An ensemble data-driven fuzzy network for laser welding quality prediction

This paper presents an Ensemble Data-Driven Fuzzy Network (EDDFN) for laser welding quality prediction that is composed of a number of strategically selected Data-Driven Fuzzy Models (DDFMs). Each model is trained by an Adaptive Negative Correlation Learning approach (ANCL). A monitoring system provides quality-relevant information of the laser beam spectrum and the geometry of the melt pool. This information is used by the proposed ensemble model to asist in the prediction of the welding quality. Each DDFM is based on three conceptual components, i.e. a selection procedure of the most representative welding information, a granular comprehesion process of data and the construction of a fuzzy reasoning mechanism as a series of Radial Basis Function Neural Networks (RBF-NNs). The proposed model aims at providing a fuzzy reasoning engine that is able to preserve a good balance between transparency and accuracy while improving its prediction properties. We apply the EDDFN to a real case study in manufacturing industry for the prediction of welding quality. The corresponding results confirm that the EDDFN provides better prediction properties compared to a single DDFM with an overal prediction performance > 78%.

[1]  A. Saah,et al.  Sensitivity and Specificity Reconsidered: The Meaning of These Terms in Analytical and Diagnostic Settings , 1997, Annals of Internal Medicine.

[2]  Dianhui Wang,et al.  Fast decorrelated neural network ensembles with random weights , 2014, Inf. Sci..

[3]  Huanhuan Chen,et al.  Regularized Negative Correlation Learning for Neural Network Ensembles , 2009, IEEE Transactions on Neural Networks.

[4]  Tony J. Dodd,et al.  Active sensorimotor control for tactile exploration , 2017, Robotics Auton. Syst..

[5]  Nasser L. Azad,et al.  Optimally pruned extreme learning machine with ensemble of regularization techniques and negative correlation penalty applied to automotive engine coldstart hydrocarbon emission identification , 2014, Neurocomputing.

[6]  Huan Liu,et al.  Feature Selection for High-Dimensional Data: A Fast Correlation-Based Filter Solution , 2003, ICML.

[7]  George Panoutsos,et al.  Iterative information granulation for novelty detection in complex datasets , 2016, 2016 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE).

[8]  George Panoutsos,et al.  Granular computing neural-fuzzy modelling: A neutrosophic approach , 2013, Appl. Soft Comput..

[9]  Xin Yao,et al.  Simultaneous training of negatively correlated neural networks in an ensemble , 1999, IEEE Trans. Syst. Man Cybern. Part B.

[10]  Neil D. Lawrence,et al.  An integrated probabilistic framework for robot perception, learning and memory , 2016, 2016 IEEE International Conference on Robotics and Biomimetics (ROBIO).

[11]  Andrzej Bargiela,et al.  Granular clustering: a granular signature of data , 2002, IEEE Trans. Syst. Man Cybern. Part B.

[12]  George Panoutsos,et al.  Fuzzy uncertainty assessment in RBF Neural Networks using neutrosophic sets for multiclass classification , 2014, 2014 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE).

[13]  George Panoutsos,et al.  Interval Type-2 Radial Basis Function Neural Network: A Modeling Framework , 2015, IEEE Transactions on Fuzzy Systems.

[14]  John Q. Gan,et al.  Low-level interpretability and high-level interpretability: a unified view of data-driven interpretable fuzzy system modelling , 2008, Fuzzy Sets Syst..

[15]  Anders Krogh,et al.  Neural Network Ensembles, Cross Validation, and Active Learning , 1994, NIPS.

[16]  Chuen-Tsai Sun,et al.  Functional equivalence between radial basis function networks and fuzzy inference systems , 1993, IEEE Trans. Neural Networks.

[17]  George Panoutsos,et al.  Development of a parsimonious GA-NN ensemble model with a case study for Charpy impact energy prediction , 2011, Adv. Eng. Softw..

[18]  George Panoutsos,et al.  A Data-driven fuzzy modelling framework for the classification of imbalanced data , 2016, 2016 IEEE 8th International Conference on Intelligent Systems (IS).