with experimental data, the deviation was found to be less than 20%. The mechanical simulation database comprises 250 sets of data, of which 80% were used for training the neural networks and 10% data for testing, 10% for validation data. This strategy provides multiple robust design options for engineers, bridging mechanical and electrical domains in MLCC placement within PCB layout design. ©2025 IEEE.