|NDT.net September 2003, Vol. 9 No.09|
The ACFM probe, operated by the manipulator's arm, is moved over the structures to be examined. The procedure measures the x and z direction values of the magnetic field intensity (Bx and Bz) above a crack and the vibration amplitude (Ax, and Az) of the ACFM probe versus position on the examined structure. Two models, online detection and amendment models have been established. Online detection model shows that the magnetic field components are related to the rates of change of the surface potential differences. With no defect present and a uniform current flowing in the y-direction, the magnetic field is uniform in the x-direction, perpendicular to the current flow, while the other two components, By and Bz, are zero. The presence of a defect diverts current away from the deepest part and concentrates it near the ends of a crack (or on either side of a pit). The effect of this is to produce strong peaks and troughs in Bx and Bz (above the ends of a crack or either side of a pit), while Bx shows a broad dip along the whole defect. When there is a crack on the structure, the curve of Bx and Bz is regularized. A templet is defined. If the curves of Bx and Bz are the same as the templet, it's sure that there is a crack on the structure. The distance away from start position, length and depth of crack can be calculated through the probe speed, the communication rate, a peak and trough, and a broad dip. The amendment model compensates the error of the magnetic field intensity caused by the probe instability perpendicular to the detected structure. The ACFM probe was installed on the URV. There will be instability. The ACFM requires lift-off between the probe and the detected structure to be constant. A function of magnetic field intensity versus distance between the probe and the structure has been established. An amendment value can be obtained by subtracting a first modular value from a second modular value. A templet of the Bx and Bz, and a function of the magnetic filed intensity versus lift-off are used in an underwater crack detection system deployed by the URV. The methods increase greatly the measurement accuracy. These have been proven in the laboratory.
Keywords: Non-Destructive Testing, Alternating Current Field Measurement, automated crack detection system, underwater structure inspection.
This paper deals with some research work performed at Robotics Research Centre, School of Mechanical & Production Engineering, NTU, based on the use of an ACFM Crack Microgauge Model U10, some ASTM-A36 structure steel specimens and a manipulator, Kawasaki JS-5. The goal is to set up an ACFM automated Crack detection system deployed by an Underwater Roving Vehicle, SUPER SAFIR, Hydro-technologie, which is able to online crack detect and improve the accuracy and reliability of the system.
1.2 ACFM technique
Alternating Current Field Measurement (ACFM) is an electromagnetic technique used for the detection and sizing of surface breaking cracks in metallic components. Figure 1 shows the basic principles of the technique. When an alternating current flows in a conductor it flows in a "skin" following the surface. With no defect present and a uniform current flowing in the Y- direction, the magnetic field is uniform in the Y- direction perpendicular to the current flow, while the other components are zero. The presence of a defect diverts current away from the deepest parts and concentrates it near the ends of a crack. The effect of this is to produce strong peaks and troughs in Bz above the end of the crack, while Bx shows a broad dip along the whole defect with amplitude related to the depth. These theoretical models have been shown to correlate well with the physical measurement of magnetic fields [1,2].
Fig 1: Current flow around a defect
Fig 2: Bx and Bz components
Fig 3: An ACFM detection system
Now weld inspections are using ACFM arrays deployed by URV [4, 5]. The probe is simply placed in a series of overlapping position around the weld. This method is generally insensitive to permeability changes and lift off. But the ACFM arrays can't access all geometric configurations encountered subsea. It's difficult to inspect weld toes or sophisticated geometric structures.
1.3 System configuration
This paper describes a way to find the crack and calculate the length and depth of it in an underwater environment, which way can be performed easily by software. Our system (Figure 3) deploys a single probe, and uses an online detection model and amendment model to detect cracks. The system considered has the following characteristics:
2.1 Characteristics of Bx, Bz and butterfly
From the Bz plot (Figure 2.), it was observed that the curves reflected from one half of the z-axis to the other half. For example, in the forward scan, the Bz values decreased steadily from background level to the first trough and subsequently increased to the maximum point which indicated the end of the crack. When the probe scanned in the reverse direction, the curve was seen to increase in value till it reached the maximum point whose absolute Bz would be close to that of the maximum, if not equal. The plot would then go down steadily toward the peak and whose value again, would be close to the peak point. Henceforth the repeatability is obvious. The difference is that the value of Bz starts at a trough and ends at a peak or the value of Bz starts at a peak and ends at a trough. The curve of Bz is insensitive to the speed and lift-off of the probe, and the length of the crack. So Bz is used to find if there are any cracks in the surface of a weld, and if there are cracks, calculate their length [6, 7].
From the Bx plot (Figure 2), the figure showed that crack depth varied with the dip of Bx. The differential Bx with background is related to the depth of the crack. It is sensitive to the lift-off of the probe and insensitive to the length of the crack and speed of the probe.
Fig 4: Plots of Bz and Bx
Fig 5: Butterfly plot
Fig 6: A variable lift-off
Fig 7: Calculation of the area of butterfly
Fig 8: A structure of online detection model
Fig 9: A structure of amendment model
Any scans producing significant loops in the butterfly plot would be rescanned for confirmation. This benefit is used to good effect for underwater inspection deployed by URV.
2.2 Lift-off analysis
Figure 6 presents that lift-off of the probe varies with time at A and B points. The unsteadiness of the probe affects the value of Bx, and Bz is affected slightly. We used ACFM Crack Microgauge Model U10 to measure the ASTM-A36 structure steel specimen. If lift-off is greater then 10 mm, the value of Bx and Bz is useless, open air values. But it doesn't affect the calculation of crack length if lift-off is smaller than 6 mm.
Distance between probe and metal surface is an important factor in determining the strength of the magnetic field measured. At zero lift-off, the accuracy of the predicted depths of the crack in specimen was high. The average percentage error was approximately 5%. The level of accuracy decreased as the probe lift-off height increased. From the experiment for 2mm lift-off, the percentage error is about 30%. In sharp contrast, the relevant percentage errors for 4mm lift-off were increased to 75% . From above information, we can see that the lift-off of probe affects seriously the accuracy of the determination of the depth of the crack.
2.3 Methodology of model
Our approach is based on the analysis of Bx and Bz above. From the curve of Bz, we assume that 1 represents increase of Bz, 0 represents decrease of Bz. So we can get a cluster of 1 and 0.
… 00001111 … 11110000 …
The point from 0 to 1 is the trough of Bz, the point from 1 to 0 is the peak of Bz. Moreover if the probe scans in the reverse direction, we see
… 11110000 … 00001111 …
This is called the templet I. When opening a window to check a cluster of consecutive 0 or 1 of Bz, we can distinguish whether there is a crack on the surface.
The templet II evaluates the area of the butterfly of Bx and Bz. We cross Bzgd of Bz axis at Bxgd of Bx, the value of the Bx and Bz background. We can break up the butterfly with each point on butterfly, shown in Figure 7. A and B points are i and i+1 two points on the butterfly. a is an angle between A point and B point. The area of the shade is about,
The online detection model is based on the templet I and templet II. The model is shown below in Figure 8. If templet I and templet II all detect a crack, we can determine that a crack is detected. Then according to the number of a cluster of 1 or 0, the speed of probe, and the communication rate, we can calculate the length of the crack.
The magnetic fields are strongest at the source i.e. at the defect. Extra lift-off and any offset from the centerline would contribute to inaccuracies in crack depth estimation. The separation would reduce the sensitivity of the sensing probe. Therefore maintaining positive contact between the probe and the examined structure is desirable. If lift-off is expected, a correction factor may be introduced so that the crack sizing would be accurate.
The amendment model will be shown as above figure 9. The model uses an ACFM lift-off sensor to determine the distance, Az between the face of the probe and the structure. Bx'(Az) is a function of the magnetic field against Az. The function is related to Az, the materials and probe type. It is independent of n. The function of Bx'(Az) is complicated. We use discrete points to set up the function. More detailed function of Bx'(Az) is developing.
The ACFM technique was initially developed to allow crack sizing underwater where other techniques were hindered by the need for good electrical contact. However, the other advantages arising from non-contact and a uniform input current meant that the technique was quickly applied to topside inspections as well, particularly on painted or coated welded structures. The technique has also been shown to be capable of detecting fatigue damage in welded materials and to be capable of identifying areas of localized micro structural change. These may be due to micro cracking, local repairs, or additional material phases all of which could be of structural significance and which certainly represents potential problem areas.