Multi-point flexible forming of composite fan blades: a novel data-driven surface compensation method for controlling process-induced deformation

Abstract Objectives Inherent process-induced deformation (PID) compromises the surface accuracy of complex-shaped composite fan blades (CFBs) with variable cross-sections. Conventional methods of iteratively modifying rigid die surfaces are time-intensive and costly, limiting manufacturing efficiency. This study aims to develop a novel data-driven surface compensation method to mitigate PID in CFBs. Methods A multi-point flexible die (MPFD) closed-loop digital control system was developed to adaptively adjust the die surface, achieving precise reconstruction. Additionally, an experiment-guided neural network approach, based on an improved sparrow search algorithm and back propagation neural network, was introduced for feature point cloud prediction. Using small-sample point cloud data, a deviation compensation model was created and validated through manufacturing experiments on a double-diaphragm flexible platform. Results The MPFD system maintained surface errors between −0.071 and 0.054 mm. For initial manufactured blades, root mean square deviations were 0.971 mm at the tip, 3.015 mm at the middle, and 1.689 mm at the root. After compensation, these deviations reduced to 0.252, 0.347, and 0.281 mm, respectively, representing improvements of 74.05%, 90.68%, and 83.36%. Over 93% of deviations fell within ±0.5 mm. Conclusion The proposed data-driven method effectively controls non-uniform PID, providing an efficient solution for high-precision manufacturing of complex composite components. Future work will focus on optimizing algorithm parameters to further enhance accuracy.