Aluminum Smelter Inspection Robot

Joe Foley, InstructorRU Hönnun-X

Vorið 2011

The Customer

Currently:• An inspector with a

thermal camera on a cart walks down the inspection route under the smelter pots

• He checks the video looking for hotspots

Aluminum Plant at Straumsvík

Inspection Corridor

• Big corridor so maintenance vehicle can fit

• Open to outside = cold• Lots of vehicles

Corridor, column 3

The Team

B.Sc. Students at Háskólann í Reykjavík

Instructors: Dr. Jón Guðnason and Dr. Joseph Foley

Structure of Work

• 12 week period– SCRUM – Teams

• Mechanical• Electrical• Sensors• Software

• 3 week intensive– Integration, testing

Design review at SCRUM

Robot Frame

Structural materials

• Stainless AISI 304.– A bit heavy but easily

available

Structural Analysis

ANSYS Simulation

Transmission

• Timing Belts• Ratio of 1:9

Wheels

• Independent two drive wheels with front caster

• Hubs custom machined(AISI 6060).

• Front wheel stock

Timing Belts

Transmission to wheel belt Motor to Transmission

Outer Shell

• Stainless Steel AISI 304.• Bent, welded, screwed

together.• Edges further sealed

with RTV rubber• Sized for 6 batteries.

Electrical Passthrough

• Partition for electrical isolation and environmental

• Easy to access motors and batteries without disturbing electronics

• Edges and passthroughs sealed

Partition on hinge

Top

• Stainless Steel AISI 304• Top angle for mounting

the RFID antenna• Nice ergonomics for

switches

Power Source

Motor power:• 2 12V Golf-cart

batteries• In series to make 24

Volts• Peak 50 Amps Electronics power:• 12V small battery• Everything else

12 Volt Motor batteries

12 Volt Electronics batteries

Motors

• 2 24 V, 120W DC motor from scooters.

• 2750 rpm through the 1:9 transmission

Motor and Battery Compartment

Motor Controllers

• Pololu 24v12. – One per motor

• RoboClaw– Originally for both

motors, but repurposed for analyzing the tachometers

– Sent to FitPC controller via serial

Polulu 24v12

RoboClaw

Spennudreifing

• 24V power– Relays for emergency

stop and general safety

• 12V power– Three relays– Eight sensors– FitPC computer– Arduino microcontroller– Audio amplifier– LED lighting

24V Safety cutoff

12V distribution

Arduino Mega

• Sensor processing is done through the Arduino Mega

• Motor commands are sent directly to the controller

• The Arduino Mega has a ATMega2560 processor which has – 56 digital IO pins– 13 analog inputs– USB– 4 serial ports

Arduino Mega

Shield for Mega

Audio Amplifier

• Audio feedback and warnings allow the robot to indicate status and get attention

10W Amplifier

Complete Electronics

Electrical Block Diagram

Fit PC

• 1,6 GHz processor• 1 GB memory• Size: 100x115x26mm• 60GB Solid State disk

Operating System

• Ubuntu server.• Non-GUI Linux.• ROS (Robotic Operating

System).

Documentation

Thermal Camera

FLIR E4:• Currently used• Max 900°C.• Stores in JPEG

Image to Temperature

• Python og OpenCV in Linux• Individual pixel processing

– 255 (= white) is max temperature

– 0 (= black) is minimum temparture

• Adjust our threshold according to the range– Example: Range set to 100°C

upp í 350°C– Value of 100:

100*(250°/255)+100° = 198 °C

Greyscale raw image

Data processing

If above our threshold (300C) then store image

Temperature statistics in Sqlite database:

• Average• Standard deviation• Maximum• Current position (ROS)

False color highlighting threshold as green

RFID• Location found using RFIDReader: • TagSense Micro UHF (800-

100MHz)• 3,8cm * 4,6cm * 0,7cm• USB interface• Voltage: 5V• Current: 250 mAAntenna:• Frequencies: 865 – 956 MHz• Gain: 6 dBi• SMA connector (inverted) • Reads tags up to 80 cm

Micro UHF reader

Antenna

RFID

Merkin:• ShortDipole Impinj

Monza 4 from UPM. • Frequency: 860-960

MHz • Size: 93mm*11mm• 96 Bits of information– Aluminum pot, column,

etc.RFID tag

RFIDLocation

Map of corridor

• One tag per pot•Each number is associated with information in the SQLite database•Database instructs possible future actions for robot

User Interface

• User can track current location of the car remotely

• Queries the SQLite database to track and find out status

Position control(e. Localization)

Encoder: • 27.600 pulses per rotation.• Wheel radius R = 13,54cm• Circumference = =

85,07cm• Resolution=

R2

Encoder

Ultra Sonic sensor

IR-Sensor

Turning (IR - feedback)

l = distance between sensorsd = delta of sensor readingsd = sk1 – sk2

l

d1tan

PID - loop

Complete!