Summary: | 碩士 === 國立中央大學 === 機械工程研究所 === 83 === A torque sensor based on system identification is proposed in
this thesis. The proposed torque sensor is constructed of two
encoders and signal processing firmware.The encoders are
magnetic type with sinusoid outputs and connected with a
torsion bar. With this arrangement, the phase difference
between these two encoders is proportional to the torque loaded
on the torsion bar.To estimate the phase difference,the signal
processing firmware acquires data,use identification algorithm
to construct a 2'nd order FIR model, and then use the FIR model
parameters to estimate the phase difference. To varify the
proposed approach, both simulation and experimental results are
provided. The simulation data are constructed of sinusoid waves
added with some distortion terms.To simulate the nonlinearity,
the distortion terms are triangular waves. To simulate the
disturbance of noise, the distortion term is pseudo random
noise. To simulate the unbalance of analogy circuite cause by
temperature or ageing,constant shift is added to sinusoid waves.
The non-perfect data are feed to both proposed approach and
Kunio's method for evaluation and comparison. The experiment is
performed on the testing mechanism designed in this thesis.The
testing mechanism is constructed of constant speed servo motor,
torque sensor clamp,reduction gear,and off-central loading.
With constant speed rotation,the loading will be a sinusoid
torque over sensor. The frequency of sinusoid loading can be
adjusted through setting the rotation speed of servo motor.
Treat the sinusoid torgue as input and the phase difference as
output, then the transfer function of our sensor is easy to get.
|