Chemical engineering systems usually involve multiple spatio-temporal scales, grouped into different levels, from the molecular scale of reactants to the industrial scale of reactors. Molecular dynamics (MD) simulation is one of the most fundamental methods for the study of such systems, but it is too costly and hence formidable for simulating large-scale behavior directly. However, there are two great potentials in extending this method. First, the logic and algorithms of traditional MD simulations can be generalized from the material level to higher levels since the elements of each level are all discrete in nature, and can be well defined, allowing an MD-style simulation based on different elements. Second, MD simulations can be accelerated by realizing the structural consistency among the problem, model, software and hardware (the so-called EMMS paradigm). These two potentials give possibilities to engineer the method of MD simulation to deal with the whole spectrum of chemical engineering phenomena. In this review, we summarize our discrete simulation studies to explore such potentials, from the establishment of a general software and hardware framework, to the typical applications at different levels, including the reactions in coal pyrolysis, the dynamics in virion, the atomic behavior in silicon at millimeter scale, and finally continuum flow. The possibility of engineering MD simulation into a virtual experiment platform is discussed finally. (C) 2014 Elsevier Ltd. All rights reserved.