Considering the strict regulations on the transport sector emissions, predictive models for engine emissions are essential tools to optimize high-efficient low-emission internal combustion engines (ICE) for vehicles. This aspect is of major importance, especially for developing new combustion concepts (e.g. lean, pre-chamber) or using alternative fuels. Among the gaseous emissions from spark-ignition (SI) engines, unburned hydrocarbons (uHC) are the most challenging species to model due to the complexity of the formation mechanisms. Phenomenological models are successfully used in these cases to predict emissions with a reduced computational effort. In this work, uHC phenomenological model approaches by the authors are further developed to improve the model predictivity for multiple variations including engine design, engine operating parameters, as well as different fuels and ignition methods. The model accounts for uHC contributions from piston top-land crevice, wall flame quenching, oil film fuel adsorption/desorption and features a tabulated-chemistry approach to describe uHC post-oxidation. With the support of 3D-CFD simulations, multiple novel modelling assumptions are developed and verified. The model is validated against an extensive measurement database obtained with two small-bore single-cylinder engines (SCE) fuelled with gasoline-like fuel, one with SI and one with pre-chamber, as well as against data from two different ultra-lean large-bore engines fuelled with natural gas (one equipped with a pre-chamber and one dual-fuel with a diesel pilot). The model correctly predicts the trends and absolute values of uHC emissions for all the operating conditions and the engines with an accuracy on average of 11.4%. The results demonstrate the general applicability of the model to different engine designs, the correct description of the main mechanisms contributing to fuel partial oxidation, and the potential to be extended to predict unburned fuel emissions with other fuels.