·Home ·Table of Contents ·Methods and Instrumentation

Inspection of Defects on the Roll Surface Using AC Magnetic Flux Leakage Method Z.S.Lim Research Institute of Industrial Science and Technology, P.O.Box 135, Pohang, Korea D.R.Son Physics Department, Hannam Univ., 133, Ojung-dong, Daejon, Korea Contact

ABSTRACT

The crack-like defects occurring on the surface of steel rolls that are used in the steel mills should be inspected intensively in order to avoid the critical failure, such as spalling accident. Since the long direction can be either axial or circumferential, if we use the conventional inspection methods, ultrasonic or eddy current method, we should arrange the probes for each direction separately, or do the inspection twice with 90 degrees rotation.

In this study we employed AC-MFL (Alternating Current Magnetic Flux Leakage) method for the defect inspection on the surface of the HSS(High Speed Steel) rolls. The roll surface was magnetized using laminated yoke at the magnetizing frequency range from 1-kHz to 10kHz (f₀), and the resulting leakage flux was measured with a Hall sensor. The sinusoidal output of a function generator was fed into a power amplifier to generate an AC magnetic potential in the yoke made of laminated silicon steel. The output of the Hall sensor was rectified and filtered using a low pass filter with a cutoff frequency about ten times smaller than f₀. We made various artificial defects (0.1 ~ 0.5 mm in depth) on the roll surface using EDM(Electro-Discharge Machining), and recorded the magnetic field for each defect. The line speed of the roll surface was varied from 0 to 1 m/sec.

With a given magnetizing frequency, a certain magnetizing current showed a peak in signal to noise ratio.

For most of the artificial defects, 6 kHz was found to be effective magnetizing frequency. In general we found that the strength of MFL depended on the depth of the defect, which showed a possibility for the defect sizing.

An interesting result was that the circumferential and axial defects could be inspected with the same magnet and sensor configuration. The reason was possibly due to combination of the MFL and induced eddy current. A further study is needed to clarify this.

KEYWORDS : alternating current magnetic flux leakage method, defect, HSS roll, defect sizing

INTRODUCTION

Rolls used in the cold rolling and hot rolling mills are exposed to heavy usage and are transferred to roll grinding machines, where they are ground to make sure that their profiles are within the requirements. The surface is to be checked in order to prevent abnormal failure due to the surface defects. Since the hardness of HSS(High Speed Steel) reaches 80/85 shore, it is highly crack-sensitive. Therefore the surface should be thoroughly inspected to make sure that it is free of defects. The ultrasonic surface wave[1] and the eddy current method[2] have been used for the nondestructive testing for this purpose. Each method has its own advantages and disadvantages. In this paper we propose a new method using AC-MFL (Alternating Current Magnetic Flux Leakage).

EXPERIMENTS

Magnetizing Yoke and Power Amplifier
We have made an AC magnetizing yoke with non-oriented electrical steel sheets (PN-18) of 0.5 mm thickness , which are laminated and molded with epoxy and wire-cut as shown in Fig.1. The magnetizing current is fed into yoke using a 500W power amplifier, which is driven by a function generator. The driving frequency was varied ranging from 1 to 10 kHz. The power amplifier was laboratory designed using PA-03 (Fig.2).

Fig 1: Shape of magnetizing yoke

Fig 2: Schematic diagram of power amplifier

Magnetic Sensor and Its Controller
We have used a commercial Hall sensor(HW-300A-F). The sensor is driven in constant voltage mode. The output signal from the sensor is rectified and low-pass filtered with cutoff frequency of 10 Hz which is much lower than the driving frequency.

Flat-bed Scanner
Before doing experiments with rotating rolls, we have done preliminary experiments with static specimen. For this purpose we have made a low speed scanning system with flat specimen.(Fig.3). Using step motor controllers, we have scanned the magnetic sensor at a constant height from the flat HSS specimen with artificial defects and recorded the magnetic leakage flux with a digital voltmeter and an IBM-PC.

Fig 3: Block diagram of flat-bed scanner

We have prepared a HSS plate specimen with 25 artificial defects made by EDM. The physical dimension is 350mm×350mm×15mm(Fig.4). The defects are listed in Table 1. We have confirmed the sizes of the defects with a laser profiler.

No.	Len.(mm)	Wid.(mm)	Dep.(mm)	Vol.(mm3)	No.	Dia.(mm)	Dep.(mm)	Vol.(mm3)	Fig 4: Plate defect specimen
A1	9.79	0.272	0.435	1.146	D1	0.500	0.269	0.211
A2	9.79	0.272	0.380	1.012	D2	0.449	0.273	0.173
A3	9.79	0.099	0.250	0.233	D3	0.335	0.222	0.079
A4	9.79	0.132	0.185	0.256	D4	0.153	0.157	0.016
A5	9.79	0.189	0.090	0.167	D5	0.079	0.047	0.001
B1	4.92	0.237	0.444	0.556	E1	0.200	0.500	0.063
B2	4.92	0.263	0.352	0.452	E2	0.265	0.375	0.083
B3	4.92	0.081	0.213	0.100	E3	0.250	0.250	0.049
B4	4.92	0.165	0.180	0.125	E4	0.191	0.165	0.016
B5	4.92	0.191	0.082	0.077	E5	0.254	0.067	0.014
C1	2.00	0.248	0.356	0.177
C2	2.00	0.233	0.324	0.151
C3	2.00	0.211	0.253	0.108
C4	2.00	0.132	0.187	0.044
C5	2.00	0.169	0.094	0.033
Table 1: Dimensions of defects on HSS plate specimen

Roll Simulator
We have constructed a roll simulator to simulate a real situation. The outer diameter of HSS roll shell is 735 mm(Fig.5). The roll shell is rotated by 4 support rubber rolls with small diameter. The roll shell has 8 EDM defects(Table 2). The magnetic sensor is axially driven with 2 step motors with variable liftoff. In the following experiments the liftoff has been kept as small as possible.

Name	Pos.(deg)	Dir.	Len.(mm)	Wid.(mm)	Dep.(mm)	Fig 5: Shape HSS roll shell
D1	0	C	0.2	5.0	0.1
D2	45	C	0.2	5.0	0.2
D3	90	C	0.2	5.0	0.3
D4	135	C	0.2	5.0	0.4
D5	180	A	5.0	0.2	0.1
D6	225	A	5.0	0.2	0.2
D7	270	A	5.0	0.2	0.3
D8	315	A	5.0	0.2	0.4
Table 2: Dimensions of defects on HSS roll shell (A: axial, C: circumferential)

RESULTS AND DISCUSSION

1. Experimental results with flat-bed scanner
We have shown AC MFL experimental results for each defect(A~E) using flat-bed scanner for different driving frequencies(2,4,6, and 8kHz) in Fig.6-10.. The magnetic potential was set to 200 amp-turns, and the liftoff was kept as 0.3 mm.

Fig 6: AC MFL with flat-bed scanner for defect type A, (a) 2 kHz , (b) 4 kHz, (c) 6 kHz, (d) 8 kHz.

Fig 7: AC MFL with flat-bed scanner for defect type B, (a) 2 kHz , (b) 4 kHz, (c) 6 kHz, (d) 8 kHz.

Fig 8: AC MFL with flat-bed scanner for defect type C, (a) 2 kHz , (b) 4 kHz, (c) 6 kHz, (d) 8 kHz.

Fig 9: AC MFL with flat-bed scanner for defect type D, (a) 2 kHz , (b) 4 kHz, (c) 6 kHz, (d) 8 kHz.

Fig 10: AC MFL with flat-bed scanner for defect type E, (a) 2 kHz , (b) 4 kHz, (c) 6 kHz, (d) 8 kHz.

We know that defects A,B,C can be detected easily. The signal change at other than defects are thought to be arisen from inhomogeneity of the specimen. As the magnetizing frequency is increased the signal to noise ratio is increased. Since defect number 1 is close to the edge, the MFL signal is not so much enhanced as expected. The peak to peak of the AC MFL signal is proportional to the depth of defects rather than the length under fixed magnetizing frequency. But defects D,E are difficult to detect due to their small size..

2. Experimental results with roll simulator


A. Effects of magnetizing frequency
The AC MFL signals with fixed rotating speeed(4 RPM) and magnetizing current (0.05A) but with variable magnetizing frequency(1kHz, 2kHz, 4kHz, 6kHz, 8kHz, and 10kHz ) are shown in Fig. 11. We can say that D1 and D2 are difficult to find due to inherent anomalies of the roll. The data from defect free region are taken to give the rms base noise level.(V_N). The peak-to-peak AC MFL signal (V) is compared with V_N. The best S/N ratio is achieved with magnetizing frequency of 6-8 kHz for circumferential defects (D1~D4) and 4-6 kHz for axial defects (D5~D8). If we take 6 kHz as the magnetizing frequency, two different types of defects(circumferential and axial cracks) can be detected in the same scan with comparable S/N ratio, which is difficult for ultrasonics and eddy current testing. .The cracks that has more than 0.2 mm depth can be detected. We can say that the crack depth can be defined from the signal level of AC MFL signal, even though the crack shape is difficult to define. Other type of inspection method can be used to define the precise shape of the defect when it is found.

(a) f= 1kHz	(b) f= 2kHz
(c) f= 4kHz	(d) f= 6kHz
(e) f= 8kHz	(f) f= 10kHz
Fig 11: Effects of magnetizing frequency, RPM=4, Ic=0.05A

B. Effects of magnetizing current
The AC MFL signals with fixed rotating speed(4 RPM) and magnetizing frequency (2kHz) but with variable magnetizing current (0.02A, 0.05A, 0.08A, 0.12A, and 0.15A) are shown in Fig. 12. We can say that if we increase the magnetizing current level, the S/N ratio shows a maximum at a certain value. It is thought that the surface condition (hardness variation due to residual stress or etc.) can play a significant role with a high ac magnetic field.

(a) I=0.02A	(b) I=0.05A
(c) I=0.08A	(d) I=0.12A
(e) I=0.15A
Fig 12: Effects of magnetizing current, RPM=4, f=2kHz

C. Effects of roll speed
The AC MFL signals with fixed magnetizing current (0,08A) and magnetizing frequency (8kHz) but with variable rotating speed (RPM=4,10,20,) are shown in Fig. 13. When RPM is lower than 10, D2, D3, D4, D6, D7, D8 are easily found, while D1 and D4 are difficult to find. When RPM is 20, D2, and D6 are difficult to find. When RPM is 30, the S/N ratio for D3, D4, D7, D8 are greatly reduced. We think that the eddy current effect, the eccentricity of the roll, and the vibration of the system are thought to interfere with the AC MFL signals. It is difficult to define the sensitivity and the detectivity at the current stage. This can be a future research work.


(a) RPM=4	(b) RPM=10
(c) RPM=20	(d) RPM=30
Fig 13: Effects of roll speed, f=8 kHz, I=0.08A

CONCLUSION

We have developed a new method for the inspection of the surface of HSS rolls to be used in the steel industry, where ultrasonics and eddy current methods have been used. An AC power amplifier and a magnetizing yoke were used to generate AC MFL in the frequency rage of 1~10 kHz. The resulting magnetic field was measured with a Hall sensor. We have setup two experimental systems, a flat-bed scanner and a roll simulator.

For static experiments we used a flat-bed scanner. Using a flat HSS specimen with artificially made defects, cracklike defects with dimension of up to 2 mm in length and 0.1 mm in depth were detected and pinhole-like defects with dimension of up to 0.2 mm in diameter and 0.3 mm in depth were detected. Anomalous signals from other than defects were thought to come from the magnetic inhomogeneity from the specimen[3]. As the magnetizing frequency was increased, the magnetic impedance from the eddy current effect was also increased and the magnetic flux was confined within the shallow region under the surface, which contributed to the enhancement of the peak-to-peak of the AC MFL signal.

We have made a roll simulator using a large hollow HSS shell with artificially made crack-like defects on its surface. With low RPM, frequency of 6~8kHz showed maximum detectivity for circumferential defects and frequency of 4~6kHz showed maximum detectivity for axial defects. Both type of defects(circumferential and axial) were detected in the same scan. Defects with depth up to 0.2mm could be detected. The depth information was closely related to the peak-to-peak of AC MFL signal. As the magnetizing current was increased the detectivity showed a maximum at a certain magnetizing current

As the rotating speed of the roll was increased the detectivity became worse. With line speed lower than 0.4 m/sec, defects up to 0.2 mm in depth could be detected.

REFERENCES

1. P. Kumar and G.V.S. Murthy, "Ultrasonic evaluation of double poured rolls", NDT&E Int'l., 3, 1 (2000) 43-45
2. W.Hill and E.Kerr, "Tool steel work roll maintenance at Dofasco", 37th MWSP Conf. Proc., ISS, Vol. XXXIII, 1996, 283-289
3. H.Fujiwara, T.Suzuma, and T.Sakamoto, "Development of numerical analysis technique for alternating current magnetic leakage flux testing", JSNDI, 47, 2(1998), 127-133

© AIPnD , created by NDT.net

|Home|    |Top|