Artificial neural network as a solution to predict 3-D workspace distortion in real-time

Machine workspace distortion is a time-variant nonlinear process. The authors present an artificial neural network based hybrid model for computer numerical control (CNC) machine system 3-D errors monitoring and distorted workspace tracking. The goal is to improve machining quality in real time. Analytical and learning approaches are used in parallel to solve this problem. An artificial neural network is a particularly useful modeling methodology for a system whose underlying property is too complicated to be modeled using classical methods. In many aspects, including fault and noise tolerant capability, a neural network approach outperforms the classical methods. Another main advantage of the neural network is the representation of the entire workspace with a single model. A multi-sensory system is incorporated with the model for real-time system monitoring and data acquisition. This research provides a feasible and robust model for a feedback control system to perform real-time machining accuracy improvement and quality control for prismatic workpieces.<>