Ref. Year Type More details Key findings [38 ] 2019 Numerical ANSYS Fluent, 2D axisymmetric steady flow, – SST, 1-phase 1-species Increasing raises ER, after a certain value performance deteriorates [39 ] 2020 Numerical Thermodynamic model, annular mixing layer Optimal depends on working condition [40 ] 2022 Experimental Ejectors with different nozzle and area ratios tested over wide ranges of operating conditions Up to 4 “optimal” observed; these values were affected by both working condition and area ratios [41 ] 2022 Mix Experimental testing and validation of proposed model, ANSYS Fluent, 2D axisymmetric steady flow, – SST, 1-phase 1-species, compound choking criterion Compound choking allowed to determine variation in secondary flow choking position under several suboptimal [42 ] 2019 Numerical ANSYS Fluent, 2D steady flow, – ; model coupled with anodic pressure drop formula, 1-phase 2-species Optimal range for (3.00–3.54) and (1–3) [43 ] 2021 Numerical 3D steady flow, RNG – , 1-phase 2-species is the first geometrical parameter to be optimized[51 ] 2014 Numerical Thermodynamic model of ejector coupled with semiempirical stack model Definition of two dimensionless parameters to guide ejector design [52 ] 2017 Experimental Ejector designed and tested at constant load and in fast transient condition Anode gas recirculation rate ranging from 40% fuel utilization per pass at 25 A stack current to 64% fuel utilization per pass at 160 A stack current [53 ] 2022 Numerical ANSYS Fluent, 2D steady flow, – SST, 1-phase 2-species Determined order of influence of geometrical parameter: (i) Low current (110 A) (ii) Middle current (275 A) (iii) High current (412.5 A) Nozzle throat length [54 ] 2013 Numerical Thermodynamic model Inlet primary flow temperature affects ejector entrainment ratio and component efficiency [55 ] 2019 Mix ANSYS Fluent, 2D axisymmetric steady flow, comparison between RNG – and – SST RNG model shows higher accuracy than SST; optimal ° and mm when stack works at its rated power [56 ] 2020 Numerical ANSYS Fluent, 2D axisymmetric steady flow, – SST, coupled with a pressure drop through anode model ER can be influenced by anode inlet temperature, relative humidity, and differential pressure [57 ] 2020 Numerical OpenFOAM, 3D transient flow, RNG , 2-phase 3-species Dynamic responses during power variations results from velocity differences between the primary and the secondary flow; increase of nitrogen mass fraction promotes total ER, while it reduces hydrogen ER [58 ] 2020 Numerical COMSOL, 3D steady flow, coupled with MATLAB/Simulink hydrogen recovery system model A lower ejector temperature is disadvantageous in removing the moisture content of the recirculated hydrogen gas, thus in practical applications; the hydrogen inlet temperature/pressure must be carefully controlled [59 ] 2022 Numerical Thermodynamic model of a 2-phase CPM ejector ER increase from 0.47 to 1.14 as mixing area ratios range from 1.0 to 1.2 under the given conditions [60 ] 2015 Numerical Integrated lumped parameter-CFD approach, ANSYS Fluent, 2D axisymmetric steady flow, – SST The model can be used for studying off-design conditions, where ejector component efficiencies are not constant [61 ] 2016 Numerical ANSYS Fluent, 2D axisymmetric steady flow, RNG – , 2-phase 1-species, optimization through genetic and evolutionary algorithm is the crucial parameter in ejector performance[62 ] 2021 Numerical ANSYS Fluent, 2D steady flow, RNG – , 1-phase 2-species Humidity and temperature of the secondary flow have a noticeable influence on the performance of the ejector [44 ] 2017 Numerical Thermodynamic model, hybrid fish swarm algorithm Optimization efficiency increased with respect to genetic algorithm [45 ] 2018 Numerical Multiobjective evolutionary algorithm coupled with a surrogate model based on CFD simulations and are the most important geometrical variable; entrainment ratio can be increased up to 110% and 35%, for air and CO2, respectively[46 ] 2023 Numerical ANSYS Fluent, 2D axisymmetric steady flow, response surface methodology Priority order in optimizing ejector geometry: [47 ] 2014 Numerical 2D axisymmetric steady flow, RNG – , 1-phase 2-species, artificial neural network and genetic algorithm to obtain optimal geometry Optimal ; optimal ; optimal °; optimal [48 ] 2021 Numerical Automated CFD workflow, ANSYS Fluent, RNG – , 2-phase 1-species, Gaussian process regression machine learning model The algorithm can be used to efficiently explore ejector designs with mean average errors between 0.07 and 0.1 [49 ] 2022 Numerical ANSYS Fluent, 2D steady flow, realizable – , optimization via adjoint method ER increased by around 37% [50 ] 2022 Numerical MATLAB and experimental dataset of a steam-centered ejector are applied to train the ANN model of a steam ejector using three different algorithms LM model yielded the best agreement; the effect of the outlet area ratio is less important with respect to throat area ratio