Intelligent Design Method of Thermal Through Silicon via for Thermal Management of Chiplet-Based System

In this article, an intelligent design method of thermal through silicon via (TTSV) for thermal management of a Chiplet-based system is proposed based on the finite element method (FEM), back propagation-neural network (BP-NN) model, and particle swarm optimization (PSO) algorithm. In order to analyze the effect of design parameters of TTSV (TTSV pitches, radius of TTSV unit, and thickness of oxide layer) on the thermal distribution of a Chiplet-based system, the detailed model is simulated by FEM. The mapping relationships between the design parameters and thermal distribution indices are described by BP-NN models according to the simulation data. In addition, according to the desired performance of the Chiplet-based system with TTSV, the multiobjective optimization criterion is formulated, and then the design parameters are optimized by the modified PSO algorithm. Based on the obtained design parameters, the FEM simulation is aimed at validating the effectiveness of the proposed method. The simulated peak temperatures (PTs) of three layers (340.81, 334.76, and 314.85 K) well agree with the desired ones (340, 335, and 315 K), which implies that the proposed method can effectively optimize the design parameters of TTSV to control the thermal distribution of Chiplet-based system. Therefore, the proposed intelligent design method can achieve the thermal management of the complex Chiplet-based system.