Support Vector Machine
Assembly Line
Process incidence monitoring in material identification during drilling stacked structures using support vector machine
Drilling of stacks comprising carbon fibre-reinforced polymers (CFRP) and aluminium in a single shot is a typical operation in the assembly of aircraft. This paper proposes a novel approach to identify incidences in CFRP/Al stack drilling with 94 % classification accuracy based on signal features and support vector machine (SVM). This enables the application of adaptive drilling which aerospace industry tries to introduce, and cutting parameters (cutting speed, feed) are automatically adjusted based on features extracted from signals obtained to achieve optimal machining. The t-distributed stochastic neighbour embedding (T-SNE) algorithm is applied to evaluate the separability and invariance of features with the significant influence of tool wear. Collinear analysis and hierarchy dendrogram are conducted to test the accuracy and robustness of the new approach, and a distance-based feature pruning is then proposed to compress data while maintaining the algorithm’s performance. The proposed SVM model achieves an accurate and reliable incidence identification, thereby enhancing the decision-making for adaptive drilling in machining stacked structures.