Springback prediction using point series and deep learning

Yang Bingqian,Yuanyi Zeng,Hai Yang,Mariluz Penalva Oscoz,Mikel Ortiz,Frans Coenen,Anh Nguyen
DOI: https://doi.org/10.1007/s00170-024-13632-6
IF: 3.563
2024-04-25
The International Journal of Advanced Manufacturing Technology
Abstract:Abstract One of the main challenges that prevent wide adoption of Single Point Incremental Forming (SPIF) is the geometric accuracy of the process resulting primarily from the effect of springback. There are various expedients that can be adopted to address this, but one of the most common is tool path correction. The challenge is then how best to predict springback so as to implement tool path correction. It is established that springback, to a large extent, is related to the geometry of the part to be manufactured. The proposed mechanism uses a novel point series representation to capture local geometries that then form a global bank of geometries for general use. Each point series can then be associated with a predicted springback value generated using deep or machine learning. Experiments are reported using a Long Short Term Memory (LSTM) model coupled with a Multilayer Perception Network (MLP), and a Support Vector Machine (SVM) regression model. A best R2, “Coefficient of Determination”, of 0.9181 was obtained indicating that the proposed approach provided a realistic solution to the current limitations of SPIF.
engineering, manufacturing,automation & control systems
What problem does this paper attempt to address?
The paper primarily aims to address a critical issue in Single Point Incremental Forming (SPIF) technology—springback prediction. SPIF is an unconventional metal sheet forming technique where the metal sheet is clamped along its periphery and deformed locally and incrementally under the action of a round tool head. Although this technique is cost-effective for small batch production or prototyping, the geometric accuracy of the formed parts is often affected by the springback effect. The method proposed in the paper aims to improve the geometric accuracy of parts in the SPIF process through tool path correction. Specifically, the authors developed a novel point sequence representation method to capture the local geometric features of the parts and used it to construct a global geometric library. Each point sequence is associated with the springback value predicted by a deep learning model. To validate the effectiveness of this method, the researchers employed a Long Short Term Memory (LSTM) network combined with a Multilayer Perception Network (MLP), as well as a Support Vector Machine Regression (SVM) model for experiments. The experimental results show that the proposed LSTM-MLP model can effectively predict springback, with the best Coefficient of Determination (\(R^2\)) reaching 0.9181, demonstrating that this method can practically address the current limitations of SPIF technology. In summary, the main objective of the paper is to propose a deep learning-based springback prediction method to improve the geometric accuracy of parts manufactured using SPIF technology.