Open Collections

Ji, Zeen

University of British Columbia

The monitoring of machining process, tool damage and machine health are crucial to the part quality while avoiding damage to the machine tool. The cutting force provides important information about the state of the process and cutting tool because it correlates to the physics-based mathematical model of the process. This thesis presents methods for the estimation and prediction of cutting forces from the spindle and feed drive motor current data extracted from CNC. The spindle motor current is correlated to inertia, friction in the bearings, and cutting torque as a disturbance in closed-loop spindle speed controller. The Frequency response function (FRF) of the spindle servo drive is measured via a built-in CNC function. The state-space model of cutting torque disturbance of the spindle drive is modeled. Kalman Filter and Regularized Convolution (RD) are used to compensate the disturbance effects of electrical and mechanical dynamics to widen the bandwidth of cutting torque estimation from spindle motor current. Automatic tuning of Kalman Filter’s covariance and RD’s regularization factor is investigated by checking the Neural Network-based online tuning of Kalman Filter and the off-line L-curve tuning of both estimators. The proposed methods are experimentally illustrated in milling operations. It is shown that Kalman Filter can estimate the cutting forces on-line during machining. RD is used only in off-line estimation due to its computational cost, although it has higher accuracy due to its compensation at a wider frequency range. L-curve only achieves offline tuning due to its computational cost. Neural Network can auto-tune the estimator, but training the network requires high computational efforts. The feed drive motor current commands are used to predict the cutting forces in Cartesian coordinate system. The previous research illustrated that Kalman filter estimates the periodic cutting forces up to about 150 Hz bandwidth, provided that the FRF is not position-dependent, and friction is compensated. This thesis presents the use of average motor current to estimate the average cutting forces after friction compensation. The average forces are used to monitor the variation in cutting force coefficients which is used to calibrate the process simulation models in a digital twin environment.

Text

2022-10-24

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/82975

Mechanical Engineering

University of British Columbia

UBCV

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