Open Collections

Abdar Esfahani, Mobin

University of British Columbia

The increasing complexity and precision requirements in Computer Numerical Control (CNC) machining necessitate advanced methods for system identification, trajectory generation, and process optimization. This thesis addresses these challenges by developing integrated techniques for accurate modeling of CNC machine dynamics, neural network-based trajectory prediction, and real-time smooth trajectory generation. A robust system identification method is developed to model the closed-loop dynamics of feed drive dynamics, friction effects, and controller characteristics. Utilizing Frequency Response Functions (FRF) and iterative pole-search algorithms, the proposed approach identifies transfer functions for both linear and rotary axes. A neural network architecture with a Bidirectional Long Short-Term Memory (Bi-LSTM) networks is designed to predict trajectory profiles and cycle times directly from G-code inputs. In CNC machining, trajectory generation involves interpolating through input G-code points to create the actual motion paths of the machine tool. However, these trajectory generation algorithms are hidden within the proprietary software of commercial CNC systems. The proposed neural network model overcomes this limitation by capturing complex interpolation behaviors without relying on proprietary algorithms. Experimental results demonstrate the prediction of corner trajectories and block-by-block cycle times, enabling the reconstruction of continuous and smooth toolpath trajectories. An advanced real-time trajectory smoothing method is proposed. An improved Finite Impulse Response (FIR) filter-based method dynamically adjusts dwell times and selectively skips points based on contour error thresholds, managing velocity overlaps and optimizing machining speed and accuracy. Additionally, a novel Half-Sine filter-based approach generates kinematic profiles with infinite continuity in acceleration and deceleration phases, mitigating vibrations and suppressing structural resonances. Combining the Half-Sine filter with FIR filtering achieves smoother transitions and shorter cycle times while maintaining contour precision. The proposed methodologies enhance machining precision and efficiency, providing solutions for process optimization. The research contributes to the field by addressing the limitations of traditional models, enabling accurate trajectory prediction and smooth motion generation in CNC machining.

Text

2024-12-16

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/89878

Mechanical Engineering

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/4.0/