Instrumentation & Measurement Magazine 26-5 - 42

INPUT PARAMETERS
6 Axis ForceTorque
Sensors
Banner
CAM
X um
Y um
Linear
Robot
Encoder
0 deg
Screw
Gap
OUTPUT
PARAMETERS
Robotic Filter
Tune System
Z um
Pen.
Depth
Move
Turn
Carry
Fig. 4. Robotic control architecture, from [19] (©2022 IEEE, used with permission).
each cavity filter. The tuning process has been completed with
two stages which are coarse and fine tuning. Coarse tuning is
based on Dishal tuning, and fine tuning is based on time and
frequency domain tuning. To minimize the tuning time, avoid
the overfitting problem, dimension redundancy should be carried
out. First, coarse tuning is implemented which is followed
by a fine-tuning stage. The control algorithm architecture
given in Fig. 4 can be divided into two sub-sections: preparations
for the tuning set-up and the tuning process. During the
preparation, all constant and variable parameters of screws are
identified. In the tuning section, steps for locking the screw,
turning, and unlocking from the screws are carried out. All
these functions are managed with three commands: " turn, "
" carry, " and " move " by the control algorithm. Also, all robotics
and sensors (encoders, Portable Network Analyzer (PNA),
and force torque) are linked to the control algorithm. All data
transfer between PC and encoders or controllers, sensors, and
image process system is performed with ethernet and USB
buses. Also, communication with PNA is provided through
GPIB. All sub-functions, methods, and classes in the control algorithm
have been constructed using DLL libraries.
Tuning Setup Preparations
Identification of Constant Parameters: Screw exact position
parameters, and X and Y coordinates, as well as Ø degree, can
be sensed with image processing by image recognition system
to. After teaching the region of interest (ROI) screw pattern, positions
are measured regarding the reference image at 17 µm/
inch resolution.
Fig. 5. Identification of the variable parameters.
42
Identification of Variable Parameters: Total z distance from
screwdriver to screw position, current screw depth positions,
and screw and driver gap angles must be determined before
tuning process. Z distance from the top and the bottom
of the filter resonator screw head surfaces are not constant, so
they vary according to the penetration depth of the screws,
as shown in Fig. 5. In the developed system, the cameras do
not have the ability to sense depth, so it is not possible to
have precise measurements. Lidar sensors cannot sense reflected
signals from the target since the filter surface material
is either silver or gold. These multiple reflected and diffracted
pulses interfere to measure highly accurate results. Therefore,
a search algorithm using FT sensors and linear encoders for
both top and bottom faces was employed. Final screw z position
values were derived during the locking stage of control
algorithm, and they were saved to a database.
IEEE Instrumentation & Measurement Magazine
August 2023
Instrumentation & Measurement Magazine 26-5

Table of Contents for the Digital Edition of Instrumentation & Measurement Magazine 26-5
