Investigation into eddy current pulsed thermography for rolling contact fatigue detection and characterization

Author's Accepted Manuscript

Investigation into eddy current pulsed thermography for rolling contact fatigue detection and characterization Jianping Peng, Gui Yun Tian, Li Wang, Yu Zhang, Kongjing Li, Xiaorong Gao

www.elsevier.com/locate/ndteint

PII: DOI: Reference:

S0963-8695(15)00058-4 http://dx.doi.org/10.1016/j.ndteint.2015.05.006 JNDT1690

To appear in:

NDT&E International

Received date: 5 July 2014 Revised date: 13 May 2015 Accepted date: 16 May 2015 Cite this article as: Jianping Peng, Gui Yun Tian, Li Wang, Yu Zhang, Kongjing Li, Xiaorong Gao, Investigation into eddy current pulsed thermography for rolling contact fatigue detection and characterization, NDT&E International, http://dx.doi.org/10.1016/j.ndteint.2015.05.006 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Investigation into eddy current pulsed thermography for rolling contact fatigue detection and characterization Jianping Penga,ba,*, Gui Yun Tianb, Li Wanga , Yu Zhanga , Kongjing Lib, Xiaorong Gaoa a

School of Physical Science and Technology, Southwest Jiaotong University, Chengdu, People’s Republic of China

b

School of Electrical and Electronic Engineering, Newcastle University, Newcastle upon Tyne, NE1 7RU, United Kingdom

Abstract: This paper reports on the use of eddy current pulsed thermography (ECPT) for detection and characterization of rolling contact fatigue (RCF). Detection mechanisms with eddy currents and heat propagation effects were discussed with RCF modeled as a simple angled defect. Two different angled defects were studied through numerical simulations and experimentally by using uniform magnetic field (UMF) excited by Helmholtz coils. Finally, a rail sample with RCF defects was inspected using UMF excitation. It is shown that ECPT with UMF excitation provides an efficient and robust method to detect angled defects, compared with nonuniform magnetic field (NUMF) excitation.

Keywords: eddy current pulsed thermography, rolling contact fatigue, Helmholtz coils, nondestructive evaluation

a

E-mail: [email protected]

1

1. Introduction Rolling contact fatigue (RCF) has become a pervasive and insidious problem for all types of railway systems, becoming a safety challenge for high-speed railways due to short examination times. The derailment at Hatfield, United Kingdom, in October, 2000, is an example of the possible consequences of RCF, which results in large economic losses and many injuries [1, 2]. RCF failure [3] involves rail damage that features multiple cracks arising from overstressing the rail material. These include shelling, squats, and gauge corner cracks found at or close to the surface, and deep-seated shells found within the subsurface. A typical RCF, gauge corner cracking (GCC) produces a series of surface cracks in the railhead [4] [Fig. 1(a)]. These are clusters of microscopic oblique cracks in the surface along the rail tread formed at an angle of 15°–40° with a typical spacing of 2–7 mm. These angled cracks also propagate a few millimeters (but less than 4 mm) deep into the railhead through heavily deformed surface layers at an angle of about 10°–30° to the rail running surface. A micrograph of an RCF defect [5] propagating into the railhead at a shallow angle [6] of 28° is presented in Fig. 1(b). Research on the application of nondestructive testing (NDT) methods for detection of defects in rails has been performed over the last 100 years. To date, more than ten NDT techniques have been used in the detection of rail defects [7]. For example, the ultrasonic technique has been proven to be a successful method to detect inner defects. In general, ultrasonic technique performs relatively well in detecting deep surface-breaking and internal defects, particularly in the railhead and web. Unfortunately, RCF defects that are smaller than 4 mm in depth are usually not detectable by high-speed systems. However, the existing ultrasonic systems are inadequate for detection of RCF defects because of the ultrasonic dead zone [8]. Sometimes, the surface defects due to RCF cause shadowing or blocking of the ultrasonic wave preventing detection of deeper, more serious defects [1]. An NDT system combining the ultrasonic technique with better probe arrangements and the eddy current technique is a possible solution for detecting rail defects [1, 9]. Nevertheless, the studies show that, in detecting rail defects, pulse eddy current (PEC) probes perform better than magnetic flux leakage sensors, which are affected by the lift-off variation [10]. In quantifying RCF, alternating current field measurement (ACFM) provides good results for surface defects even when obtained under lift-off and time constraints [11], but not for subsurface rail defects or multiple cracks [12]. An electromagnetic acoustic transducer can generate surface waves to detect RCF in rails at small lift-off and has the advantage of non-contacting and rapid operation, but is limited when operating at a large lift-off and for identifying multiple cracks [13]. However, the 2

characterization of RCF, such as size, shape, and orientation in the case of multiple defects, including surface and subsurface angled cracks [3], is important for the rail maintenance and quality control [1]. Eddy current pulsed thermography (ECPT) combining pulsed eddy current and thermographic nondestructive testing techniques provides an efficient method for detection and characterization of defects over a relatively wide area in comparison with other active thermography methods, such as flash heating, laser or vibration thermography [14]. When the eddy currents encounter a discontinuity, they are forced to divert, leading to eddy current density variation, in which higher levels of Joule heating are achieved in the regions of increased current density. Thus, the defect can be identified from a characteristic heat distribution in the thermal image/video [14]. In previous work, several modeling and experimental studies of ECPT have been performed at Newcastle University. Two fundamental defect models, namely slot and notch, were constructed to determine the eddy current and heat distributions around any given defect, based on the theoretical eddy current distributions and Joule heating [15]. The physical relationship between the eddy current distribution and thermal processes in the heating and cooling stages was studied based on thermal transient features [16]. Successful detection of surface cracks by eddy current inductive thermography, including artificial and natural cracks with the depth of 0.1–4 mm, was reported [17, 18]. Furthermore, the ECPT technique was used in inspecting metal corrosion under paint coatings [19], fatigue evaluation in gears [20], and damage evaluation in carbon-fiberreinforced plastic [21]. Identification of an RCF defect in a railhead using ECPT was studied using an angled slot model in the COMSOL Multiphysics software platform [22]. Spatial thermal features, e.g., thermal gradients and the maximum temperature amplitude distribution were extracted from the numerical results. The results showed that these spatial features provide means to characterize the angled defect. Actual RCF defects in a rail sample have been imaged experimentally using ECPT in a previous experimental work [23], where two RCF defects with different sizes were studied through the transient response and were characterized to some extent. Previous work attempted to use induction coils to generate a nonuniform magnetic field (NUMF) with the field strength falling off inversely proportional to the distance, i.e., ~1/r. However, the spatial thermal features were only studied beneath the induction coil. Additionally, only a small area near the induction coil was used to avoid the influence of NUMFs in ECPT [14], therefore, only a small part of the RCF defect was imaged [23], and the coil was required to be scanned over the area to generate a complete 3

image of the defect. To resolve or alleviate these difficulties, generation of a uniform magnetic field (UMF) using Helmholtz coils [24] is a potential solution to these issues in ECPT. Helmholtz coils provide a flexible inductive heating solution independent of the geometry of the sample in ECPT applications, such as detection of failure in bonded wires in insulated-gate bipolar transistors [25] or identification of natural cracks in a casting sample [26]. In the present work, Helmholtz coils were used to produce a UMF excitation in the sample to overcome the limitation of linear induction coils. The spatial features, e.g., thermal gradients of an angled defect and the maximum temperature amplitude, were then studied for the cases of two differently sized angled slots in a COMSOL simulation based on the UMF excitation. The results for UMF and NUMF excitations were compared to analyze the robustness of those features. Additionally, a cutoff rail sample containing RCF defects was experimentally tested using ECPT with Helmholtz coils. The rest of this paper is organized as follows: Section 2 discusses the principle of ECPT and detection of RCF; Section 3 introduces the simulation procedure in COMSOL; Section 4 introduces the experimental study of RCF; Section 5 presents the conclusion and proposes future work.

2. Methodology 2.1. Induction heating theory in ECPT In an ECPT system [14], eddy currents are induced in the sample by the excitation field, which penetrates it to a certain depth because of the skin effect. This skin depth is defined as:

δ=

1

(0)

πµσ f

where f, σ, and µ denote the excitation frequency (Hz), electrical conductivity (S/m), and magnetic permeability (H/m), respectively. The conductive material is heated through Joule heating by eddy currents. The generated resistive heat power Q is proportional to the square of the magnitude of the electric current density Js. Current density, in turn, is proportional to the electric field vector E. Hence, we obtain:

Q=

1

σ

2

Js =

1

σ

σE

2

(0)

where the electric conductivity σ is dependent on the temperature T, and is given by:

4

σ=

σ0 1 + α (T − T0 )

(0)

where σ0 is the conductivity at reference temperature T0 and α is the temperature coefficient of resistivity, which describes how the resistivity varies with temperature. The conductivity σ is thus inversely proportional to the temperature T. The heat conduction equation describing a specimen heated inductively by a source Q is:

ρC p ∂T 1 ∂ 2T ∂ 2T ∂ 2T = q( x, y, z, t ) + ( 2 + 2 + 2 ) λ ∂t λ ∂x ∂y ∂z

(0)

where ρ, Cp, and λ are density, heat capacity, and thermal conductivity, respectively. Here, T(x,y,z,t) is the temperature distribution, which is caused by the Joule heating sources and heat diffusion in the material; q(x,y,z,t) is the internal heat generation per unit volume as a result of the eddy current. In ECPT measurements, two kinds of thermal processes, Joule heating from eddy currents and heat diffusion inside the material are apparent. In the heating stage, the existence of surface defects distorts the eddy current density distribution (which depends on q(x,y,z,t)) and results in a variation in temperature. However, in the cooling stage, the temperature changes occur only through heat diffusion in the material. The geometry of an angled defect can affect the temperature difference at the slot boundary during the heating process and lead to a change in the heat diffusion pattern. Therefore, if the value of temperature T(x,y,z,t) and the spatial and transient temperature profile can be obtained, they can be applied to detect and characterize the defects inside the sample. 2.2. RCF detection using ECPT In general, a simplified model of a defect is used in numerical studies to represent real defects. In the present work, the real RCF defect was represented as an angled defect, which propagates into the railhead at a shallow angle (10°–30°) to the rail running surface [Fig. 1]. Fig. 2 illustrates the principle of the ECPT system using Helmholtz coils to detect the RCF defect. An angled slot is defined as a surface crack over the full width of the sample, with an angle of propagation into the sample θ, but finite in depth (or slot propagation length into the metal) and width. Using Helmholtz coils, the eddy current is induced for a short period of time (typically, from 40 to 1000 ms), which is distributed symmetrically on the two sides of the railhead. The eddy current is perturbed by the angled slot, leading to the corresponding regions of

5

varying eddy current density. According to a study of rotation of a linear coil relative to the crack orientation [27], the eddy current in the railhead induced by the Helmholtz coils has a small oblique angle with the direction of the RCF crack and can be employed to heat the crack. For the case of an angled slot, eddy current deflection results in a higher eddy current density at the slot corner, augmenting the Joule heating [15] to the levels much greater than that found at a typical straight crack in the sample [28]. Meanwhile, the heat diffusion in the region around the angled slot is trapped between the slot edge and the sample surface, so that heat flux is limited to this bounded area. Therefore, the highest temperature occurs at the corner region of the angled slot. Thus, the defect can be identified from a characteristic heat distribution in the thermal image/video captured by an infrared camera. In this manner, induction heating and spatial heat distribution are used for characterization and quantitative nondestructive evaluation of defects. For the present work, spatial thermal features associated with the temperature distribution were studied under UMF excitation.

3. Numerical studies 3.1. Simulation setup The numerical studies were conducted using the finite element method (FEM) modeling package COMSOL Multiphysics. As shown in Fig. 3(a), a 3D FEM model comprised a sample (carbon steel) with an angled slot defect (filled with air) and Helmholtz coils. According to the previous work [18] and experimental setup, the sample surface was covered with thick black paint, and the high emissivity value of 0.95 was set in the simulation leading to a more homogeneous energy emissivity and image. Table 1 shows the material parameters that were used in the simulations. TABLE 1. Parameters used for the FEM model Parameters

Air

Steel

Conductivity (S/m)

0

5.0 × 106

Relative Permeability

1

100

Density (kg/m )

1.205

7.7 × 103

Heat capacity (J/(kg K))

1.005 × 103

490

Thermal conductivity (W/(m K))

0.0257

90

3

2

Thermal diffusivity (m /s)

2.12 × 10

Temperature coefficient α (1/℃)

--

−5

1.172 × 10−5 5.0 × 10−3

Using data from [14], a metal sample of dimensions 100 × 60 × 30 mm3 was placed at the center of the 6

Helmholtz coils with the radius of 100 mm. The excitation frequency and current were set at 256 kHz and 350 A, respectively, which are the same values as those chosen for the experiment (see Section 4). The heating period was set to 200 ms, the entire recording time was 2000 ms, and the initial temperature was set to 273 K (19.85 °C). The angled defect was an angled slot, which was cut along the width of the sample [Fig. 3(b)]; the width of the slot was 0.2 mm, the length l was set to 1.0, 1.5, 2.0, and 2.5 mm, and the slot angle θ of either 45° or 70° was chosen. As shown in Fig. 3(b), the angle θ is measured normal to the sample surface, so as the angle is increased, the slot is moved closer to the surface. From Eq. (1), the eddy current penetration depth was estimated to be approximately 0.05 mm, so the skin depth could be neglected; the induction heating was on the surface of the material and on the slot edge. The Helmholtz coils generated a UMF at the metal sample surface [Fig. 3(c)]. 3.2. Infrared image using UMF excitation Figure 4 shows the simulated infrared (IR) image for an angled slot (θ = 70°, l = 2.5 mm) using UMF excitation at the maximum Joule heating time of 200 ms. As shown in Fig. 4(a), the entire slot defect was heated using UMF excitation, and there were areas of high and low temperatures around the angled defect. Fig. 4(b) displays the slice view for the angled crack and a defect-free area. Fig. 4(a) shows the temperature measurements along four lines HN running in the y-direction with 1 mm offset were used to develop the temperature distribution at the defect crossing points, PosN (N=1,2,3,4) between HN and the slot corner. The temporal temperature profiles for PosN were extracted and plotted in Fig. 4(c); they are similar to the measured profiles at the four points with the maximum relative deviation of less than 1%. The temperature distribution in the y-direction was extracted at HN and is shown in Fig. 4(d); note that the spatial temperature distribution profiles are in a similar agreement at each position, and the maximum relative deviation is less than 1%. The thermal gradients for these four curves (dotted line in Fig. 4(d)), as plotted in Fig. 4(e), have equal values, which indicates approximately uniform heating. Therefore, the Helmholtz coils generate a uniform spatial and temporal temperature distribution for all HN positions. However, from Fig. 4(a), inhomogeneous heating and a higher temperature appeared at the two ends of the slot along the edge of the block because of the edge effects [27]. This means that although the UMF excitation using Helmholtz coils was employed in ECPT, the boundaries of the metal sample lead to its inhomogeneous heating. This could be removed using algorithms used in, for example, principal component analysis (PCA) and independent component analysis (ICA) [29]. In this manner, the heating pattern for an angled defect, which includes high temperatures at the slot 7

side and low temperatures at the opposing side with a maximum temperature at the slot corner, can be observed using ECPT with Helmholtz coils and extracting the thermal gradients from the temperature distribution at the slot side. 3.3. Influence of the angled slot geometry Temperature distributions were studied under UMF excitation for different slot angles and different slot lengths. The spatial thermal curves illustrate the temperature distribution for the 45° and 70° angled slots with the slot propagation lengths into the sample of l = 1.0, 1.5, 2.0, 2.5 mm (see Fig. 5(a) for 45° slot and Fig. 5(b) for 70° slot). The temperatures clearly change with the variation of the slot length. For the same angle of the slot, longer slots exhibit higher temperature variations than the shorter ones. This implies slower heat diffusion at the slot corner caused by the deep slot, which leads to higher temperature at the slot corner. The thermal gradients can also be obtained from the curves at the slot direction side and for a deep slot of longer length. The depth of the angled slot is dependent on the angle and length, and hence maintains the same rate of change in temperature. From Fig. 6, for the same slot length but larger slot angle, the curve has a steeper thermal gradient than that for the smaller slot angle, because the former has a higher eddy current density at the slot corner that induces higher Joule heating. Moreover, for the larger slot angle, heat trapped along the slot edge diffuses more effectively between the sample surface and the edge. In Fig. 7, the maximum temperatures for the 45° and 70° angled slots are compared after the heating time of 200 ms. The maximum temperature is dependent on the slot length and angle. For a given slot angle, such as 70°, the maximum temperature rises as the length of the slot increases. 3.4. Comparison of ECPT using UMF and NUMF excitation The spatial features for an angled slot using NUMF excitation from a linear coil were extracted in ECPT and compared with the features extracted using UMF excitation, as discussed above. As shown in Fig. 8(a), the 3D Finite Element Model (FEM) consisted of the same block with an artificial angled crack, as presented in Section 3.2, and a linear coil (typical diameter 4 mm; other excitation parameters in the simulation were the same as UMF excitation in section 3.2). The linear coil was placed above the specimen. The lift-off distance between the coil and sample was 1 mm. Fig. 8(b) shows an IR image for a 70° 2.5 mm-long angled slot at 200 ms, which was obtained in the simulation, subject to the same method of analysis as described in Section 3.2. Only a portion of the angled slot and the defect-free area under the linear coil was heated and could be used for analysis.

8

Four lines of measurements over the angled slot along the y-direction with offsets of 0.0, 1.0, 2.0, 3.0 mm from the y-axis were used to construct the temperature distribution after the heating time of 200 ms [Fig. 8(c)]. There was a large temperature change, when the position LN moved towards the central area. However, line L1 closest to the coil had the highest temperature, because of NUMF effects from the linear coil. In addition, the thermal gradient [Fig. 8(c)] drops from 1.5 to 1.18 between positions L1 and L4, which demonstrates a correlation between the offset and thermal gradient [Fig. 8(d)]. Therefore, it can be deduced that ECPT using UMF excitation provides a wider detection area in imaging and characterizing angled defects despite the persisting inhomogeneous heating phenomenon. This method overcomes the limitations of NUMF excitation, leading to a more efficient inspection without the problems associated with inductive shielding for large samples. Also, thermal gradients and temperature maxima are demonstrated to be a robust spatial feature of UMF excitation correlated to the angled slot geometry, which can be used to characterize angled defects.

4. Experimental Studies 4.1. ECPT setup An experimental ECPT system was constructed to inspect the RCF defects in a rail sample using UMF excitation. It included three key units: an inductive heating unit with Helmholtz coils, a test sample, and an IR camera unit with associated software. In this experiment, the excitation signal generated by the induction heater (Easyheat 224 from Cheltenham Induction Heating) had the frequency of 200 kHz and the current of 350 Arms. The IR camera (FLIR SC655sc) had an uncooled detector with the maximum frame size of 640 × 120 pixels on a 7.5–14.0 µm sensor with the sensitivity of ≤50 mK. The IR image was stored at a 200 Hz frame rate with the size of 640 × 120 pixels, and 2 s of video or IR image sequence was recorded during the experiment. Two samples were used in the experiment. One was a block with two artificial angled slots; the other was a rail sample with RCF defects on one side of the railhead gauge corner. 4.2. Block with artificial angled slot To verify the result of simulation analysis, a block with two angled cracks of the same size but different angles of propagation into the metal and a block with two artificial angled slots were tested. The sample size was 100 × 60 × 30 mm3, the two angled slots had the width of 0.3 mm, angles of 45° and 70°, and 2.5 mm length of propagation into metal. The samples were placed at the center of the Helmholtz coils, and an eddy current signal of 350 A, 200 kHz, and 200 ms heating time was applied, which are the values 9

used in the numerical studies. Figure 10 shows the IR image of the temperature distribution at 200 ms for heating of the 45° and 70° angled slots. The spatial temperature distributions at 200 ms heating time for the two angled slots with the same length (design value 2.5 mm) are presented in Fig. 11. The larger angled slot exhibits a higher temperature and steeper thermal gradient than the smaller angled slot. These spatial temperature features shown in Fig. 11 are similar to those in Fig. 6, Section 3.3, although the temperature profiles from the experimental data are somewhat broader than the theoretical ones. The machining error, which resulted in an altered real angled crack length (about 2.8–2.9 mm) in Fig. 10, is considered to result in the broader profile phenomenon. Hence, the simulation model discussed in this paper is demonstrated to be effective in simulating an actual ECPT system, and the simulation results are of practical significance. Similar work was performed to verify the results of ECPT using NUMF excitation [22]. Therefore, we can infer that ECPT can identify crack propagation angle of angled defects. 4.3. Rail sample with natural RCF Figure 12(a) shows an IR image of a railhead using UMF excitation with Helmholtz coils at 200 ms heating time. The RCF defect area is observed on the right-hand side of the rail from the variation in temperature; the area on the left-hand side is defect-free, where the significant temperature increase is observed under Joule heating because of edge effects. This defect-free area provided a good reference in RCF defect analysis, because most of the defects are encountered on only one side of the rail. Figure 12(b) shows an IR image using NUMF excitation with a linear coil. A narrow image area under the linear coil marked with a white rectangle is heated over a smaller area to enable a comparison with Fig. 12(a). UMF excitation provides a broad inspection area for RCF detection, according to the analysis results in Section 3.2. Figure 13 shows IR images of the RCF defects under UMF excitation [Fig. 12(a)] at 30, 100, 200 (maximum heating), and 300 ms (cooling stage). The shape of the defects became more apparent during the heating stage with the best profile seen at 200 ms when the temperature reached a maximum. At 200 ms, there were at least two RCF defects according to the pattern from the angled slot, which had hot and cold areas with a maximum temperature over the angled slot, in accordance with the simulation results (Sections 3.2 and 3.3). Moreover, the entire defect shape was obtained under UMF excitation without any shielding problems (cf. Fig. 13(c) with Fig. 12(a) obtained using a linear coil).

10

Figure 14 shows the temperature distribution curves obtained at defect A indicated in Fig. 13.(c), where the three measurement lines (denoted Ln, n = 1..3) run parallel to the x-axis, each with a position offset of 2 pixels (1 pixel ≈ 0.5 mm). The position shift of these three linear temperature distributions with respect to the x-axis is caused by the real defect at the rail surface being oblique to the rail track direction (x-axis), as shown in Fig. 13(c). The spatial temperature distributions of L1, L2, and L3 have similar profiles for the angled defect that range from 45 to 55 mm along the x-axis and have similar thermal gradients, compared to those in Fig. 4, Section 3.2. Moreover, the temperature distribution over the defect area from these three lines of measurement also maintains a similar distribution under UMF excitation. The experimental results indicate that UMF excitation can be employed for RCF defect detection and for evaluation of spatial thermal features by the analysis presented above. A UMF excited by Helmholtz coils is more suitable for defect detection in railheads than a NUMF excited by a linear coil. Using a UMF excitation, a larger inspection area is achieved with a stable temperature distribution. The spatial thermal pattern for the RCF defect at the crack side can be clearly observed in the IR images (Fig. 13(c) and Fig. 14), supporting the conclusions derived from simulation results in Section 3. However, the inhomogeneous heating phenomenon is still observed on the two sides of the railhead in Fig. 12(a) because of edge effects, as discussed in Sections 3.2 and 4.2. To reduce the influence of railhead geometry, optimization of inductive coils geometry can prove to be effective [26].

5. Discussion and Conclusions In this paper, an investigation of ECPT using Helmholtz coils for UMF excitation was performed to detect angled defects in railhead samples. The results were compared with those from NUMF excitation with a linear coil. The spatial thermal features, such as thermal gradients along the slot direction and the maximum temperature, were analyzed in simulations of metal samples with two angled defects in the form of angled slots. Finally, RCF defects in an actual rail sample were detected and imaged using the proposed ECPT method. The RCF in the railhead was detected by ECPT in a static way. Although there are still some technical issues needed to be overcome, the creation of a viable in-service system for rail inspection in a high-speed [23] or dynamic way becomes possible now, as new techniques are being developed, such as cameras with uncooled IR sensors with a high image frame rate up to 1 kHz, extraction of thermal features only at the maximum heating time, inductive heating coil design [26]. The slow heating time to reach the

11

maximum temperature can be overcome by a preheating process. Some image processing techniques could be used for motion deblurring of infrared images [9]. The following conclusions were obtained: (1) In ECPT, a UMF excited by Helmholtz coils was proven to provide a large and stable detection area compared with that performed by NUMF excitation with a linear coil; (2) The extracted spatial thermal features can be used to identify and characterize angled defects by ECPT, as verified by the simulation and experimental results. These spatial features include high- and low-temperature regions around the angled defect that identify it, a thermal gradient on the angled defect side, which depends on the angle of propagation into the metal, and the maximum temperature related to the slot geometry. For a fixed angle of the slot, longer slots exhibit higher temperature variations than the shorter ones. For the same length of the slot, a larger slot angle results in a curve with steeper thermal gradient than that for the smaller slot angle. Moreover, these features are more stable and robust under UMF excitation compared with NUMF excitation, although there is still the inhomogeneous heating phenomenon caused by edge effects under UMF excitation. (3) The RCF defects in the rail sample were tested and imaged under UMF excitation by Helmholtz coils, and the experimental results support the above conclusions. UMF excitation provides a means to characterize RCF defects, based on the entire defect image and spatial features. In further studies, quantitative characterization of angled defects will be investigated and employed to characterize RCF defects in railheads. Furthermore, the interaction of RCF defects in railheads will also be studied to characterize the properties of multiple cracks.

Acknowledgement The authors would like to thank the China Scholarship Council (CSC) for sponsoring a visit by Jianping Peng to Newcastle University, UK. The authors also would like to thank the EPSRC for funding the study through Novel Sensing Network for Intelligent Monitoring (EP/J012343/1).

12

References

[1] Cannon DF, Edel KO, Grassie SL, Sawley K. Rail defects: an overview. Fatigue & Fracture of Engineering Materials & Structures. 2003;26:865-86. [2] Grassie SL. Rolling contact fatigue on the British railway system: treatment. Wear. 2005;258:1310-8. [3] Ekberg A, Kabo E. Fatigue of railway wheels and rails under rolling contact and thermal loading—an overview. Wear. 2005;258:1288-300. [4] U.S. Department of Transportation FRA. Rolling Contact Fatigue: A Comprehensive Review. 2011. [5] Nicholson GL, Kostryzhev AG, Hao XJ, Davis CL. Modelling and experimental measurements of idealised and light-moderate RCF cracks in rails using an ACFM sensor. NDT & E International. 2011;44:427-37. [6] Farris TN, Keer LM, Steele RK. The effect of service loading on shell growth in rails. Journal of the Mechanics and Physics of Solids. 1987;35:677-700. [7] INNOTRACK. D4.4.1 Rail inspection Technologies. INNOYRACK. University of Birmingham ed2008. p. 42. [8] Ph Papaelias M, Roberts C, Davis CL. A review on non-destructive evaluation of rails: State-of-the-art and future development. Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid Transit. 2008;222:367-84. [9] Oswald-Tranta B, Sorger M, O’Leary P. Motion deblurring of infrared images from a microbolometer camera. Infrared Physics & Technology. 2010;53:274-9. [10] Tian GY, Sophian A. Defect classification using a new feature for pulsed eddy current sensors. NDT & E International. 2005;38:77-82. [11] Li W, Guoming C, Xiaokang Y, Chuanrong Z, Tao L. Analysis of the lift-off effect of a U-shaped ACFM system. NDT & E International. 2013;53:31-5. [12] Nicholson GL, Davis CL. Modelling of the response of an ACFM sensor to rail and rail wheel RCF cracks. NDT & E International. 2012;46:107-14. [13] Edwards RS, Fan Y, Papaelias M, Dixon S, Davis CL, Roberts C. ultrasonic detection of surface breaking railhead defects. AIP Conference Proceedings. 2008;975:602-9. [14] Tian GY, Wilson J, Cheng L, Almond DP, Kostson E, Weekes B. Pulsed Eddy Current Thermography and Applications. In: Mukhopadhyay S, editor. New Developments in Sensing Technology for Structural Health Monitoring: Springer Berlin Heidelberg; 2011. p. 205-31. [15] Wilson J, Tian GY, Abidin IZ, Yang S, Almond D. Modelling and evaluation of eddy current stimulated thermography. Nondestructive Testing and Evaluation. 2010;25:205-18. [16] Yin A, Gao B, Tian GY, Woo WL, Li K. Physical interpretation and separation of eddy current pulsed thermography. Journal of Applied Physics. 2013;113. [17] Netzelmann U, Walle G. Induction thermography as a tool for reliable detection of surface defects in forged components. 17th WCNDT. 2008. [18] Noethen M, Wolter KJ, Meyendorf N. Surface crack detection in ferritic and austenitic steel components using inductive heated thermography. Electronics Technology (ISSE), 2010 33rd International Spring Seminar on2010. p. 249-54. [19] He Y, Tian GY, Pan M, Chen D, Zhang H. An investigation into eddy current pulsed thermography for detection of corrosion blister. Corrosion Science. 2014;78:1-6. [20] Tian GY, Yin A, Gao B, Zhang J, Shaw B. Eddy current pulsed thermography for fatigue evaluation of gear. AIP Conference Proceedings. 2014;1581:1652-62. [21] Cheng L, Tian GY. Transient Thermal Behavior of Eddy-Current Pulsed Thermography for Nondestructive Evaluation of Composites. Instrumentation and Measurement, IEEE Transactions on. 2013;62:1215-22. [22] Ilham ZA, Tian GY, John W, Yang S, Almond D. Quantitative evaluation of angular defects by pulsed eddy current thermography. NDT & E International. 2010;43:537-46. [23] Wilson J, Tian GY, Mukriz I, Almond D. PEC thermography for imaging multiple cracks from rolling contact fatigue. NDT & E International. 2011;44:505-12. [24] Mathiak G, Egry I, Hennet L, Thiaudière D, Pozdnyakova I, Price D. Aerodynamic Levitation and Inductive Heating – A New Concept for Structural Investigations of Undercooled Melts. International Journal of Thermophysics. 2005;26:1151-66. [25] Li K, Tian GY, Cheng L, Yin A, Cao W, Crichton S. State Detection of Bond Wires in IGBT Modules using Eddy Current Pulsed Thermography. Power Electronics, IEEE Transactions on. 2013. [26] Oswald-Tranta B, Sorger M. Localizing surface cracks with inductive thermographical inspection: from measurement to image processing. Quantitative InfraRed Thermography Journal. 2011;8:149-64.

13

[27] Yang S, Tian GY, Abidin IZ, Wilson J. Simulation of edge cracks using pulsed eddy current stimulated thermography. Journal of Dynamic Systems, Measurement and Control, Transactions of the ASME. 2011;133. [28] Walle G, Netzelmann U. Thermographic crack detection in ferritic steel components using inductive heating. 2006. [29] Cheng L, Gao B, Tian GY, Woo WL, Berthiau G. Impact Damage Detection and Identification Using Eddy Current Pulsed Thermography Through Integration of PCA and ICA. Sensors Journal, IEEE. 2014;14:1655-63.

14

Highlights: •

ECPT with UMF excitation provides an efficient and robust method for angled crack.

• •

The results for UMF and NUMF excitations were compared to analyze the spatial feature’s robustness. ECPT with UMF excitation was employed to characterize RCF in railhead.

15

Fig. 1. RCF in rails: (a) GCC defect, (b) micrograph for RCF with a 28°-angle propagation into the railhead Fig. 2. The Diagram for angled defect detection by ECPT: (a) work principle of ECPT, (b) top view for angled slot Fig. 3. 3D FEM model for an angled slot: (a) 3D model for angled defect with Helmholtz coils, (b) angled slot, (c) uniform magnetic field distribution Fig. 4. Simulation results for angled slot by UMF excitation: (a) slot image at 200ms in slice view, (b) slot image at 200ms in top view, (c) PosN temporal temperature profile, (d) HN spatial temperature distribution at 200ms, (e) thermal gradient Fig. 5. Spatial temperature profile with different slot lengths at 200ms by simulation: (a) 45° slot, (b) 70° slot Fig. 6. Spatial temperature profile with different slot angles by simulation: (a) length 1.0mm, (b) length 2.5mm Fig. 7. Maximum temperature for various angled slot Fig. 8. Angled slot detection by a linear coil by simulation: (a) 3D FEM model, (b) slot image in top view at 200ms, (c) temperature distribution for LN at 200ms, (d) thermal gradient for LN Fig. 9. RCF detection experiment Fig. 10. IR image using UMF excitation for different angled slot at 200ms by experiment: (a) slot 45°, (b) slot 70° Fig. 11. Spatial temperature profile with different slot angle Fig. 12. IR image for RCF detection in rail sample at 200ms heating by experiment: (a) UMF excitation using Helmholtz coils, (b) NUMF excitation using line coil Fig. 13. IR image by UMF excitation at different times: (a) 20ms, (b) 100ms, (c) 200ms, (d) 300ms Fig. 14. Line temperature distribution for defect A marked in Fig. 13(c)

16

Fig. 1. RCF in rails: (a) GCC defect, (b) micrograph for RCF with a 28°-angle propagation into the railhead

IR Camera

a

Transient thermography

Image sequence

Transient Feature

PC

Spatial distribution

Defect characterization and features

Helmholtz Coil Eddy Current Generator eddy currents angled defect Induction Heater

Fig. 2. The Diagram for angled defect detection by ECPT: (a) work principle of ECPT, (b) top view for angled slot

Fig. 3. 3D FEM model for an angled slot: (a) 3D model for angled defect with Helmholtz coils, (b) angled slot, (c) uniform magnetic field distribution

17

Temperature response

Temperature (degC)

25 Pos1 Pos2 Pos3 Pos4

24 23 22 21 20

24.55 24.5 24.45 195

19 0

100

200

205

200 300 Time (ms)

400

500

2 1.5 1 0.5 0

1

2

3

4

Fig. 4. Simulation results for angled slot by UMF excitation: (a) slot image at 200ms in slice view, (b) slot image at 200ms in top view, (c) PosN temporal temperature profile, (d) HN spatial temperature distribution at 200ms, (e) thermal gradient at 200ms.

18

25

25

1.0mm 1.5mm 2.0mm 2.5mm

24 23

23

22

22

21

21

20

20

-10

-5

0

1.0mm 1.5mm 2.0mm 2.5mm

24

5

-10

-5

0

5

Fig. 5. Spatial temperature profile with different slot lengths at 200ms by simulation: (a) 45° slot, (b) 70° slot

25

23 70 degree 45 degree

22.5

70 degree 45 degree

24

22

23 21.5

22

21 20.5

21

20

20

19.5 -10

-5

0

-10

5

-5

0

5

Fig. 6. Spatial temperature profile with different slot angles by simulation: (a) length 1.0mm, (b) length 2.5mm

25 Slot 45 degree Slot 70 degree Temperature (

24

23

22

21 0.5

1

1.5

2

2.5

3

Slot length (mm)

Fig. 7. Maximum temperature for various angled slot geometries

19

25

)

24 Temperature (

1.6

L1=0mm L2=1mm L3=2mm L4=3mm

23

1.5 1.4 1.3

22

1.2

21 20

1.1 0

5

10

1

15

1

2

3

4

Fig. 8. Angled slot detection by a linear coil by simulation: (a) 3D FEM model, (b) slot image in top view at 200ms, (c) temperature distribution for LN at 200ms, (d) thermal gradient for LN

Fig. 9. RCF detection experiment 20

Fig. 10. IR image using UMF excitation for different angled slot at 200ms by experiment: (a) slot 45°, (b) slot 70°

38

45 degree 70 degree

36 34 32 30 30

35

40

Fig. 11. Spatial temperature profile with different slot angle at 200ms

21

Fig. 12. IR image for RCF detection in rail sample at 200ms heating by experiment: (a) UMF excitation using Helmholtz coils, (b) NUMF excitation using line coil

Fig. 13. IR image by UMF excitation at different times: (a) 20ms, (b) 100ms, (c) 200ms, (d) 300ms

22

Fig. 14. Line temperature distribution for defect A marked in Fig. 13(c)

23