In this paper two immersed boundary methods (IBM), specifically a continuous forcing method (CFM) and a discrete forcing method (DFM), are applied to perform direct numerical simulations (DNSs) of heat transfer problems in tubular fluid-particle systems. Both IBM models are built on the well-developed models utilized in momentum transfer studies, and have the capability to handle mixed boundary conditions at the particle surface as encountered in industrial applications with both active and passive particles. Following a thorough verification of both models for the classical Graetz-Nusselt problem, we subsequently apply them to study a much more advanced Graetz-Nusselt problem of more practical importance with a dense stationary array consisting of hundreds of particles randomly positioned inside a tube with adiabatic wall. The influence of particle sizes and fractional amount of passive particles is analyzed at varying Reynolds numbers, and the simulation results are compared between the two IBM models, finding good agreement. Our results thus qualify the two employed IBM modules for more complex applications. (C) 2018 The Authors. Published by Elsevier Ltd.