AOBL-PSO-BPNN-based electrical and thermal performance prediction of TSV-based 3D inductor

Through-Silicon Via (TSV)-based three-dimensional (3D) inductors have gained considerable attention in 3D integrated circuits due to their high inductance density. However, their compact structure results in significant heat accumulation, making electro-thermal coupling analysis essential. This study first employs ANSYS and COMSOL finite element software to investigate the impact of individual design parameters on the electrical and thermal characteristics of TSV-based 3D inductors. To improve the efficiency and accuracy of electro-thermal analysis, an Adaptive Opposition-Based Learning Particle Swarm Optimization-BP Neural Network (AOBL-PSO-BPNN) method is proposed for rapid prediction of electro-thermal characteristics of TSV-based 3D inductors. Compared with finite element simulation results, the proposed method achieves a maximum error of less than 3.23% while reducing memory resources by 78.13% and computation time by 93.67%, enabling efficient electro-thermal analysis of TSV-based 3D inductors.