Heavy fuel oil (HFO) as a fuel in industrial and power generation plants ensures the availability of energy at economy. Coke and cenosphere emissions from HFO combustion need to be controlled by particulate control equipment such as electrostatic precipitators, and collection effectiveness is impacted by the properties of these particulates. The cenosphere formation is a function of HFO composition, which varies depending on the source of the HFO. Numerical modelling of the cenosphere formation mechanism presented in this paper is an economical method of characterising cenosphere formation potential for HFO in comparison to experimental analysis of individual HFO samples, leading to better control and collection. In the present work, a novel numerical model is developed for understanding the global cenosphere formation mechanism. The critical diameter of the cenosphere is modelled based on the balance between two pressures developed in an HFO droplet. First is the pressure (Prpf) developed at the interface of the liquid surface and the inner surface of the accumulated coke due to the flow restriction of volatile components from the interior of the droplet. Second is the pressure due to the outer shell strength (PrC) gained from van der Walls energy of the coke layers and surface energy. In this present study it is considered that when PrC ≥ Prpf the outer shell starts to harden. The internal motion in the shell layer ceases and the outer diameter (DSOut) of the shell is then fixed. The entire process of cenosphere formation in this study is analysed in three phases: regression, shell formation and hardening, and post shell hardening. Variations in pressures during shell formation are analysed. Shell (cenosphere) dimensions are evaluated at the completion of droplet evaporation. The rate of fuel evaporation, rate of coke formation and coke accumulation are analysed. The model predicts shell outer diameters of 650, 860 and 1040 µm, and inner diameters are 360, 410 and 430 µm respectively, for 700, 900 and 1100 µm HFO droplets. The present numerical model is validated with experimental results available from the literature. Total variation between computational and experimental results is in the range of 3–7%. © 2016 Taylor & Francis.