Developing representative thermo-mechanical power loss models for engineering systems, including gearboxes, is crucial for optimizing the performance of electric vehicles. These models, when computationally efficient, can significantly reduce design costs by minimizing the number of experiments needed to identify unknown model parameters. In this case, the model design and measurement plan can be evaluated using identifiability techniques, which refer to the ability to estimate unknown parameters in lumped models based on measurements uniquely [1]. In this study, the identifiability of parameters in a developed model is examined by performing singular value decomposition (SVD) on the output sensitivity matrix [2, 3]. SVD provides a tool to analyze the sensitivity matrix by decomposing it into orthogonal components. This decomposition helps show if the parameters are correlated by finding locally identifiable parameters and highlighting gaps in parameter identifiability that can complicate model accuracy. A state-space model is first derived from the governing differential equations to achieve this. Later, the sensitivity matrix of the measurement equation with respect to unknown parameters, including the heat capacities and thermal resistances, is constructed. The structural identifiability of the model is then evaluated by decomposing the sensitivity matrix using SVD and analyzing the singular values. Additionally, the right singular vectors are examined to assess parameter correlations. The results indicate that singular values vary with different parameter definitions, suggesting that model identifiability is sensitive to parameterization. Furthermore, the singular values depend on the parameter range.