So far thermoset based composites have been used in primary aircraft structures, owing to their good performance and vast research on the subject. Recently thermoplastic based composites also gained an increased attention and there is a push towards adopting these for primary aircraft structures as well. This interest for thermoplastic composites in aeronautical structures is driven by the material's advantages over thermoset parts in joining methods, damage tolerance, sustaining, shelf live, recyclability, or chemical stability. The butt-joint feature is one of the novel joining techniques made possible by using thermoplastic composites, joining technique which can used to assemble stiffened panels by co-consolidating different, typically flat, laminates. With the potential use of thermoplastic structures in primary aircraft structures being relatively recently considered, the available literature on the damage behaviour of thermoplastic stiffened panels under compression is still scarce, to say the least, especially when using this novel butt-joint feature. In this project, the building-block approach is used to develop a robust modeling technique able to model skin-stringer separation in butt-joined thermoplastic panels under compression. Here the modeling strategy is developed in a step-by-step fashion, going from a material characterisation coupon scale, up to a sub-component scale. The separation was first modeled in a DCB coupon used to describe the Mode I material fracture toughness, then in a single-stringer specimen with skin-stringer de-bond and finally in a multi-stringer panel with skin-stringer de-bond. The base of the damage modeling strategy was defined by studying the load-displacement response of a DCB specimen using multiple models. First, an experimental load-displacement response of a DCB specimen was captured by an analytical approach, with which the accuracy of the numerical models was compared. The numerical approaches tried were based on cohesive zone models (contact and element based) and on the Virtual Crack Closure Technique, with the latter proving to be better suited for this particular task. The VCCT separation damage model was preferred here since it allowed the use of coarser meshes than the cohesive zone models, with a superior overall accuracy as well. As the material of the studied DCB specimen was identical with the one identical of the single-stringer specimen and multi-stringer panel aforementioned, the VCCT model used to model separation was also kept identical when these larger structures were studied. The accuracy of the FE model developed for the single-stringer specimen was also assessed by comparing its load-displacement curve and failure load with the existent experimental data, improving the modeling technique for its use in the multi-stringer panel. As an FE model which correlated well with the experimental data already existed within Fokker, using a cohesive surface based damage model, the shortcomings of this damage model was shown by applying it to the lower scale DCB specimen. Implementing this damage model for the DCB specimen led to a very poor load-displacement correlation, owing to the altered fracture toughnesses values with respect to the measured one. With the shortcomings of the old separation damage model shown, the VCCT one was implemented in this already existent model, which gave a poor correlation with the experimental data. Since the VCCT damage model used needed to be kept unmodified, the butt-joints of the single stringer were refined and the elements used for these areas were updated in a new FE model, these changes leading to a near excellent experiment-simulation correlation in terms of load-displacement curves and failure loads. The modeling technique developed was then used further to study skin-stringer separation in the multi-stringer panel. A sensitivity study on the panel's compressive response with and without the de-bond damage modeled revealed the high sensitivity of its skin buckling pattern. The occurring skin buckling pattern appeared to be highly influenced by the skin's imperfect non-flat shape, asymmetric load introduction, ends boundary conditions, as well as by the de-bond damage itself. This very sensitive skin buckling pattern was attributed to the specific combination of panel's skin bays boundary conditions and skin bays aspect ratio, which likely gave a buckling coefficient governed either by a 3 or 4 longitudinal buckling half-waves. The sensitivity study on the different possibly occurring buckling patterns revealed that the ones promoting de-bond growth from a certain side of the stringer have a strong influence on the panel's failure load, decreasing the strength of the panel by as much as 25% for a 4 half-waves skin buckling pattern, the most likely to occur however having a 3 half-waves one. Similarly, the most significant strength increase was also due to a 4 half-waves skin buckling pattern, the difference between this and the aforementioned one being the stringer side from which the de-bond grew. Furthermore, this sensitivity study was also used to select a best blind prediction based on which the experimental test plan was defined. The test measurement set-up to correlate and validate the developed FE model was also defined based on this selected best blind prediction and it was successfully used during the experimental test to accurately capture the panel's compressive response. The test-model correlation showed that the FE model was able to predict the compressive response of the panel with great accuracy in what concerns its load-compression response, failure load, buckling behaviour, skin out-of-plane deflections and the qualitative aspect of the de-bond growth. These qualitative aspects of the de-bond growth were its skin buckling pattern influenced location, de-bond growth front shape and its growth soon after skin buckling occurred. A initially very good correlation in terms of load-strain response was also achieved, this correlation being highly influenced by two aspects. First source of test-model strain correlation mis-match was considered to be the significant difference in the panel's skin buckling load, while the second one was the model's poor prediction of the panel's stiffness drop associated with de-bond growth. The de-bond growth was overestimated by the FE model prediction, while the panel stiffness loss as a result of the de-bond growth was underestimated. Overall, the presented modeling strategy to model and study skin-stringer separation in thermoplastic butt-joint stiffened panel under compression proved to be robust. The robustness of this method comes mainly from the same modeling strategy used at all the addressed scales and with the same VCCT damage model using the measured material fracture toughnesses. This implies that the presented method could be successfully transferred to similar panels with similar skin-stringer damage, provided that skin-stringer separation is the main failure mode.