The pressure drop and collection efficiency are generally the most relevant parameters when evaluating the performance of gas-solid cyclone separators. In optimization procedures, mathematical cyclone models are used to find the best possible compromise between the pressure drop and the collection efficiency, based on the design preferences. Although several geometry optimization studies on single cyclone configurations have been reported, multiple cyclones in series remain relatively unexplored in this regard. Highly sophisticated models are described in the literature, but researchers mostly use lower fidelity models due to the shorter computational times involved. Also, when more sophisticated models are used, the authors tend to employ surrogate modeling approaches, which reduce the computational times but add another layer of uncertainty. In addition, several authors have optimized either the efficiency or the pressure drop, usually by adding a constraint on the other parameter. In contrast, herein, we describe the multi-objective optimization of three cyclones in series based on high fidelity computational fluid dynamics (CFD) cyclone modeling. A fully automated methodology solving this problem is evaluated, with the COBYLA optimization method and an Eulerian-Eulerian six-phase two-way gas-solid CFD approach simultaneously minimizing the pressure drop and maximizing the efficiency. No surrogate modeling is used here, which means that the results of the CFD simulations are directly used in the optimization procedures. The results obtained indicate that the optimized trios of cyclones outperform the conventional Stairmand (high efficiency) and Lapple (moderate pressure drop) geometries. The minimization of the emission resulted in 30 times less solids being emitted compared with the Stairmand trio. When minimizing the energy consumption, the pressure drop was 50% that of the Lapple geometry. When balancing the pressure drop and emission several results were considerably better than both geometries simultaneously (e.g. 33% less emission than Stairmand combined with 30% less pressure drop than Lapple). This demonstrates that the multi-objective optimization of cyclones in series delivers excellent results and is highly feasible in industrial timescales without compromising the fidelity of the mathematical cyclone model used. (C) 2017 Elsevier B.V. All rights reserved.