[en] This report deals with an evaluation of three pulse irradiation tests, which were carried out in 1996-1998 in the Japanese Nuclear Safety Research Reactor (NSRR) to study the behaviour of high-burnup boiling water reactor fuel under reactivity initiated accidents. The tests, FK-1, FK-2 and FK-3, are evaluated by use of computer simulations. The base irradiation of the test rods to a fuel burnup of 41-45 MW_d/kg U is first simulated with the FRAPCON-3.2 program, in order to establish the pre-test fuel rod conditions. The pulse irradiation tests are then analysed by use of the SCANAIR-3.2 code, which is specifically designed to model the thermo-mechanical behaviour of light water reactor fuel rods under reactivity initiated accidents. The work presented here is primarily intended to examine the correctness and applicability of models used in these computer programs. The validation is done by comparing the calculated fuel rod conditions with measurements made in the tests. The considered tests are well suited for computer code validation, since the test rods were carefully characterized both before and after the pulse irradiation tests. Moreover, the test rods were extensively instrumented, and in-pile measurements of temperatures and deformations as functions of time were made under the pulse tests. In general, the calculated results agree fairly well with measured data from both the base irradiation of the test rods and from the actual pulse tests. The most striking exception is the transient fission gas release, which is underestimated by SCANAIR for FK-1 and FK-2, while it is overestimated for FK-3. Calculated deformations of both the fuel pellet stack and the clad tube under the pulse tests are well in line with measurements, except for the axial deformations in FK-3. This exception is, however, due to unsuccessful measurements, and not a result of model deficiencies. Satisfactory agreement between calculated and measured coolant and clad surface temperatures under the pulse tests is reached only when the extended, two-phase, coolant channel model is used in SCANAIR. With the original single-phase model, clad temperatures are considerably overestimated. The original coolant channel model foresees clad dry-out for all three tests, whereas the extended model excludes dry-out for all cases. The in-pile measurements of clad surface temperature, as well as post-test measurements of clad hardness, show that local dry-out occurred in FK-1 and FK-3, but not in FK-2. Fuel rod survival is predicted for FK-1 and FK-2, in accordance with the test results. Clad tube failure is however erroneously foreseen for FK-3. The survival of FK-3, in spite of its significant clad hoop plastic deformation, is probably due to the fact that the axial load imposed on the clad tube is surprisingly low in this test. The stress biaxiality ratio, σ_zz/σ_θθ is estimated to 0.63, base on the axial-to-hoop ratio of measured residual strains. This is far from the equal biaxial (σ_zz/σ_θθ) stress state, which is expected under pellet-clad mechanical interaction in high-burnup fuel rods