Please use this identifier to cite or link to this item: https://hdl.handle.net/10419/336838 
Year of Publication: 
2024
Citation: 
[Journal:] Journal of Intelligent Manufacturing [ISSN:] 1572-8145 [Volume:] 37 [Issue:] 1 [Publisher:] Springer US [Place:] New York, NY [Year:] 2024 [Pages:] 31-43
Publisher: 
Springer US, New York, NY
Abstract: 
Incremental hole flanging by industrial robots requires an accurate estimation of process forces to ensure the safe use of robots and develop tool path strategies that enable defect-free forming. Existing analytical models for force prediction in single-point incremental forming (SPIF) usually predict only the maximum process force, which is insufficient for real-time force prediction necessary for process control. This research investigates the application of neural network models for real-time prediction of the process forces in hole flanging by incremental forming. Experiments and finite element simulations (synthetic data) serve as training data to compare the performance of four time series machine learning (ML) algorithms: a nonlinear autoregressive model with exogenous inputs (NARX), a convolutional neural network (CNN), a long short-term memory (LSTM), and a hybrid CNN-LSTM neural network in predicting real-time forces. The experimental forces predicted by the NARX had a regression ( ) of 0.998, while the other ML algorithms achieved greater than 0.75, indicating strong correlations between the predicted and measured data. All the ML algorithms trained on synthetic data achieved greater than 0.98. Surrogate models built by the integration of synthetic to experiment data led to a decline in the performance of the ML models compared to models composed of only experimental data. The NARX model exhibits superior performance in the investigated scenarios and can be applied to predict online process forces while minimizing experimental effort through the utilization of synthetic data.
Subjects: 
Real-time force prediction
Incremental hole flanging
Surrogate model
CNN
LSTM
NARX
Hybrid CNN-LSTM
Persistent Identifier of the first edition: 
Creative Commons License: 
cc-by Logo
Document Type: 
Article
Document Version: 
Published Version

Files in This Item:
File
Size





Items in EconStor are protected by copyright, with all rights reserved, unless otherwise indicated.