In this research, effect of nanoparticles dimeter on free convection of aluminum oxide-water was investigated in a cavity by single phase and two phase models. The range of Rayleigh number is considered 105-107 in volume fractions of 0.01 to 0.03 for nanoparticles with various diameters (25, 33, 50 and 100 nm). Given that the two phase nature of nanofluids, necessity of modeling by this method is increasing. Single phase approach (in contrary of two phase) for nanofluids is based on that the behaviors of each two solid phase (nanoparticles) and liquid phase (base fluid) are completely similar. In this study, Eulerian-Eulerian approach and mixture model was used given that Brownian motion and thermophoresis effects. Brownian motion and thermophoresis creates under influences of volume fraction gradient and temperature gradient, respectively that cause to creating slip between nanoparticles and base fluid; thus, kind of non-uniformity creates on behavior between nanoparticles and base fluid. This non-uniformity leads to significant effects on results of two phase modeling that creates better agreement to single phase modeling with experimental results. Results indicate that heat transfer decreases with increasing diameter and volume fraction of nanoparticles. Also, effect of nanoparticle diameter on flow and heat transfer is tangible. Manuscript profile

In the present research, an artificial neural network was designed and conducted to thermodynamically analyze performance variables of two-shaft radial flow gas turbine model GT185. To do this, firstly the needed tests were conducted at different operating conditions and the essential variables like temperature, pressure, rotational speed, mass flow rate which totaled 17 inputs were recorded. Then, using the relations regarding radial flow turbines and the laboratory results, performance variables including compressor, gas turbine and free turbine power and efficiency and finally the cycle heat efficiency were calculated. After calculation of these variables for all laboratory data, a neural network was designed and tested using Matlab software toolbox in order to facilitate the obtaining of performance variables in different operating conditions. In this network, highest errors absolute values of training, verification and testing data were 0.32, 0.86 and 1.39 respectively. Error value of the produced function of sample laboratory results and manual calculations was less than 0.1% Manuscript profile

a Manuscript profile