Development of a Scalable Machining Feature Recognition System

In this thesis, various pre-processing and training techniques were applied to improve \nthe performance of a model trained with an existing machining feature recognition approach \nby Yeo et al. using a smaller dataset that more effectively mimics the complexity of \nCAD models used in industry. \n \nA GUI tool was developed to tag faces in CAD models with the corresponding machining \nfeatures which would be necessary to resolve those faces. Using the encoding algorithm \noutlined by Yeo et al., a tool was developed to generate feature vectors from tagged \nCAD models. Two CAD datasets were compiled. First, a dataset of generic CAD models \nwas filtered from a larger dataset compiled by Koch et al., selecting those models which \ncould be manufactured using a 3-axis CNC machine. Second, a dataset of real-world CAD \nfiles used in CNC manufacturing was compiled from models contributed by individuals \nfrom the Unviersity of Waterloo, Hurco Inc. and Perfecto Inc. \n \nUsing the first dataset, three potential improvements to the feature recognition algo- \nrithm developed by Yeo et al. were explored: the incorporation of dropout to improve model \nstability and accuracy, the incorporation of ID3 tree pre-classification to reduce training \ntime by reducing the size of the deep learning dataset without impacting classification \naccuracy, and the incorporation of crossover data generation to improve classification ac- \ncuracy by reducing overfitting due to insufficient training data. It was determined that \nincorporating dropout improved the stability of the model and improved 5-fold cross val- \nidation accuracy. Further, it was determined that incorporating a 2-deep ID3 decision \ntree pre-classification marginally improved classification performance and was effective in \nreducing the size of deep learning training dataset. Crossover data generation did not \nimprove model performance, and so was rejected. Using the model trained on the generic \nCAD dataset, and incorporating 10% dropout and a 2-deep ID3 tree, models from the \nreal-world dataset were classified. This classifier was effective in classifying some simple \nfeatures, but had poor accuracy overall. To improve this accuracy, an incremental learning \ntechnique was applied. The generic model was re-trained using samples from the real-world \ndataset, which improved the classification accuracy of the system.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.