This poster presents a novel hybrid model for the optimization and prediction of surface waviness of components produced by wire and arc additive manufacturing. It consists of an artificial neural network optimized by the rank-Gaussian particle swarm optimization (PSO) and combined with an RGPSO algorithm. The novelty is that RGPSO is not only used to optimize the hyperparameters of the ANN model improving its prediction performance, but also to solve the problem of optimizing surface waviness.
Maya Abi Akl, Florence Marie Muller, Jens Maebe, Meysam Dadgar, Boris Vervenne, Nadia Withofs, Christian Vanhove, Stefaan Vandenberghe
abstract
Extending the axial field-of-view (aFOV) of PET scanners to achieve higher sensitivity and improved image quality has been the focus of several research groups worldwide. Little consideration has been given to the high cost of increasing detector coverage (8-10 versus 2-3 MEuro for standard PET). Despite needing faster PET scans without compromising quality, most hospitals cannot afford to acquire/maintain such high-cost scanners. Moreover, the high
sensitivity (enabling shorter acquisitions) of
Lightweight fiber-reinforced polymer composites offer a promising alternative to metal-based engineering solutions. However, understanding and predicting their complex nonlinear mechanical behavior poses challenges due to intricate microstructures and experimental limitations. Developing constitutive models for accurate Finite Element (FE) simulations demands significant expertise and time investment.
This research proposes the integration of Artificial Intelligence (AI) into constitutive mater
Drag to adjust frozen columns
Alert
Lorem ipsum
Okay
Prediction and Optimization of Surface Waviness in Wire and Arc Additive Manufacturing Using ANN+PSO/PSO Hybrid Model
Loaded
Title
Prediction and Optimization of Surface Waviness in Wire and Arc Additive Manufacturing Using ANN+PSO/PSO Hybrid Model
@
{name}
Poster ID
3
Session
1
authors
Jun Cheng, Wim De Waele
abstract
This poster presents a novel hybrid model for the optimization and prediction of surface waviness of components produced by wire and arc additive manufacturing. It consists of an artificial neural network optimized by the rank-Gaussian particle swarm optimization (PSO) and combined with an RGPSO algorithm. The novelty is that RGPSO is not only used to optimize the hyperparameters of the ANN model improving its prediction performance, but also to solve the problem of optimizing surface waviness. Experimental data of waviness from literature are used to define training data, and the K-fold cross-validation strategy is applied to train, test, and validate the prediction model. The results indicate that the performance of the developed RGPSO/ANN model reaches a higher accuracy than a PSO/ANN model and an ANN model. The prediction errors of the RGPSO+ANN are within ±0.05 mm for all samples, while the PSO+ANN model has some errors that are outside of this range. The accuracy of the PSO/ANN prediction model is quantified as 0.019, 0.990, 0.013, and 3.46% in terms of RMSE, R^2, MAE, and MAPE. The RGPSO, PSO, and other optimization algorithms are applied to optimize the WAAM process parameters to reach the minimal value of waviness. The lowest value for waviness is obtained with the RGPSO algorithm and is equal to 0.1631 mm.