Motion position prediction and machining accuracy compensation of galvanometer scanner based on BWO-GRU model

This paper proposes a new prediction model based on an advanced deep learning approach that integrates the beluga whale optimization (BWO) algorithm with the gated recurrent unit (GRU) to enhance the laser machining accuracy. This fusion aims to achieve precise prediction of galvanometer scanner motion trajectories. Initially, the collected dataset is subjected to normalization and denoising procedures, followed by division into distinct training and testing sets. Subsequently, the GRU model is harnessed for effective trajectory planning and prediction due to the intricate nature of the galvanometer scanner's dynamic model and the profound influence of previous moments' positional, velocity, and acceleration information on its motion trajectory. Furthermore, in consideration of factors such as neuron count, learning rate, and data batch size, which significantly impact prediction performance within the GRU model, the implementation of the BWO algorithm is proposed to optimize the model's structure. The proposed BWO-GRU model is benchmarked against the BWO-BP (Back Propagation) model and the BWO-LSTM (Long Short-Term Memory) model through experimental evaluation. Encouragingly, the results underscore the superiority of the BWO-GRU model, manifesting exceptional predictive precision and streamlined runtime efficiency. Ultimately, the trajectory prediction based on the BWO-GRU model helps to improve the machining accuracy of the laser galvanometer, signifying its potential for advancing laser-based manufacturing processes.