Structural Reliability Evaluation of Thermoelectric Generator Modules: Influence of End Conditions, Leg Geometry, Metallization, and Processing Temperatures

Geometry optimization for structural reliability and performance of a thermoelectric generator

“…Numerical simulations were carried out to simulate the performance and the thermomechanical stresses of a single couple TEG using the commercial FEA software ANSYS [28]. The probability-of-failure (Reliability, R = 1 − P f ) was estimated using a macro implemented within ANSYS via ANSYS Parametric Design Language (APDL) whose details are presented in our previous work [26]. The thermoelectric and subsequent thermomechanical simulations were sequentially performed in ANSYS where the thermoelectric performance is solved first using SOLID226 elements with only TEMP and VOLT degrees of freedom (DOF), and the resulting temperatures distributions are later imposed on a structural model with SOLID186 elements that have only displacement DOFs (UX, UY, UZ).…”

“…To investigate the potential for plastic strains in metallization, the Cu ICs and Ni barrier are modeled as elastoplastic materials with bilinear isotropic hardening behavior. The material properties and constitutive models of the single couple TEG models considered in this work are similar to those considered in [11,26]. A load resistance between the end terminals is specified in the single couple models using a CIRCU124 element.…”

“…A load resistance between the end terminals is specified in the single couple models using a CIRCU124 element. The details of the TE materials, metallization and the substrate and the thermoelectric and mechanical properties of all the materials used in the FEA are presented in [26].…”

“…Their study did not optimize the stresses or relate stresses to reliability which is necessary to estimate the tradeoff in performance while arriving at truly optimal and reliable TEG configurations. In our previous study [26] we addressed the challenge of linking the TEG leg stresses to reliability using statistical brittle material failure theory and demonstrated that the failure probability of TEG approaches zero at a certain length or aspect ratio (the ratio of length to area L/A of p-and ntype legs). Any improvement beyond that limit will only diminish the performance, especially the power output, and it is necessary to have insights on such aspect ratios to limit the tradeoff in the performance of an optimized TEG configuration.…”

“…The DOE based RSM with facecentered central composite design (CCD) was used to fit second order quadratic regression models for the power output (P), efficiency (η), the probability-of-failure (P f ) and their simultaneous optimization. The factors and ranges for the CCD were chosen based on a two-level full-factorial effects screening study as well as the insights from the one-factor-at-a-time analysis conducted in our previous study [26].…”