Storage rack structures are pre-engineered light weight steel structures with standard details which can be quickly assembled because all building components arrive on site pre-fabricated in line with the agreed design. Since they are made using thin-walled perforated steel sections, mostly cold-formed, with mechanical joints, generally partial-strength, semi-rigid, and with a low structural redundancy, these structures are vulnerable to any action capable to induce a local damage of a member or connection. When heavily loaded, the local damage may spread progressively, generating an overall collapse or disproportional damages. Difficulties in predicting the structural behavior of storage pallet racks are amplified by the specific geometry of the structural components: members made by high slenderness thin-walled and open-section profiles (hence prone to global, local and distortional buckling problems), flexible beam-to-upright and baseplate connections with a non-linear behavior. Due to their peculiarities, additional modelling and design rules are required for these non-traditional steel structures and reference cannot be made to usual structural design recommendations and standards. In case of Selective Pallet Rack (SPR) structures, except for the earthquakes, collision of forklift trucks or other moving equipment with front upright is considered as one of the most frequent causes of local failure, with potential to develop into a progressive collapse. In the study, the robustness of SPR structures under accidental loading situations involving collision with forklift truck is assessed using both notional upright removal and explicit forklift impact approaches. Structural configurations are varied to consider different connection properties (upright base and beam-to-upright) and brace arrangements (spine bracing, top plan bracing). SPR structures are susceptible to global failure, especially if the spine bracing is in just a few spans and the rigidity of connections is low. The explicit modelling of the forklift impact provides the most accurate results, as the effects associated with the forklift impact are not properly captured when the response is evaluated using the dynamic analysis and notional upright removal approach. Non-linear static pushdown analysis can provide satisfactory results at the least computational effort, but the dynamic increase factors may require corrections.