Integrated Additive Manufacturing Trajectory Control and Repair Management for Spiral Bevel Gear Using Fractional-Order PID and Adaptive Fuzzy Neural Inference System

Design optimization of mechanical system components, such as spiral bevel gears, is important due to their complex geometry and design constraints. However, the complex geometry of these components introduces great challenges in manufacturing and maintenance. As inspired by the need to improve gear maintenance practices through advanced technologies, the integration of additive manufacturing (AM) is an opportunity to revolutionize gear repair processes, which potentially reduces downtime and extends gear life cycles. The research objectives concentrate on constructing an adaptive control algorithm for gear repair using metal powder bed fusion technology and assessing these algorithms against state-of-the-art methods. The repair phase provided an approach with a control system combining fractional-order PID (FOPID) controllers with adaptive neuro–fuzzy inference systems (ANFIS) to manage the complex and nonlinear process of gear repair using AM. By integrating these techniques in ANFIS, neural networks can adapt to varying conditions and fuzzy logic can efficiently handle imprecise data and mimic human decisions. This integration produces a robust and adaptive system that can handle the uncertainties inherent in gear inspection and repair. The FOPID controllers outperform traditional PID controllers in terms of rise time, settling time, and steady-state error. The ANFIS model achieves good overall tracking of amplitude variations and accurate classification of repair methods. The adaptive control system utilizing FOPID controllers with ANFIS demonstrates superior performance in handling the complex repair process. The framework established for gear repair offers a blueprint for future advancements in predictive maintenance and automated repair systems. Its potential applications extend beyond gear systems to other essential mechanical components.

Abstract licence: CC BY-NC-ND

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.