This work presents a model-based optimization approach to determine the configuration of a petroleum refinery for a new (grassroots) or an existing site for heavy oil processing of atmospheric distillation unit residue. We develop a high-level superstructure representation for the refinery configuration that encompass many possible alternatives comprising the technologies and their interconnections. The superstructure is postulated by decomposing it to include representative heavy oil processing scheme alternatives that center on the technologies for atmospheric residual hydrodesulfurization (ARDS), vacuum residual hydrodesulfurization (VRDS), and residual fluid catalytic cracking (RFCC). We formulate a mixed-integer nonlinear optimization model (MINLP) appended with logic propositions devised from experience and heuristics to aid solution convergence. Implementation on a case study of Kuwaiti refineries demonstrate the applicability and practicality of the proposed approach when validated against existing real-world refinery configurations. © 2019 Elsevier B.V.