SUMMARY - Solex Robotics tested Maverick in a tank containing 100,000 bbls of jet fuel which was the servicing Atlanta airport. 
The tank remained in operation during the inspection.  In two hours, more than 31,000 discrete ultrasonic data points were collected. 


September 1st - 4th Maverick was demonstrated successfully in a large fixed-roof jet fuel storage tank at the Atlanta Junction Colonial Pipeline facility.  Testing targeted refinements to the control software, UT transducer frequencies, sonar blockage, and other operational issues.

The tank inspection provided several significant operational challenges.  The tank serves as a primary source of jet fuel for Hartsfield Airport.  The drain piping was connected to pontoon floats and would pivot up and down with the fuel level.  During the inspection, jet fuel was drawn and filled with Maverick having no impact on their operations.  The tank had no internal drawings.  There were five fixed columns with the possibility of six which made the calculation of the internal geometry difficult.  The sump was open, without any guard. From a quick visual inspection conducted in June, 1998, the tank was known to have heavy top-side corrosion. Hurricane Earl moved through Atlanta during the demonstration which caused the Solex team to put into effect emergency shut down of operations.

With respect to the demonstration plan, the first item for testing was the control system.  A new driving system, including software and the integration of a sophisticated joystick, was installed into the system and tested for operability.  After minor adjustments to the software, the system performed flawlessly and now provides greater maneuverability to the operator.  The value of the upgrade was evident later on in testing when the new capabilities enabled the operators to extract the robot from several tight situations.

The second major test was the UT system against the difficulties of getting thickness readings from corroded tank floor plates. Initially, no thickness readings could be obtained with the 5 MHz transducers.  The robot was pulled from the tank, and every portion of its UT system was checked and recalibrated.  All portions of the system were operating properly. The plates directly below the entry hatch were heavily pitted on the top surface and the sound was scattering too much to obtain thickness measurements.  The UT analysis software was reconfigured to measure the surface pitting and to generate profiles of the top surface of the plates.  This revealed the extent of the top-side corrosion and yielded a survey analogous to a rough visual inspection.  Other areas of the tank floor were less corroded and were fully mapped using the 5 MHz transducers.  To improve the system's performance on heavily corroded regions, a set of 2.25 MHz transducers were installed and tested.  The lower frequency transducers were able to penetrate the rough surfaces, allowing the system to map the roughest regions of the tank floor successfully.  Future inspection operations will include full sets of transducer arrays to best match the surface condition encountered. 

The UT data software can be set up to directly measure two out of three characteristics: top-side surface profile; bottom-side profile; and, metal thickness.  The third characteristic can then be developed computationally by direct export to Excel.  Scanning and data acquisition is in real time and continuous.  Within a discrete time or distance traveled interval, Maverick records the worst data point observed.  Currently, Maverick makes a record of the data acquired during one-inch intervals.  The recording system is infinitely adjustable.  In two hours of actual scanning, 31,500 UT data points were collected.  This amounted to 390 square feet (3.5%) of the tank floor.  The inspection company that had provided the previous visual inspection considered the data acquired by Maverick to be reliable.

The sonar tracking system was also challenged by the tank's internal geometry.  The roof was supported by several pillars which occasionally blocked sound from reaching the listening transducers.  As a result, there were blind spots within the tank where the positioning system was unable to function.  The operators were able to work effectively around this deficiency in most situations.  On one occasion, the robot was driven too close to the sump due to positioning error and fell in, but was pulled out with minimal effort.  The sound blockage problem had been previously anticipated.  To solve the problem, additional listening transducers will be used eliminate the blind spots.

The tank roof pillars also posed a navigational problem inside the tank.  Both the location and the number of pillars within the tank were in question, since drawings of the tank were unavailable.  Throughout the week, visual surveys of the inside of the tank were conducted to identify obstacles like the roof pillars and to demonstrate the visual capabilities of the robot.  The pillars are constructed from large back-to-back welded steel channels, with support angles along the tank floor at the base, as seen clearly in the robot's cameras.  The cable tended to catch on the edges of the channels when the robot was driven around a pillar.  In the future, surveys will be conducted in a way to avoid wrapping the cabling around roof pillars.

The demonstration proved to be successful and concluded a series of important engineering developments to make Maverick commercially ready.  We want to say thank you to the Department of Energy, Lockheed Martin, General Electric, Texas Natural Resource Conservation Commission, American Petroleum Institute, Independent Liquid Tank Association, Mobil, Colonial Pipeline, Amoco, Shell Oil, Exxon, Al Mansoori Specialized Engineering, United Arab Emirates Offset Group, Regional Development Alliance, Idaho Department of Commerce, City of Idaho Falls, and the many others who have helped by being a part of the beginning of this success.

References available upon request.