Protein hydrolysates often contain bioactive peptides with a broad range of properties, such as antimicrobial, antihypertensive or opioid activities. Hydrolysates are therefore of great interest for the formulation of nutraceutics or food safety products. However, bioactive peptides are present in low concentration amongst many other peptides. Therefore, fractionation processes are required to enrich initial mixtures in targeted peptides. For this purpose, ultrafiltration (UF) is most often carried out but in an empirical way. In this work, an original methodology aimed at predicting UF performances in terms of extraction yield and enrichment of targeted peptides of known molar weight is described. The methodology is simply based on classical mass balance equations, a size exclusion analysis and the aminogram of the starting hydrolysate and a UF membrane calibration. The methodology was validated experimentally with an industrial bioactive hydrolysate of caseins (Lactium®), in batch concentration and continuous diafiltration mode with a 3 kDa membrane at different operating conditions (pH, transmembrane pressure, initial concentration, recirculating flow rate). The results show that the methodology enables to predict the yield and the enrichment in 7 different peptides during ultrafiltration with relative errors ranging from 2 % to 30 %. Moreover, results demonstrate that a unique calibration of the membrane can be used in the range of operating conditions studied. Eventually, the methodology was applied to calculate the optimal conditions for enriching peptide of interest from several hydrolysates (alfalfa white proteins, bovine hemoglobin…). For example, the best extraction yield (95%) and enrichment (120 %) of an antioxidant peptide from bovine hemoglobin hydrolysate could be obtained in permeate after 7 volumes of continuous diafiltration with the 3 kDa membrane.