Session: CIE-15/16: SEIKM Joint Topics

Paper Number: 117184

117184 - Data Fusion Cognitive Computing for Characterization of Mechanical Property in Friction Stir Welding Process 

High intricacy of the friction stir welding (FSW) process and severe plastic deformation typically result in serious part quality concerns, including thinning, groove, and kissing bond. These defects, in turn, impact mechanical properties (e.g., Young's modulus and yield strength), thereby deteriorating the performance of a joint part. Machine learning methods such as feature-based support vector machines and end-to-end deep neural networks are recently coupled with in-situ sensing to link process parameters with the final properties, yet the low accuracy and precision, poor sample efficiency, and lack of interpretable learning limit their capabilities for reliable characterization. This research introduces hyperdimensional cognitive computing (HCC) that mimics human brain functionalities to fuse power, torque, and force data and provide robust, sample-efficient, and explainable learning for process-property characterization in FSW. Data augmentation is integrated with HCC to tackle the imbalanced data problem. The experimental results show that HCC seamlessly fuses in-situ data to predict the ultimate tensile strength with an F1-score of 0.899 +/- 0.048, which is 21.579%, 47.779%, 1.933%, and 68.139% better than support vector machine, logistic regression, Naive-Bayes and multilayer perceptron, respectively. Besides the superior prediction capability, HCC symbolic representation reveals that measured force is the critical factor in the characterization of ultimate tensile strength. The proposed HCC is shown to be effective for data fusion and single-pass learning and eliminates the necessity of costly and long retraining in various manufacturing processes such as 3D printing and bio-fabrication.

Presenting Author: Farhad Imani University of Connecticut

Presenting Author Biography: Farhad Imani is an Assistant Professor in the Department of Mechanical Engineering at the University of Connecticut (UConn). He received his Dual-title Ph.D. in Industrial Engineering and Operations Research from the Pennsylvania State University in 2020. He is an expert on sensing, control and automation, computational intelligence, and cognitive learning with applications in advanced manufacturing. His research interests encompass mechatronics, robotic integration, signal and image processing and data analytics, system identification and uncertainty quantification, and modeling and analysis of systems with multi-field couplings. Imani’s ongoing research includes designing cognitive digital twins (sponsored by DepEd).

Authors:

Danny Hoang University of Connecticut
Ruimin Chen University of Connecticut
Debasish Mishra University of Connecticut
Surjya K Pal Indian Institute of Technology Kharagpur
Farhad Imani University of Connecticut