Design of an Electric Bolt Tightening Robot Using Wavelet Neural Networks and Feedback Linearization

In industrial settings, achieving high-precision bolt tightening is essential to guaranteeing the dependability and safety of products. Inadequate torque management could result in inefficiency or mechanical issues. Standard bolt-tightening robots frequently have trouble adjusting to different kinds of bolts, joint properties and operation environments. This can result in inaccurate torque management and possible overtightening or loosening. The Wavelet Neural Networks (WNNs) provide superior pattern recognition and learning capabilities while feedback linearization makes up for nonlinearities in the tightening process to ensure more accurate control and lower error margins. Our research designed an electric tightening robot using a method based on feedback linearization and WNN. Initially, the dataset was gathered from the Research Institute’s railway basis. A ballastless track was used for the railway in the test site, and WJ-8-type fasteners were executed. Once the data has been normalized using min–max normalization, pre-possess it to ensure it is aligned with track fasteners. The proposed system leverages WNN for adaptive control, allowing the robot to adjust to varying bolt properties and environmental conditions dynamically. Feedback linearization is employed to linearize the tightening process’s nonlinear dynamics, improving the torque control’s precision. The robot demonstrates significant improvements in tightening accuracy, overshoot, RMSE, MAPE and operational efficiency through simulations and experimental validation. This research provides a foundation for developing advanced robotic systems in automated assembly lines, offering potential applications in the aerospace, automotive and heavy machinery industries.