Hydraulic power is a renewable energy with a very advantageous price, and its flexibility of use coupled with its storage possibilities easily enables to meet the peak demand. Refurbishment of old hydraulic power plants is an important topical issue and the upgrading of the power stations of the years 1940 to 1960 can considerably increase both efficiency and power with minimal costs. The current challenge for the turbine manufacturers is the accurate flow prediction in the whole turbine and across a broad operating range, taking into account the unsteady features of the flow. For further developing flow computation tools, the major hydraulic turbine manufacturers : ALSTOM Power Hydro, GE Hydro, VATECH Escher Wyss, VOITH-SIEMENS Hydro have joined Electricité de France and EPFL, in the European initiative framework EUREKA n° 1625. This project has been financially supported by CTI – Swiss Commission for Technology and Innovation. The aim of the project was to analyze the flow within the draft tube of a turbine theoretically, experimentally and numerically, and the present work is a part of the project. In this general framework, the objective of the thesis was to set up PIV measurements for analyzing the flow behavior in the draft tube of a Francis turbine. Several enhancements of the PIV technique have been carried out for acceding to the 3D velocity field, both in single and two-phase flows, in the cone and at the outlet of the draft tube. Due to the complex geometry of the draft tube, particular attention has been given to calibration, allowing obtaining 3% uncertainty on the measured velocity field. In two-phase flow, corresponding to partial load operation in cavitation conditions, two measurement methods have been set up, allowing simultaneous measurement of the velocity field and volume of the vapors-core vortex. The first one relies on the 2D PIV method: two cameras focused on the same measurement zone provide separately the shape of the vapors core and the velocity of fluorescent seeding particles in the flow. The second method, 3D PIV, is based on the separation on the same image of the fluorescent seeding particles from the shape of the vapors core, taking advantage of background lighting in a specific wavelength. Using two cameras in stereoscopic arrangement allows, in this case, obtaining the third component of the velocity. Adaptive image processing algorithms have been developed in both cases for detecting the vapors core boundary and calculating its geometrical characteristics in the measurement section. These experimental developments have allowed studying several phenomena and characterizing several flow types in the draft tube. For this draft tube, the efficiency characteristic shows a severe drop close to the best efficiency point, and this part has been particularly addressed. The 3D velocity field has been measured at the runner outlet and diffuser outlet for a flow rate range of ± 20 % of the best efficiency point flow rate, and these results allowed explaining the source of the hydraulic losses that lead to the drop of the efficiency characteristic. A second issue concerns the spatial distribution of the non-uniformity of the velocity field at the runner outlet, synchronous with the runner blades passage. The velocity field development in the cone of the draft tube from the runner outlet to the cone outlet has been determined and the mixing length of the sheared flow has been studied. In low charge operating conditions (70% of the best efficiency point flow rate), relevant results of unsteady velocity field and related vapors volume are obtained for different pressure levels in the draft tube, which correspond to different cavitation conditions from single-phase flow to vortex-circuit interaction in two-phase flow. The shape of the vortex core has been determined and a parametrical model is proposed. Its influence on the flow, as well as its stability, has been evaluated depending on the pressure level. Particular attention has been paid to the characterization of the interaction of the synchronous part of the vortex-induced pulsations and the circuit. These results have been used to calibrate and validate flow computation codes and they form a database available to the project partners. An example is given of the validation of the vortex flow simulation at partial load in single-phase flow conditions. Unlike the usual data comparison based on wall pressure signals, the velocity field and vortex filament are also assessed.