A fast decision-making method for process planning with dynamic machining resources via deep reinforcement learning

Application of Multisensor Information Fusion Technology in the Measurement of Dynamic Machining Errors of Computer Numerical Control (CNC) Machine Tools

“…Although the individual neurons are extremely structured and limited in function, numerous neurons achieve extremely colorful behaviors and are widely used. Artificial neural networks have good prospects for development for the following reasons: first, neural networks avoid the shortcomings of expert systems, that is, the need to previously establish a knowledge system, while neural networks obtain knowledge by learning samples to get the weights and thresholds that contain the laws and only need to store the weights and thresholds to achieve the simulation of input and output relationships and to achieve the prediction function, that is, the nonlinear mapping function; second, the neural network has good online learning and parallel operation function, good fault tolerance for parameters, and good inference, especially for nonlinear information, imperfect information, and inaccurate information and has very good simulation ability, as a new modeling technology is widely used [19]. The establishment of CNC machine tools with accurate mathematical models is the focus of error compensation, but with a complex structure of CNC machine tools, access to accurate mathematical models is very difficult, plus the environment of the CNC machining center is always changing, resulting in machining errors constantly changing, so the compensation value of the error is not a simple addition or subtraction of individual components.…”