Two phase gas-liquid flows occur in a wide variety of situations, i.e. , in water supply systems, petroleum industry, pressurized tunnels and pipelines, culverts and water conduits. Air entrainment into a pipeline is the result of vortices at water intakes. Due to the severity of two-phase air-water flow at hydraulic systems, estimation of the pressure and pressure loss in such systems is of great importance. Most of the former studies have focused on two-phase gas-liquid flow in micro-channels and small pipelines. Lack of studies exists on pipelines with larger dimensions as used in hydraulic engineering. This situation becomes even more complicated because the exact shapes and locations of the regions occupied by air and water are unknown, as the air-water interface deforms. The most common two-phase flow regimes that are attainable in pipelines are bubbly, slug, wavy, stratified, and annular flow. The mutual effects of gas and liquid specifications make the behavior of two-phase flows complex and periodic with different parameters involved. Among different two-phase gas-liquid flow patterns, slug flow is the most complex flow with extremely variation of pressure. Therefore, more researches on this flow pattern seems to be necessary. In this study, the variations of pressure and pressure loss coefficient are estimated using artificial neural networks (ANN) and adaptive neuro fuzzy inference system (ANFIS). Recently, these models have been used in different practical situations to predict complex nonlinear relations among the parameters. The present model also combines particle swarm optimization (PSO) with adaptive neuro fuzzy inference system, namely hybrid PSO-ANFIS(HPA), thereby optimizing the results. For this purpose, the antecedent parameters of ANFIS estimating pressure and pressure loss coefficient were optimized using PSO algorithm whereas the consequent parameters were optimized using least-square algorithm. ANFIS model includes complex processes with different parameters involved, thereby produces so many rules. To overcome this complication, fuzzy clustering preprocessing was employed to decrease the number of fuzzy rules and to increase the efficiency of ANFIS model. To these models with subtractive clusters, a PSO algorithm was applied to optimize the clustering parameters and to control the fuzzy- if- then rules in subtractive clustering. Another PSO algorithm was also employed to tune the fuzzy rule parameters associated with fuzzy if-then rules. The SUBCLUST fuzzy clustering technique was used for estimating the variation on pressure in two-phase air-water flows. Performance of the models was evaluated using root mean square error (RMSE) and coefficient of determination (R 2 ). Results show that the ANFIS model is more accurate compared to the empirical correlations presented in the literature. Furthermore combining PSO algorithm and subtractive fuzzy clustering resulted in improving the results. Key words: Two phase flow, Pressure loss, Adaptive Neuro Fuzzy Inference System, Fuzzy Clustering, Optimization