The flow velocity profiles in most of the central air-conditioning pipelines are, in general, not fully developed flow and difficult to obtain the accurate flow rates by flowmeters, which are used for measuring average velocity. Especially for being at the outlet of an elbow, the accuracy of flow rate by measurement is quite low. Therefore, there are some limitations for measurements of flow rate and velocity profile by the present flow measuring technologies. The objective of this study was to establish an approach on accurate predictions of velocity profiles at different measured locations of central air-conditioning pipelines for nonuniform flow measurements by simulations of computational Fluid Dynamics (CFD). All the velocity profiles will integrate as a database for predictions by neural network algorithm for smart measurement further. In the present work initially, international experiments were employed to validate the accuracy of CFD approach. The calculations were carried out by different turbulence models. The results compared with the experimental data by Realizable k-ε turbulence model with less computing resources have great agreements. Realizable k-ε turbulence model was, therefore, determined for the predictions of central air-conditioning pipeline. According to various pipings and pipe sizes, the results for three cases show that the velocity profiles in the pipelines would not be symmetrical and has strong secondary flow. Therefore, all of the flow profiles would be integrated and analyzed as a database and assist to get accurately the measured locations of ultrasonic flowmeters. Further, this database will be combined with algorithm of artificial neural network for smart predictions.