[en] Highlights: • Automated input calibration and code validation using genetic algorithm is presented. • Predictions generally overlap experiments for individual system response quantities (SRQs). • It was not possible to predict simultaneously experimental maximum flow rate and oscillation period. • Simultaneous consideration of multiple SRQs is important for code validation. - Abstract: Validation of system thermal-hydraulic codes is an important step in application of the codes to reactor safety analysis. The goal of the validation process is to determine how well a code can represent physical reality. This is achieved by comparing predicted and experimental system response quantities (SRQs) taking into account experimental and modelling uncertainties. Parameters which are required for the code input but not measured directly in the experiment can become an important source of uncertainty in the code validation process. Quantification of such parameters is often called input calibration. Calibration and uncertainty quantification may become challenging tasks when the number of calibrated input parameters and SRQs is large and dependencies between them are complex. If only engineering judgment is employed in the process, the outcome can be prone to so called “user effects”. The goal of this work is to develop an automated approach to input calibration and RELAP5 code validation against data on two-phase natural circulation flow instability. Multiple SRQs are used in both calibration and validation. In the input calibration, we used genetic algorithm (GA), a heuristic global optimization method, in order to minimize the discrepancy between experimental and simulation data by identifying optimal combinations of uncertain input parameters in the calibration process. We demonstrate the importance of the proper selection of SRQs and respective normalization and weighting factors in the fitness function. In the code validation, we used maximum flow rate as the SRQ of primary interest. The ranges of the input parameter were defined based on the experimental data and results of the calibration process. Then GA was used in order to identify combinations of the uncertain input parameters that provide maximum deviation of code prediction results from the experimental data. Such approach provides a conservative estimate of the possible discrepancy between the code result and the experimental data.