One of the hotspots of fragmentation of plastics in the sea are beaches where sand acts as an abrasive. Artificial fragmentation tests without abrasives have the advantage that the resulting particles can be easily analyzed. However, these test conditions differ from those on the beach. Therefore, tests in which sand or gravel is used are necessary. We use glass bottles that are partly filled with sand and rotated during the experiment. By setting the rounds per minute, the energy input can be adjusted to a defined value. For the analysis of the formed particles a separation from the non-polymeric particles is necessary. A visual selection is not an accurate method and cannot consider the particle size range below 1 mm. The analysis of the material loss from the sample is also not sufficient. Thus, we have developed a separation setup to separate the polymer particles from the sand particles. Subsequently, the particle size distribution can be measured using appropriate methods. The separation setup consists of a funnel in which the particles separate due to their different densities. The required high-density liquid is an aqueous solution of potassium carbonate (K2CO3). The potassium carbonate solution is advantageous as the ions also occur in the sea. In addition, the monovalent potassium ions do not react with the particle surfaces, which can be a problem with divalent ions such as zinc. An almost saturated solution with a density of 1.52 g/cm3 is used. Thus, nearly all polymers, including PET and PE, can be separated from the sand particles. In combination with a previous fragmentation step, recovery rates of at least 98 % were achieved in preliminary tests. With this setup we investigate the fragmentation behavior of plastic pellets in different weathering conditions. In addition, naturally and artificially weathered samples are compared.