| Περίληψη: | Metal-based AM processes, although mature enough for small-batch production, are seldom first-time-right. This means that the user needs to print at least a few parts to test and validate that the process is performing as intended, while modifying both the geometry and the process plan (process parameters, path planning strategy etc.). This thesis aims at using pre-existing data (from metrology, real time process monitoring, inspection of the final part) and knowledge generated through additively manufacturing different components and apply them in an intelligent way to assess the manufacturability of a new part. A twofold methodology is proposed that evaluates the manufacturability of a part in terms of shape accuracy and structural integrity. Although, the inputs for both individual models are coming from different sources, however, they are combined in a common data structure, also being registered in time and place, in order to facilitate the data processing and to understand how the process parameters and the part design can lead to defects (cracks-non manufacturable features). Prior to the modelling, detailed analysis has been conducted to prove which of the available datasets could be used for predictors of the machine learning model, ensuring the high accuracy of the results. The first model aims to give as output the probability for the initiation or the propagation of the crack and the exact location by utilizing metrics and indicators that rely on existing knowledge. In addition, the second model utilizes the .stp format of the 3D CAD so as to detect critical geometries and to compare them with experimentally obtained thresholds that indicate if a part is manufacturable as is or design modifications are needed. Both models are related to the machine specifications and the processed material. However, re-calibration of model is possible by following the described methodology. The intuitive user interface of the developed software tool will give an insight about the manufacturability of the part, indicating where the defect is found, either manufacturing or shape.
|
|---|
