NDT tool makes light work of tank inspection

TWI Bulletin, May - June 2005

Inspecting large steel storage tanks can be a risky business - but not if the inspectors adopt a remote mechanical friend...

Bruce Blakeley is a Senior Project Leader in the NDT Technology Group at TWI. He joined the company in 2002 after spending six years at Corus in Port Talbot, where he completed a PhD in Audio Acoustic Condition Monitoring. He then went on to become the Condition Monitoring Engineer for the Cold Mill. Since joining TWI he has worked on a variety of projects from automated ultrasonics, robotics and digital radiography.

Dubbed Robot Inspector it presents a revolutionary way of inspecting suspect structures without the drawbacks of some existing inspection techniques. As Bruce Blakeley reveals the system adopts an arm's length approach to inspection of sheet steel structures...using a wheeled robot.

Above ground storage tanks are used to store bulk fluids, such as oil, foodstuffs, pesticides and fertilisers, as shown in Fig.1.

The tanks are built from steel plates formed into a simple cylinder with a sheet steel floor. If one of these tanks were to leak, the damage to the environment could be catastrophic. To prevent such a disaster the tanks are regularly inspected to ensure that the steel plates have not corroded, which could cause a failure of the tank.

The welds used to join the plates are inspected during fabrication. The ceiling and walls of the tank are relatively easy to inspect during service, using magnetic crawling devices. These robots have magnetic wheels to attach on to the outside of the tank, and a simple longitudinal ultrasonic probe to measure the thickness of the steel plates. These crawlers are very useful for taking spot measurements on the walls and ceilings of the tanks, but lack the sophistication to map corrosion in the tank floor autonomously.

There are currently several solutions to this problem:

• Manual inspection
• Magnetic Flux Leakage (MFL)
• Submersible robots
• Acoustic emission

Manual inspection

To inspect the tank manually, it must be drained of fluid, cleaned, defumed and degreased to allow access to NDT technicians, who then use manual ultrasonic techniques to measure the thickness of the floor-plates to ensure they have not corroded during service. This emptying and cleaning of the tank is very expensive and results in loss of production. Manual inspection is time consuming and dangerous, and may give false readings as the monotonous nature of the work can lead to incorrect readings. A single spot measurement in the centre of each plate is sometimes the only measurement taken. If the corrosion is nearer the edge of the plate the operator is likely to miss it.

Magnetic flux leakage

Magnetic Flux Leakage (MFL) is the most commonly used technique. Again the tank must be drained, cleaned, defumed and degreased, to allow NDT technicians to enter the tank safely. A trolley containing heavy duty magnets is pushed across the plates, inducing a saturated magnetic field within the steel floor. Sensors then monitor the change in flux as the trolley is pushed across areas of corroded plate. Again, this is a manually operated technique, and is prone to operator error due to fatigue.

Submersible robots

Sophisticated submersible robots can be lowered into the tank from the roof, without the need to empty the tank. These robots are connected to a sophisticated positioning system complete with a CAD model of the tank, so that it is able to navigate around internal pipework without hitting any obstacles.

These robots are extremely sophisticated, but the NDT sensors they carry can be rather basic; they often consist of just a single ultrasonic probe, no more sophisticated than the probes used by the manual operators. These robots also depend on the CAD model of the tank, which can often be out of date or incomplete.

Acoustic emission

Acoustic emission sensors are placed around the outside edge of the tank floor. These sensors are claimed to be able to detect possible signs of corrosion. However, they can give false failures.

Robot inspector

All of the above techniques have their place within the market, but all have disadvantages; either they do not provide 100% inspection or they are prone to operator errors or false failures. The best they can do is to indicate that the tank's condition is questionable. Once a tank has been failed by one of these methods the tank operator must determine the precise nature of the problem, how severe it is and where the corrosion has occurred. What is required is a robot capable of providing a 100% coverage, giving quantifiable measurements of the floor thickness. Ideally the same robot would be able to inspect the welds of the plates during fabrication.

These features would also make the robot ideally suited for the inspection of ships during fabrication and the inspection of large rolled sheet steel for delamination at the steel mill. Fig.2 shows the Robot Inspector system.

The Robot Inspector consortium

The Robot Inspector Consortium consists of European companies involved in the supply chain servicing the tank-inspection market, including equipment and NDT manufacturers, service companies, research establishments, universities and robotics experts. They've been partly funded by the European Commission under the CRAFT co-operative research program. For the last two years work has been done on an inspection robot with the objective of providing tank operators with detailed knowledge of the condition of their tanks, to help them reduce leakage and spills that could severely damage the environment.

The three stages of operation

The Robot Inspector system consists of a rugged four-wheel drive robot capable of working in harsh environments. A laser range-finder is used for mapping its environment, a wheel-probe containing an ultrasonic sensor is used to measure the thickness of the steel floor, and a side-scanner, complete with ACFM sensors and phased-array sensors is used to inspect welds.

Stage One - Mapping the environment

Once the condition of the tank has been proven to be suspect using one of the alternative techniques described above, the tank must be cleaned and degreased as normal. The robot is placed in the tank through a manhole in the side of the tank. The autonomous robot then wanders around the tank in a random manner, using its laser navigation system to prevent it colliding with unknown objects in the tank. The laser range-finder, shown at the front of the robot in Fig.2, scans in a 180° arc, creating a virtual map of the environment it is to inspect - this map is then used for navigation in the following stages. Fig.3 is a map of the laboratory at TWI, created by the laser navigation system.

Stage Two - Corrosion mapping

The ultrasonic wheel-probe, shown in Fig.4 is then attached to the front of the robot. This is used to provide 100% inspection of the potentially damaged plates, highlighted by the previous techniques used.

The wheel probe is a water filled rubber wheel, whose ultrasonic properties are similar to the Perspex shoes normally used in ultrasonics. This wheel contains a longitudinal ultrasonic probe for measuring the thickness of the steel plate floor. The wheel probe is moved from one side to the other, while the robot inches forward to produce a raster scan of the plate.

A threshold is set to determine the minimum allowable plate thickness. Any areas of the floor that fail to meet this standard are plotted on the colour-coded map produced in stage one. Fig.5 shows the wheel-probe.

Stage Three - weld inspection

The welds may be inspected at any time, most usefully during fabrication of that tank or ship. A side-scanner is fitted to the side of the robot. It uses an ACFM array to locate surface breaking cracks in the weld, and a phased array probe to locate volumetric faults, such as internal cracks and lack of side-wall fusion, that could lead to the premature failure of the tank. At the end of the inspection automated defect recognition software locates the faults and adds this information to the map created in stage one.

Fig.6 shows the robot's side-scanner, containing the ACFM and phased-array sensors

The ACFM array shown in Fig.7 consists of a curved array of eight ACFM sensors.

The curvature of the array makes it possible to inspect butt, fillet and overlapped welds. Semi-automated defect recognition assists the operator in selecting suspect areas of the weld. The position of the robot is known from the laser navigation system, which allows the location of the faults to be added to the colour coded map.

Fig.8 shows the phased-array sensors, used to detect the presence of volumetric weld faults.

Two phased array sensors are used. One for the overlap and butt welds, and one for the fillet weld. The phased-array sensor used for the fillet weld is positioned against the outer wall of the tank, as shown in Fig.8. It is able to detect faults in the inner and outer annular weld of the tank, in just one pass. Automated defect recognition software is capable of analysing the phased-array data, and returns the positions and lengths of the accumulated defects along the length of the weld. This information is added to the map.

Sensors within the side-scanner provide feedback to the navigation system so that the robot can move parallel to the outer wall or weld. When the robot moves too close, the feedback sensors relay this information to the on-board controller, which uses fuzzy-logic to correct its path.

Results

The end result is an accurate map of the tank with areas of corrosion and weld defects clearly indicated. To date Robot Inspector has been validated in the laboratory, and it is hoped that validation within a tank will be realised in the future. To validate the system in the lab, several 8mm steel plates were welded together to form butt, fillet and overlap welds, similar to those found in tanks. Genuine corrosion samples, cut out of the floors of older tanks were also examined. The plates were placed on the floor of the lab, and the robot instructed to inspect them. The result of this inspection is shown in Fig.9.

The colour-coded faults indicate where the defect detection software was able to determine faults such as cracking and porosity in the welds and corrosion in the floor plates.