Learning the Part Shape and Part Quality Generation Capabilities of Machining and Finishing Processes Using a Neural Network Model

Automatically acquiring knowledge of manufacturing process capabilities from existing data is essential for automated process selection in digital and cyber manufacturing. In this work, we present a neural network model to automatically learn the capabilities of discrete manufacturing processes such as machining and finishing from design and manufacturing data. Concatenating a 3D Convolutional Neural Network (3D CNN) with an artificial neural network, the combined model can learn the part shape and part quality generation capabilities of the manufacturing processes. Specifically, the proposed method takes the voxelized part geometry and part quality information as inputs and utilizes a mixed neural network model (3D CNN + artificial neural network) to predict the manufacturing process label as output. The manufacturing process capability knowledge embedded in the neural network model is scalable and updatable as new manufacturing data becomes available. We present an example implementation of the proposed method with a synthesized manufacturing dataset to illustrate how the method enables automatic manufacturing process selection. The high prediction accuracy shows its predictive strength for use in Computer Aided Process Planning (CAPP).