Rapid prediction of thrust force coupling scale-span model and revised ANN in drilling CFRPs

Abstract

This study develops an integrated methodology to rapidly predict the thrust force with a tapered drill-reamer (TDR) by coupling a scale-span model and revised artificial neural networks (ANN) in drilling carbon fiber reinforced polymers (CFRPs). First, the optimum mesh size of the scale-span finite element (FE) model of CFRPs was optimized to enhance simulation efficiency on the premise of ensuring accuracy in drilling. Then, an order-driven FE computation approach was first proposed to improve computing efficiency for batch samples and maximize utilization of the available computing resources. Modeling and solving of the weight indices of material property parameters (MPPs) and machining parameters for the thrust force were first carried out entirely based on a feature selection model. A multi-layer revised ANN architecture model which considers the material properties of CFRPs and the corresponding initial weight indices was first designed for the thrust force prediction in Python software. Finally, drilling experiments involving T700S-12K/YP-H26 CFRPs specimens with different machining parameters were carried out, which more than 25 prediction results of the fresh samples showed that the established ANN prediction model with a 16-18-18-18-16-1 architecture is highly prediction precision, and the maximum absolute deviation is only 4.56% with the comparisons of experiments.