Open Collections

Automated design and implementation of Kalman observer for spindle torque estimation in CNC machining

University of British Columbia

Milling is a subtractive manufacturing technique, where the material is continuously removed from a workpiece until the desired part shape is obtained. Heavily used in the automotive and aerospace industries, Computer Numerical Control (CNC) milling machines occupy a large chunk in the manufacturing process. The current research focuses towards machining and machine tool monitoring systems that are more self-sufficient and self-adjusting to adapt the processes to machine tools. Cutting torque delivered by the machine tool spindle is one of the key sensory signals for machining process monitoring. This thesis presents a method that automatically reconstructs cutting torque from motor current commands generated by the servo controller of the machine tool. To estimate the cutting torque from commanded spindle current, the dynamics between torque to current relationship must be modeled and compensated. The thesis first presents an automated identification of spindle dynamics using data-driven system identification methods. The frequency response function (FRF) of the spindle dynamics is measured manually using CNC internal diagnostic tools. The identified FRF is then automatically converted to a state-space model using the Eigensystem Realization Algorithm (ERA). To reduce overfitting, an optimal threshold is applied to the ERA method to limit the identified system order. And to ensure the stability of the identified system, unstable eigenvalues of the system are removed using Schur decomposition. The identified system is then augmented such that the unknown torque input is modeled as a state changed by a random process noise. This augmented system is used to create a Kalman Observer, which compensates the spindle dynamics and estimates the torque from the spindle nominal current. The Kalman Observer is tuned automatically by estimating the noise covariance values using machining simulations. The method was eventually validated on a Quaser UX 600 industrial CNC system.

Text

2020-12-08

Attribution-NoDerivatives 4.0 International

http://hdl.handle.net/2429/76698

Mechanical Engineering

University of British Columbia

UBCV

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