A novel approach to boring feed force estimation using Pearson-correlation-based weighted interpolation and virtual sampling

Feed force has a direct impact in boring operations where surface quality and geometric accuracy are critical. Since the boring operation is the final operation, if the feed force cannot be controlled, productivity will decrease while secondary operations, scrap amount, and cost will increase. Therefore, the feed force needs to be modeled and estimated effectively. Due to time and cost constraints, obtaining datasets from numerous experiments is often unfeasible. While this situation causes small sampling problems, establishing highly successful models and making predictions is a great challenge. The aim of this investigation is to precisely model the boring feed force (Ff) and obtain good prediction performance utilizing weighted interpolation for virtual sample generation. Contrary to the methods used in the literature, it was suggested to use Pearson correlation coefficients (PCC) in determining the weights. According to the findings, it is found that the proposed method significantly improves the drilling feed force estimation accuracy and outperforms other weighting methods. The Huber regressor showed higher performance among the algorithms compared both with and without virtual samples.