sensors Article
A Flexible Arrayed Eddy Current Sensor for Inspection of Hollow Axle Inner Surfaces Zhenguo Sun 1,2, *, Dong Cai 1 , Cheng Zou 1 , Wenzeng Zhang 1 and Qiang Chen 1,2 1
2
*
Department of Mechanical Engineering, Tsinghua University, Beijing 100084, China; [email protected] (D.C.); [email protected] (C.Z.); [email protected] (W.Z.); [email protected] (Q.C.) Yangtze Delta Region Institute of Tsinghua University, Jiaxing 314006, China Correspondence: [email protected]; Tel.: +86-10-6277-3860
Academic Editor: Manuel Quevedo Received: 19 April 2016; Accepted: 21 June 2016; Published: 23 June 2016
Abstract: A reliable and accurate inspection of the hollow axle inner surface is important for the safe operation of high-speed trains. In order to improve the reliability of the inspection, a flexible arrayed eddy current sensor for non-destructive testing of the hollow axle inner surface was designed, fabricated and characterized. The sensor, consisting of two excitation traces and 28 sensing traces, was developed by using the flexible printed circuit board (FPCB) technique to conform the geometric features of the inner surfaces of the hollow axles. The main innovative aspect of the sensor was the new arrangement of excitation/sensing traces to achieve a differential configuration. Finite element model was established to analyze sensor responses and to determine the optimal excitation frequency. Experimental validations were conducted on a specimen with several artificial defects. Results from experiments and simulations were consistent with each other, with the maximum relative error less than 4%. Both results proved that the sensor was capable of detecting longitudinal and transverse defects with the depth of 0.5 mm under the optimal excitation frequency of 0.9 MHz. Keywords: eddy current; flexible arrayed sensor; hollow axle; frequency optimization
1. Introduction High-speed railways integrating a number of advanced rail technologies have been developed in many countries for passenger transportation and freight services, and they have become increasingly popular around the world due to their convenient, efficient, reliable and comfortable features. Axles are key components which are subject to dynamic loads in bogie systems, and may directly affect the safety of train operations. The traditional solid axles have been replaced by hollow axles to decrease the unsprung mass and consequently reduce the forces between rails and wheels. Failures and defects caused by material, processing, assembling, fatigue loading and complicated operating environments, can not only be observed on the outer surfaces of the hollow axles, but also on the inner surfaces. Thus, inspection of the inner surfaces of hollow axles plays a key role in the quality inspection process of hollow axles. So far, several non-destructive testing (NDT) techniques have been designed and developed for the inspection of hollow axles in practice. The ultrasonic testing (UT) technology has been widely used in the in-service NDT equipment to monitor hollow axles. Several ultrasonic transducers with different incident angles were arranged in a probe holder, which could rotate and move within a hollow axle, enabling the detection of transverse (circumferential) and longitudinal (axial) defects at different locations of the hollow axle covering both inner surfaces and outer surfaces [1,2]. The ultrasonic phased array technique has also been used in inspection of hollow axles along with the synthetic aperture focusing technique to improve the lateral resolution and testing efficiency [3,4]. In spite of Sensors 2016, 16, 952; doi:10.3390/s16070952
www.mdpi.com/journal/sensors
Sensors 2016, 16, 952
2 of 9
the high thickness-to-wavelength ratio, Li et al. [5] designed a guided wave-based structural health monitoring method for damage detection of hollow axles. Another inspection method using guided wave phenomenon in combination with a modified pulse-echo approach was presented by Ziaja et al. [6] to detect the cracks within the specific sections of a hollow axle. Ultrasonic inspection techniques play a dominant role in flaw detection of the hollow axle outer surface. However, UT is not sufficiently sensitive to identify shallow defects on the inner surfaces of axels due to the existence of natural waves, caused by the interface between the coupling liquid and the hollow axle, leading to the issue of false-calls [7]. Cavuto et al. [8] proposed a laser-ultrasonic technique for the inspection of the hollow axle outer surface. An air-coupled ultrasonic probe was utilized to detect the ultrasonic waves generated by a high-power pulsed laser. However, no inspection application on the inner surface was demonstrated. Induction thermography is an alternative method with high detection sensitivity and high testing speed for the outer surface inspection of a hollow axle [4], yet no inspection application on inner surfaces of hollow axles was mentioned. The electromagnetic method, such as eddy current testing, can work as an alternative to address the limitations of UT. Chady et al. [9] designed two kinds of probes based on flux leakage and eddy current, respectively, for the inspection of the hollow axle inner surface. One of the configuration consisted of the permanent magnet for magnetizing the ferromagnetic axle and a matrix of anisotropic magneto resistive three axis sensors. The other one used excitation coil to generate electromagnetic field which was picked up by Hall effect sensors. An integrated electromagnetic testing system combining eddy current testing with magnetic memory testing was developed to detect cracks and stress concentrations at the inner surfaces of hollow axles [7]. Flexible arrayed eddy current sensors, which are suitable for the inspection of components with complex geometries, are becoming a research hotspot [10–12]. Crouch et al. [13] used a flexible printed board with multiple eddy-current coil pairs to produce a rapid mapping of the external pipeline corrosion. Flexible arrayed eddy current sensors have also been applied in measurements of bond coat and top coat material properties and thicknesses, such as the flexible Meandering Winding Magnetometer array presented in [14]. Endo et al. [15] proposed a flexible arrayed eddy current probe with several coil pairs and a 12-decibel drop method for crack length evaluation. Another flexible arrayed eddy current sensor for condition-based maintenance of key components of aircraft was presented in [16] where the sensor was comprised of 64 elements with resolution of 0.8 mm and an algorithm used for sizing the crack length based on the sensor was also presented. A new flexible arrayed eddy current sensor for the inspection of the hollow axle inner surface is proposed in this paper. The design and the principle of the sensor are presented in Section 2. The validation of the sensor behaviors and the optimization of the excitation frequency by simulation and experiments are introduced subsequently. 2. Sensor Design Differential-mode eddy current probes are more sensitive to small discontinuities and have the advantage of suppressing lift-off effect, and thus have been widely used in eddy current testing [17]. Therefore, a flexible arrayed eddy current sensor in a differential mode, characterized by high detection efficiency due to multiple sensing elements, was designed to inspect the hollow axle inner surface. In order to obtain higher signal-to-noise ratio and better detection sensitivity, the sensor was configured as transmit/receive type. Copper traces in FPCB were used to form the excitation elements and the sensing elements, to guarantee the flexibility and consistency of the sensors. The four-layered FPCB with excitation traces and sensing traces was rolled and mounted on the sensor holder with 28.6 mm in diameter as shown in Figures 1 and 2. The sensor is symmetric about the x axis to work in the differential configuration. There are two independent excitation traces with the same alternating current flowing in the direction shown in Figure 2. The sensor consists of 28 sensing traces, among which every two adjacent traces along y axis are connected to function as a differential sensing pair. Each of the sensing elements has nine windings which are surrounded by the
Sensors 2016, 16, 952
3 of 9
Sensors 2016, 16, 952
3 of 9
excitation trace. Adjacent sensing traces along x axis and y axis are winded in the opposite direction. Sensors 2016, 16, 952 3 of 9 Excitation traces are 1 mm in width while the width of sensing traces is 0.1 mm with gaps of 0.2 mm direction. Excitation traces are 1 mm in width while the width of sensing traces is 0.1 mm with gaps direction. Excitation traces are 1 mm in width while the width of sensing traces is 0.1 mm with gaps between windings, and the element spacing is 6.5 mm. is 6.5 mm. of 0.2 mm between windings, and the element spacing of 0.2 mm between windings, and the element spacing is 6.5 mm.
(a) (a)
(b) (b)
(c) (c)
Figure 1. Novel flexible arrayed sensor (a) Bottom view of the unfolded sensor; (b) Top view of the Figure 1. 1. Novel flexible view of of the theunfolded unfoldedsensor; sensor;(b)(b) Top view Figure Novel flexiblearrayed arrayedsensor sensor (a) (a) Bottom Bottom view Top view of of thethe unfolded sensor; (c) Actual sensor; 1—Differential sensing pair, 2—Sensor holder. unfolded sensor; (c)(c) Actual 2—Sensorholder. holder. unfolded sensor; Actualsensor; sensor;1—Differential 1—Differential sensing sensing pair, 2—Sensor
Figure2. Schematic diagram of the sensor. A—Top layer; B—Mid-layer 1; C—Mid-layer 2; D—Bottom Figure2. Schematic diagram of the sensor. A—Top layer; B—Mid-layer 1; C—Mid-layer 2; D—Bottom Figure 2. 1—Sensing Schematic diagram of the sensor. A—Top layer; B—Mid-layer 1; C—Mid-layer 2; D—Bottom layer; traces; 2—Excitation traces; 3—Terminals of sensing traces; 4—Terminals of layer; 1—Sensing traces; 2—Excitation traces; 3—Terminals of sensing traces; 4—Terminals of layer; 1—Sensing traces; 2—Excitation traces; 3—Terminals of sensing traces; 4—Terminals of excitation excitation traces; 5—via hole. excitation traces; 5—via hole. traces; 5—via hole.
When an alternate current I flows in excitation traces, an alternating magnetic field B is When an alternate current I flows in excitation traces, an alternating magnetic field B is generated, which induces a voltage on each of the traces, sensingan traces given by: When anwhich alternate current I flowson ineach excitation alternating magnetic field B is generated, generated, induces a voltage of the sensing traces given by: which induces a voltage on each of the sensing traces by: N dgiven N d t ,Wi (1) V t ,Wi (1) Vtt iN 1 dt ÿ d t i 1 dφt,W i Vt “ ´ (1) where subscript t denotes the sensing trace number, Ndtis the number of windings, and t ,Wi is the where subscript t denotes the sensing trace number, i “ 1 N is the number of windings, and t ,W is the i magnetic field flux depending on the area defined by each winding Wi . For each of the differential W magnetic field flux depending on the area defined by each winding . For each of the differential where subscript t denotes the sensing trace number, N is the numberi of windings, and φt,Wi is the sensing pairs, the output voltage V out can be expressed as: V outarea can defined be expressed as: winding Wi . For each of the differential sensing field pairs,flux the depending output voltage magnetic on the by each sensing pairs, the output voltage as: Nout N d N d V dcan be expressed 1,Wi N d 1,Wi N d 2,Wi N d 2,2,WWi i V W 1, (2) W W 1, 2, i i ˜V out ¸ i ˜ ˘ (2) dt ¸ d`t N N N i 1 dt ÿ i 1 i 1 ÿ outÿ dφ dφ d φ ` φ t t t d d d 1,W 2,W 1,W 2,W 1 i i i 1 i ` ´ i1 i “i ´ Vout “ ´ (2) dt dt dt where subscripts 1 and i2“denote two sensing of a differential sensing pair, respectively. If 1 i “ traces 1 i“1 where subscripts 1 and 2 denote two sensing traces of a differential sensing pair, respectively. If sensing traces of a differential sensing pair are under the same circumstance, the induced current on sensing traces of a differential sensing pair are underathe same circumstance, therespectively. induced current on where subscripts 1 and denote twomagnitude, sensing traces differential sensing pair, sensing sensing traces will be2of the same but of the output voltage will remain close to zeroIfdue to sensing will besensing of the same butsame the output voltage will remain close to zero to traces a traces differential pair magnitude, aretraces under the induced current ondue sensing the of winding direction of the sensing and the flowcircumstance, direction of thethe excitation current. Otherwise, the winding direction of the sensing traces and the flow direction of the excitation current. Otherwise, traces will be contribution of the same butwill thenot output voltage remain close to the the voltage ofmagnitude, a sensing trace be canceled bywill the contribution ofto thezero otherdue sensing the voltage contribution of a sensing trace will not be canceled by the contribution of the other sensingthe winding direction of the sensing traces and the flow direction of the excitation current. Otherwise, trace leading to nonzero output voltages. trace leading to nonzero output voltages. voltage contribution of a sensing trace will not be canceled by the contribution of the other sensing trace leading to nonzero output voltages.
Sensors 2016, 16, 952
Sensors 2016, 16, 952 3. Finite Element Modeling
4 of 9
4 of 9
3.The Finite Elementsensor Modeling proposed was simulated using a finite element modeling (FEM) software, COMSOL Multiphysics, to validate the sensor behaviors and optimize the modeling excitation(FEM) frequencies. ToCOMSOL simplify the The proposed sensor was simulated using a finite element software, simulation model, excitation traces and a differential pair offrequencies. the sensor To were included Multiphysics, toonly validate the sensor behaviors and optimizesensing the excitation simplify in the with the hollow the artificial defect and the airsensing as shown 3. Thewere air gap themodel simulation model, onlyaxle, excitation traces and a differential pairinofFigure the sensor between the sensing traces and the hollow axle inner surface was 0.5 mm. A Lumped Element included in the model with the hollow axle, the artificial defect and the air as shown in Figure 3. Thenode gapto between traces the hollow axle inner surface was 0.5 mm. Element wasair used mimicthe thesensing insertion ofand a resistor between output boundaries of Aa Lumped differential sensing was usedthe to mimic insertion of a resistor output boundaries a differential sensing as pairnode to measure outputthe voltage generated bybetween eddy current. Excitationoftraces were modeled pair to measure the output voltage by eddy current. traces were modeled asaxle Single-Turn Coil nodes subjected to agenerated current excitation. Only Excitation a small segment of the hollow Single-Turn Coil nodes subjected to a current excitation. Only a small segment of the hollow axle was was simulated at the frequency domain with the consideration of time consumption and hardware simulated at the frequency domain with the consideration of time consumption and hardware requirement. The Magnetic Insulation interface was applied at all of the outer boundaries of the model, requirement. The Magnetic Insulation interface was applied at all of the outer boundaries of the which set the tangential components of the magnetic potential to zero at these boundaries. The Free model, which set the tangential components of the magnetic potential to zero at these boundaries. Tetrahedral node was added to create an unstructured tetrahedral mesh with a minimum element The Free Tetrahedral node was added to create an unstructured tetrahedral mesh with a minimum size element of 0.048size mm maximum elementelement growthgrowth rate of 1.35. TheThe hollow axle the air of and 0.048amm and a maximum rate of 1.35. hollow axlesteel steel and and the wereairapplied in the domain 3 and34,and respectively. The The electrical conductivity of of thethe hollow were applied in the domain 4, respectively. electrical conductivity hollowaxle axlesteel 6 S/m.6 Both excitation traces and sensing traces were defined as copper. The (ASTM 1050) is 5.655 ˆ 10 steel (ASTM 1050) is 5.655 × 10 S/m. Both excitation traces and sensing traces were defined as copper. domain of the artificial defect defect was considered as the The domain of the artificial was considered asair. the air.
Figure 3. Simulation model of the sensor. 1—Excitation traces; 2—Sensing traces; 3—Segment of
Figure 3. Simulation model of the sensor. 1—Excitation traces; 2—Sensing traces; 3—Segment of hollow axle; 4—Air; 5—Artificial Defect; 6—Lumped element. hollow axle; 4—Air; 5—Artificial Defect; 6—Lumped element.
At first, the model was simulated at 1 MHz with a defect located under one of the sensing traces. At first, the model was simulated at 1 MHz with a defect underin one of the traces. The distribution of the eddy current density in the hollow axle located steel is shown Figure 4a.sensing As can be eddy current at the inner surface of the hollow are similar to the4a. shape of be The seen, distribution of theloops, eddygenerated current density in the hollow axle steel axle, is shown in Figure As can traces.loops, The current density regions two segments of excitation is seen,excitation eddy current generated at of the inner below surface of adjacent the hollow axle, are similar totraces the shape much higher since thecurrent excitation current of these below adjacent segments is segments of the same of excitation traces. The density of regions two adjacent ofdirection. excitationThe traces existence of defect the distribution of the eddy current,segments making the is much higher since changes the excitation current of these adjacent is eddy of thecurrent same concentrate direction. The on two ends of the defect. These features can also be observed by the distribution of the magnetic flux existence of defect changes the distribution of the eddy current, making the eddy current concentrate density as shown in Figure 4b. on two ends of the defect. These features can also be observed by the distribution of the magnetic flux To obtain the sensor response at different excitation frequencies, the defect was positioned at density as shown in Figure 4b. different locations relative to the sensor. Figure 5a,b shows the change of the output voltage of a To obtain sensing the sensor at different was defect positioned differential pairresponse for different excitation excitation frequenciesfrequencies, and positionsthe of adefect transverse and at different locationsdefect, relative to the sensor. Figure 5a,b of shows the change offor thethese output a longitudinal respectively. Limited difference the shape of curves two voltage kinds of of a differential sensing pair for different excitation frequencies and positions of a transverse defect and a defects with the same defect depth was observed because of the different defect orientations. The longitudinal defect, respectively. Limited difference of the shape of curves for these two kinds of defects peak-to-peak values of the output voltages were measured to determine the optimal excitation withfrequency the same as defect depth was observed the the different defect orientations. The peak-to-peak shown in Figure 6. Figure because 6 shows of when excitation frequency is 0.9 MHz, the best response be obtained all ofmeasured the simulated defects. the optimal excitation frequency as shown values of the can output voltagesfor were to determine
in Figure 6. Figure 6 shows when the excitation frequency is 0.9 MHz, the best response can be obtained for all of the simulated defects.
Sensors 2016, 16, 952 Sensors 2016, 16, 952 Sensors 2016, 16, 952 Sensors 2016, 16, 952
5 of 9 5 of 9 5 of 9 5 of 9
(a) (b) (a) (b) (a) (b) Figure 4. (a) Distribution of the eddy current density (A/m22), f = 1MHz; (b) Distribution of the Figure ), ff = Figure 4. 4. (a) (a) Distribution Distribution of of the the eddy eddy current current density density (A/m (A/m2), = 1MHz; 1MHz; (b) (b) Distribution Distribution of of the the 2
magnetic density (T),ffof = 1the MHz. Figure 4. flux (a) Distribution eddy current density (A/m ), f = 1MHz; (b) Distribution of the magnetic magnetic flux flux density density (T), (T), f == 11 MHz. MHz. magnetic flux density (T), f = 1 MHz.
(a) (b) (a) (b) (a) (b) Figure 5. Output voltage of a differential sensing pair with the position sweep of defects at different
Figure 5. Output voltage of a differential sensing pair with the position sweep of defects at different Figure 5. Output voltage(a) of Transverse differential sensing pair with the position sweep of defects defects at at different excitation frequencies. defect with 0.5with mmthe in position depth; (b) Longitudinal defect with Figure 5. Output voltage of aa differential sensing pair sweep of different excitation frequencies.(a)(a)Transverse Transverse defect with 0.5 in mm in depth; (b) Longitudinal defect with excitation frequencies. defect with 0.5 mm depth; (b) Longitudinal defect with 0.5with mm 0.5 mm in frequencies. depth. excitation (a) Transverse defect with 0.5 mm in depth; (b) Longitudinal defect 0.5 mm in depth. in depth. 0.5 mm in depth.
Figure 6. Peak-to-peak voltage for different kinds of defects at different frequencies from simulation. Figure 6. Peak-to-peak voltage for different kinds of defects at different frequencies from simulation. (“0.5 mm, Longitudinal” denotes longitudinal defect with 0.5 in depth). Figure 6. voltage for different of at mm different frequencies Figure 6. Peak-to-peak Peak-to-peak voltage forlongitudinal different kinds kinds of defects defects different frequencies from from simulation. simulation. (“0.5 mm, Longitudinal” denotes defect with 0.5atmm in depth). (“0.5 mm, Longitudinal” denotes longitudinal defect with 0.5 mm in depth). (“0.5 mm, Longitudinal” denotes longitudinal defect with 0.5 mm in depth).
Sensors 2016, 16, 952
6 of 9
Sensors 2016, 16, 952
6 of 9
4. Experimental Validation 4. Experimental Validation
4.1. Experimental Set-up
4.1. Experimental Set-up
An experimental system was set up to validate the simulation results and the feasibility of the An experimental system was set up to validate the simulation results and the feasibility of the sensor as shown in Figure 7. The motion control platform consisted of stepper motors, stepper motor sensor as shown in Figure 7. The motion control platform consisted of stepper motors, stepper motor drivers and and a corresponding moduletoto implement the linear and rotary drivers a correspondingtransmitting transmitting module implement the linear and rotary motionsmotions of the of the sensor. generationofof excitation signal and detection of the differential output voltage sensor. The The generation thethe excitation signal and detection of the differential output voltage were were realized realizedby byaa RITEC RITECRAM-5000 RAM-5000measurement measurement system, which featured a broadband gated system, which featured a broadband gated RF RF amplifier, a unique tracking receiver, sensitivedetectors, detectors, gated integrators amplifier, a unique tracking receiver,quadrature quadrature phase phase sensitive gated integrators and and multiple frequency synthesizers. Sincethe themeasurement measurement system only two reception channels, a multiple frequency synthesizers. Since systemhad had only two reception channels, a multiplexer module designed forthe thereception reception of of of 14 differential sensing pairs.pairs. multiplexer module waswas designed for of output outputvoltages voltages 14 differential sensing the excitation andreception the reception channels configured the impedance matching Both Both the excitation and the channels were were configured with with the impedance matching network network to make the status of excitation and reception at their best. Processed data was transmitted to make the status of excitation and reception at their best. Processed data was transmitted to a PC to a PC with a LabVIEW interface for inspection configuration, motion control and results with a LabVIEW interface for inspection configuration, motion control and results visualization. visualization.
Figure Diagramblock block of of the system. Figure 7. 7. Diagram the experimental experimental system.
Experiments were conducted by using a hollow axle specimen made of ASTM 1050 steel. Both
Experiments were conducted by using hollow axleofspecimen made 1050 steel. longitudinal and transverse artificial defectsa with depths 0.5 mm, 1.0 mm of andASTM 2.0 mm studied in Both longitudinal transverse artificial defects depths of 0.5 mm, 1.0All mm and 2.0 studied the FEM and simulation were reproduced on the with specimen by electro-erosion. defects are mm 0.35 mm in in the FEM simulation reproduced on the specimen by activated electro-erosion. Allthe defects 0.35 mm in width and 10 mmwere in length. A differential sensing pair was to measure sensorare responses at different excitation frequencies for verification of simulation results. The inspection the width and 10 mm in length. A differential sensing pair was activated to measure the sensorofresponses specimen was carried out by using the differential sensing pairs with the and circumferential at different excitation frequencies for all verification of simulation results. Theaxial inspection of the specimen sampling intervals of 1 mm and 1.29°, respectively. In order to achieve the resolution, the sensor was carried out by using all the differential sensing pairs with the axial and circumferential sampling needed to rotate 20 times with a rotation angle of 1.29° after each axial step with a step distance of ˝ intervals of 1 mm and 1.29 , respectively. In order to achieve the resolution, the sensor needed to rotate 1 mm and then rotated reversely after next axial step with the same motion parameters. The excitation 20 times with a rotation angle of 1.29˝ after each axial step with a step distance of 1 mm and then and reception of the sensor were implemented after each rotation. The gain of the RITEC RAM-5000 rotated reversely after next step with the same motion parameters. The excitation and reception measurement system wasaxial set as 13 dB. of the sensor were implemented after each rotation. The gain of the RITEC RAM-5000 measurement system set as 13Results dB. 4.2.was Experimental Figure 8 shows the change of the output voltage for a linear sweep of the sensor position through the whole specimen. When the first sensing trace of the differential sensing pair approaches a defect, the output voltage from to itsvoltage minimum. it starts to of increase and approximates Figure 8 shows thedecreases change of thezero output for aThen linear sweep the sensor position through to zerospecimen. when the center thefirst pairsensing is aligned withofthe center of the defect. Subsequently, the voltage the whole Whenofthe trace the differential sensing pair approaches a defect, increases and then decreases as the sensor leaves the defect. The shapes of the curves for longitudinal the output voltage decreases from zero to its minimum. Then it starts to increase and approximates and transverse defects are different, which is consistent with the simulation results. The peak-to-peak to zero when the center of the pair is aligned with the center of the defect. Subsequently, the voltage values of the output voltages for different kinds of defects and different excitation frequencies were increases and then decreases as the sensor leaves the defect. The shapes of the curves for longitudinal obtained experimentally as shown in Figure 9. The peak-to-peak voltage achieves the maximum
4.2. Experimental Results
and transverse defects are different, which is consistent with the simulation results. The peak-to-peak values of the output voltages for different kinds of defects and different excitation frequencies were obtained experimentally as shown in Figure 9. The peak-to-peak voltage achieves the maximum value
Sensors 2016, 16, 952 Sensors 2016, 16, 952 Sensors 2016, 16, 952
7 of 9 7 of 9 7 of 9
value the when the excitation excitation frequency is 0.9 0.9which MHz is which is the theassame same as the the one mentioned mentioned in simulation simulation when excitation frequency is 0.9 MHz the same the one mentioned in simulation section value when the frequency is MHz which is as one in section above. The corresponding values from experiments and simulations are different because the above. The corresponding values from experiments and simulations are different because the voltage section above. The corresponding values from experiments and simulations are different because the voltage values from experiments were pre-amplified by the RITEC RAM-5000 measurement system values experiments were pre-amplified by the RITEC RAM-5000 measurement system while the voltagefrom values from experiments were pre-amplified by the RITEC RAM-5000 measurement system while the the voltage values from simulations simulations were measured measured directly from two ends of the thetrace. sensing trace. voltage values from simulations were measured directly from two from ends two of the sensing In order while voltage values from were directly ends of sensing trace. In quantitatively order to to quantitatively quantitatively compare results fromand experiments and simulations, the results results were from to compare results from results experiments simulations, thesimulations, results from simulations In order compare from experiments and the from simulations were multiplied by the gain used in the experiments and then the relative errors were multiplied the gain used inby thethe experiments then the relativeand errors were as shown simulationsbywere multiplied gain used and in the experiments then the calculated relative errors were calculated as shown in the Table 1. The maximum relative error is less than 4%. in the Table 1. The maximum relative error is less than 4%. calculated as shown in the Table 1. The maximum relative error is less than 4%.
Figure 8. 8. Output Output voltage voltage of of aa differential differential sensing sensing pair pair along along aa linear linear sweep sweep of of the the sensor sensor position position at at Figure 8. linear sweep the sensor position 0.9 MHz. 1—2.0 mm, longitudinal defect; 2—1.0 mm, longitudinal defect; 3—0.5 mm, longitudinal 0.9 MHz. MHz. 1—2.0 mm, longitudinal defect; 2—1.0 mm, longitudinal longitudinal defect; defect; 3—0.5 3—0.5 mm, mm, longitudinal longitudinal defect; 4—0.5 4—0.5 mm, mm, transverse transverse defect; defect; 5—1.0 5—1.0 mm, mm, transverse transverse defect; defect; 6—2.0 6—2.0mm, mm,transverse transversedefect. defect. defect; transverse defect; 5—1.0 mm, transverse defect; 6—2.0 mm, transverse defect.
Figure 9. 9. Peak-to-peak Peak-to-peak voltage voltage for for different different kinds kinds of of defects at at different frequencies frequencies from experiment. experiment. Figure Figure 9. Peak-to-peak voltage for different kinds of defects defects at different different frequenciesfrom from experiment.
relative errors errors of of results results from from experiment experiment and and simulation simulation (%). (%). Table 1. 1. The relative Table Table 1. The The relative errors of results from experiment and simulation (%). Frequency (MHz) (MHz) Frequency 0.6 Type 0.7 0.7 0.8 0.9 1.0 1.1 1.2 Defect 0.6 0.8 0.9 1.0 1.1 1.2 Defect Type 0.6 0.7 0.8 0.9 1.0 1.1 1.2 Defect Type Frequency (MHz) 2.0 mm, mm, longitudinal longitudinal 1.79 11 0.95 1.58 3.79 2.06 1.35 2.64 2.0 1.79 0.95 1.58 3.79 2.06 1.35 2.64 2.0 mm, longitudinal 1.79 1 0.75 0.95 1.58 3.79 2.17 2.06 1.352.232.64 2.71 1.0 mm, longitudinal 2.14 2.23 2.20 1.0 mm, longitudinal 2.14 2.23 0.75 2.20 2.17 2.23 2.71 1.0 mm, longitudinal 2.23 0.75 2.17 2.232.842.71 1.05 0.5 mm, mm, longitudinal longitudinal 2.72 0.76 2.14 0.53 0.53 2.59 2.20 0.07 0.07 0.5 2.72 0.76 2.59 2.84 0.5 mm, longitudinal 0.76 0.53 0.07 2.842.381.05 1.05 2.0 mm, mm, transverse transverse 1.28 1.54 2.72 1.43 1.43 3.14 2.59 1.35 1.35 0.76 2.0 1.28 1.54 3.14 2.0 mm, transverse 1.28 1.54 1.43 3.14 1.35 2.382.380.76 0.76 1.0 mm, transverse 0.23 0.11 3.84 3.95 2.15 0.23 2.56 1.0 mm, transverse 0.23 0.11 0.23 3.84 3.95 3.95 2.15 1.0 mm, transverse 0.11 3.84 2.15 0.230.232.56 2.56 0.5 mm, transverse 1.59 2.29 0.31 3.44 2.83 2.84 2.06 0.5 mm, transverse 1.59 2.29 1.59 0.31 3.44 3.44 2.83 0.5 mm, transverse 2.29 0.31 2.83 2.842.842.06 2.06 1 The relative errors are calculated using ˇ = V Vpp ,Exp denotes where V denotes V Vppˇˇ,Sim V Vpp ,Exp 100% 100% ,, where 1 The 1 relative errors are calculated using ˇ = Vpp , Exp The relative errors are calculated using δ= Vpp,Exppp ´ {Vpp,Exp the pp , Exp , ExpVpp,Sim pp ,Sim pp ,ˆ Exp100%, where Vpp,Exp denotes peak-to-peak voltage for different kinds of defects at different frequencies from experiments, V the pp,Sim the peak-to-peak peak-to-peak voltage voltage for for different different kinds kinds of of defects defects at different different frequencies frequencies from from experiments, experiments, the peak-to-peak voltage for different kinds of defects at differentatfrequencies from simulations.
V pp ,Sim the the peak-to-peak peak-to-peak voltage voltage for for different different kinds kinds of of defects defects at at different different frequencies frequencies from from V pp ,Sim
simulations. simulations.
Sensors 2016, 16, 952 Sensors 2016, 16, 952
8 of 9 8 of 9
All of of the 6 artificial defects were The inspection of the specimen was conducted at 0.9 MHz. All detected as shown in Figure 10, where the maximum noise amplitude is 0.012 V and the response longitudinal defect defect with with 0.5 0.5 mm mm depth depth is is 0.038 0.038 V. V. The defect signal amplitude is three amplitude of longitudinal times higher than the noise amplitude, which proves the feasibility of the proposed sensor. The noise mainly comes from electronic noises of the the RITEC RITEC RAM-5000 RAM-5000 measurement measurement system system and and lift-off lift-off effect. effect. The latter one is the dominant factor resulting from the processing error of the hollow axle inner be noticed, noticed, the the sensor sensor is ismore moresensitive sensitive to tothe thetransverse transversedefects. defects. surface and the sensor holder. As can be
Figure 10. Inspection result of the specimen. 1—2.0 mm, longitudinal defect; 2—1.0 mm, longitudinal Figure 10. Inspection result of the specimen. 1—2.0 mm, longitudinal defect; 2—1.0 mm, longitudinal defect; defect; 3—0.5 3—0.5 mm, mm, longitudinal longitudinal defect; defect; 4—0.5 4—0.5 mm, mm, transverse transverse defect; defect; 5—1.0 5—1.0 mm, mm, transverse transverse defect; defect; 6—2.0 transverse defect. defect. 6—2.0 mm, mm, transverse
5. Conclusions 5. Conclusions A flexible arrayed eddy current sensor has been presented. This new design is suitable for the A flexible arrayed eddy current sensor has been presented. This new design is suitable for the inspection of the hollow axle inner surfaces. The optimal excitation frequency, which is 0.9 MHz, is inspection of the hollow axle inner surfaces. The optimal excitation frequency, which is 0.9 MHz, is determined by FEM simulation, whose results are in good agreement with the experimental results. determined by FEM simulation, whose results are in good agreement with the experimental results. The maximum relative error between simulations and experiments is less than 4%. Results from The maximum relative error between simulations and experiments is less than 4%. Results from simulations and experiments show the sensor is capable of detecting both longitudinal and transverse simulations and experiments show the sensor is capable of detecting both longitudinal and transverse defects with depths as small as 0.5 mm. The sensor is more sensitive to the transverse defects, defects with depths as small as 0.5 mm. The sensor is more sensitive to the transverse defects, therefore therefore future work is required to increase the sensibility of the sensor to the longitudinal defects. future work is required to increase the sensibility of the sensor to the longitudinal defects. Author Contributions: Contributions: Z.S. Author Z.S. and and D.C. D.C. proposed proposed the the idea; idea; Z.S. Z.S. and and D.C. D.C. conceived conceived and and designed designed the the experiments; experiments; D.C. and C.Z. performed the experiments; D.C. also contributed to the writing and editing the manuscript. D.C. and C.Z. performed the experiments; D.C. also contributed to the writing and editing the manuscript. Z.S., Z.S., research andand were charged withwith the critical revisions of the of manuscript; all of the Q.C and and W.Z. W.Z.guided guidedthe the research were charged the critical revisions the manuscript; allauthors of the analyzedanalyzed the results. authors the results. Conflicts of Conflicts of Interest: Interest: The The authors authors declare declare no no conflict conflict of of interest. interest.
References References 1. 1.
2. 2. 3. 3. 4.
4. 5.
Marty, P.N.; Engl, G.; Krafft, S.; Spinelli, J.M. Latest development in the UT inspection of train wheels and Marty, P.N.; Engl, G.; Krafft, S.; Spinelli, J.M. Latest development in the UT inspection of train wheels and axles. In Proceedings of the 18th World Conference on Nondestructive Testing, Durban, South Africa, 16–20 axles. In Proceedings of the 18th World Conference on Nondestructive Testing, Durban, South Africa, April 2012. 16–20 April 2012. Carboni, M.; Cantini, S. Advanced ultrasonic “Probability of Detection” curves for designing in-service Carboni, M.; Cantini, S. Advanced ultrasonic “Probability of Detection” curves for designing in-service inspection intervals. Int. J. Fatigue 2016, 86, 77–87. [CrossRef] inspection intervals. Int. J. Fatigue 2016, 86, 77–87. Rudlin, J.; Raude, A.; Völz, U.; Conte, A.L. New methods of rail axle inspection and assessment. In Proceedings Rudlin, J.; Raude, A.; Völz, U.; Conte, A.L. New methods of rail axle inspection and assessment. of the 18th World Conference on Nondestructive Testing, Durban, South Africa, 16–20 April 2012. In Proceedings of the 18th World Conference on Nondestructive Testing, Durban, South Africa, 16–20 Kappes, W.; Hentschel, D.; Oelschlägel, T. Potential improvements of the presently applied in-service April 2012. inspection of wheelset axles. Int. J. Fatigue 2016, 86, 64–76. [CrossRef] Kappes, W.; Hentschel, D.; Oelschlägel, T. Potential improvements of the presently applied in-service inspection of wheelset axles. Int. J. Fatigue 2016, 86, 64–76. Li, F.; Sun, X.; Qiu, J.; Zhou, L.; Li, H.; Meng, G. Guided wave propagation in high-speed train axle and damage detection based on wave mode conversion. Struct. Control. Health Monit. 2015, 22, 1133–1147.
Sensors 2016, 16, 952
5.
6.
7.
8. 9.
10. 11. 12.
13. 14.
15.
16. 17.
9 of 9
Li, F.; Sun, X.; Qiu, J.; Zhou, L.; Li, H.; Meng, G. Guided wave propagation in high-speed train axle and damage detection based on wave mode conversion. Struct. Control. Health Monit. 2015, 22, 1133–1147. [CrossRef] Ziaja, A.; Cheng, L.; Radecki, R.; Packo, P.; Staszewski, W. Cylindrical guided wave approach for damage detection in hollow train axles. In Proceedings of the 10th International Workshop on Structural Health Monitoring, Palo Alto, CA, USA, 1–3 September 2015; pp. 2054–2061. Lin, J.; Lin, F.; Lin, C.; Zhang, K.; Yu, X. Hollow shaft array electromagnetic testing system based on integrated nondestructive testing techniques. In Electromagnetic Nondestructive Evaluation (XVIII); IOS Press: Washington, DC, USA, 2015; Volume 40, pp. 296–302. Cavuto, A.; Martarelli, M.; Pandarese, G.; Revel, G.M.; Tomasini, E.P. Experimental investigation by laser ultrasonics for high speed train axle diagnostics. Ultrasonics 2015, 55, 48–57. [CrossRef] [PubMed] Chady, T.; Psuj, G.; Kowalczyk, J.; Spychalski, I. Electromagnetic system for nondestructive evaluation of train hollow axles. In Proceedings of the 2013 Far East Forum on Nondestructive Evaluation/Testing: New Technology & Application, Szczecin, Poland, 17–20 June 2013. Bureau, J.; Ward, R.C.; Jenstead, W. Advances in eddy current array sensor technology. In Proceedings of the 18th World Conference on Nondestructive Testing, Shanghai, China, 25–28 October 2008. Lepage, B. Development of a flexible cross-wound eddy current array probe. Int. J. Appl. Electromagn. Mech. 2014, 45, 633–638. Marchand, B.; Decitre, J.M.; Sergeeva-Chollet, N.; Skarlatos, A. Development of flexible array eddy current probes for complex geometries and inspection of magnetic parts using magnetic sensors. In Proceedings of the Conference on Review of Progress in Quantitative Nondestructive Evaluation, Denver, CO, USA, 15–20 July 2012. Crouch, A.; Goyen, T.; Porter, P. New method uses conformable array to map external pipeline corrosion. Oil Gas. J. 2004, 102, 55–59. Sheiretov, Y.; Evans, L.; Schlicker, D.; Zilberstein, V.; Goldfine, N.; Sikorski, R. TBC characterization using magnetic and electric field sensors. In Proceedings of the ASME Turbo Expo 2007: Power for Land, Sea, and Air, Montreal, QC, Canada, 14–17 May 2007; pp. 97–103. Endo, H.; Nishimizu, A.; Tooma, M.; Ouchi, H.; Yoshida, I.; Nonaka, Y.; Otani, K. Signal evaluation system of flexible array ECT probes for inspecting complexly shaped surfaces. In Proceedings of the Conference on Review of Progress in Quantitative Nondestructive Evaluation, San Diego, CA, USA, 18–23 July 2010. Xie, R.; Chen, D.; Pan, M.; Tian, W.; Wu, X.; Zhou, W.; Tang, Y. Fatigue crack length sizing using a novel flexible eddy current sensor array. Sensors 2015, 15, 32138–32151. [CrossRef] [PubMed] García-Martín, J.; Gómez-Gil, J.; Vázquez-Sánchez, E. Non-destructive techniques based on eddy current testing. Sensors 2011, 11, 2525–2565. [CrossRef] [PubMed] © 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC-BY) license (http://creativecommons.org/licenses/by/4.0/).
A Flexible Arrayed Eddy Current Sensor for Inspection of Hollow Axle Inner Surfaces Zhenguo Sun 1,2, *, Dong Cai 1 , Cheng Zou 1 , Wenzeng Zhang 1 and Qiang Chen 1,2 1
2
*
Department of Mechanical Engineering, Tsinghua University, Beijing 100084, China; [email protected] (D.C.); [email protected] (C.Z.); [email protected] (W.Z.); [email protected] (Q.C.) Yangtze Delta Region Institute of Tsinghua University, Jiaxing 314006, China Correspondence: [email protected]; Tel.: +86-10-6277-3860
Academic Editor: Manuel Quevedo Received: 19 April 2016; Accepted: 21 June 2016; Published: 23 June 2016
Abstract: A reliable and accurate inspection of the hollow axle inner surface is important for the safe operation of high-speed trains. In order to improve the reliability of the inspection, a flexible arrayed eddy current sensor for non-destructive testing of the hollow axle inner surface was designed, fabricated and characterized. The sensor, consisting of two excitation traces and 28 sensing traces, was developed by using the flexible printed circuit board (FPCB) technique to conform the geometric features of the inner surfaces of the hollow axles. The main innovative aspect of the sensor was the new arrangement of excitation/sensing traces to achieve a differential configuration. Finite element model was established to analyze sensor responses and to determine the optimal excitation frequency. Experimental validations were conducted on a specimen with several artificial defects. Results from experiments and simulations were consistent with each other, with the maximum relative error less than 4%. Both results proved that the sensor was capable of detecting longitudinal and transverse defects with the depth of 0.5 mm under the optimal excitation frequency of 0.9 MHz. Keywords: eddy current; flexible arrayed sensor; hollow axle; frequency optimization
1. Introduction High-speed railways integrating a number of advanced rail technologies have been developed in many countries for passenger transportation and freight services, and they have become increasingly popular around the world due to their convenient, efficient, reliable and comfortable features. Axles are key components which are subject to dynamic loads in bogie systems, and may directly affect the safety of train operations. The traditional solid axles have been replaced by hollow axles to decrease the unsprung mass and consequently reduce the forces between rails and wheels. Failures and defects caused by material, processing, assembling, fatigue loading and complicated operating environments, can not only be observed on the outer surfaces of the hollow axles, but also on the inner surfaces. Thus, inspection of the inner surfaces of hollow axles plays a key role in the quality inspection process of hollow axles. So far, several non-destructive testing (NDT) techniques have been designed and developed for the inspection of hollow axles in practice. The ultrasonic testing (UT) technology has been widely used in the in-service NDT equipment to monitor hollow axles. Several ultrasonic transducers with different incident angles were arranged in a probe holder, which could rotate and move within a hollow axle, enabling the detection of transverse (circumferential) and longitudinal (axial) defects at different locations of the hollow axle covering both inner surfaces and outer surfaces [1,2]. The ultrasonic phased array technique has also been used in inspection of hollow axles along with the synthetic aperture focusing technique to improve the lateral resolution and testing efficiency [3,4]. In spite of Sensors 2016, 16, 952; doi:10.3390/s16070952
www.mdpi.com/journal/sensors
Sensors 2016, 16, 952
2 of 9
the high thickness-to-wavelength ratio, Li et al. [5] designed a guided wave-based structural health monitoring method for damage detection of hollow axles. Another inspection method using guided wave phenomenon in combination with a modified pulse-echo approach was presented by Ziaja et al. [6] to detect the cracks within the specific sections of a hollow axle. Ultrasonic inspection techniques play a dominant role in flaw detection of the hollow axle outer surface. However, UT is not sufficiently sensitive to identify shallow defects on the inner surfaces of axels due to the existence of natural waves, caused by the interface between the coupling liquid and the hollow axle, leading to the issue of false-calls [7]. Cavuto et al. [8] proposed a laser-ultrasonic technique for the inspection of the hollow axle outer surface. An air-coupled ultrasonic probe was utilized to detect the ultrasonic waves generated by a high-power pulsed laser. However, no inspection application on the inner surface was demonstrated. Induction thermography is an alternative method with high detection sensitivity and high testing speed for the outer surface inspection of a hollow axle [4], yet no inspection application on inner surfaces of hollow axles was mentioned. The electromagnetic method, such as eddy current testing, can work as an alternative to address the limitations of UT. Chady et al. [9] designed two kinds of probes based on flux leakage and eddy current, respectively, for the inspection of the hollow axle inner surface. One of the configuration consisted of the permanent magnet for magnetizing the ferromagnetic axle and a matrix of anisotropic magneto resistive three axis sensors. The other one used excitation coil to generate electromagnetic field which was picked up by Hall effect sensors. An integrated electromagnetic testing system combining eddy current testing with magnetic memory testing was developed to detect cracks and stress concentrations at the inner surfaces of hollow axles [7]. Flexible arrayed eddy current sensors, which are suitable for the inspection of components with complex geometries, are becoming a research hotspot [10–12]. Crouch et al. [13] used a flexible printed board with multiple eddy-current coil pairs to produce a rapid mapping of the external pipeline corrosion. Flexible arrayed eddy current sensors have also been applied in measurements of bond coat and top coat material properties and thicknesses, such as the flexible Meandering Winding Magnetometer array presented in [14]. Endo et al. [15] proposed a flexible arrayed eddy current probe with several coil pairs and a 12-decibel drop method for crack length evaluation. Another flexible arrayed eddy current sensor for condition-based maintenance of key components of aircraft was presented in [16] where the sensor was comprised of 64 elements with resolution of 0.8 mm and an algorithm used for sizing the crack length based on the sensor was also presented. A new flexible arrayed eddy current sensor for the inspection of the hollow axle inner surface is proposed in this paper. The design and the principle of the sensor are presented in Section 2. The validation of the sensor behaviors and the optimization of the excitation frequency by simulation and experiments are introduced subsequently. 2. Sensor Design Differential-mode eddy current probes are more sensitive to small discontinuities and have the advantage of suppressing lift-off effect, and thus have been widely used in eddy current testing [17]. Therefore, a flexible arrayed eddy current sensor in a differential mode, characterized by high detection efficiency due to multiple sensing elements, was designed to inspect the hollow axle inner surface. In order to obtain higher signal-to-noise ratio and better detection sensitivity, the sensor was configured as transmit/receive type. Copper traces in FPCB were used to form the excitation elements and the sensing elements, to guarantee the flexibility and consistency of the sensors. The four-layered FPCB with excitation traces and sensing traces was rolled and mounted on the sensor holder with 28.6 mm in diameter as shown in Figures 1 and 2. The sensor is symmetric about the x axis to work in the differential configuration. There are two independent excitation traces with the same alternating current flowing in the direction shown in Figure 2. The sensor consists of 28 sensing traces, among which every two adjacent traces along y axis are connected to function as a differential sensing pair. Each of the sensing elements has nine windings which are surrounded by the
Sensors 2016, 16, 952
3 of 9
Sensors 2016, 16, 952
3 of 9
excitation trace. Adjacent sensing traces along x axis and y axis are winded in the opposite direction. Sensors 2016, 16, 952 3 of 9 Excitation traces are 1 mm in width while the width of sensing traces is 0.1 mm with gaps of 0.2 mm direction. Excitation traces are 1 mm in width while the width of sensing traces is 0.1 mm with gaps direction. Excitation traces are 1 mm in width while the width of sensing traces is 0.1 mm with gaps between windings, and the element spacing is 6.5 mm. is 6.5 mm. of 0.2 mm between windings, and the element spacing of 0.2 mm between windings, and the element spacing is 6.5 mm.
(a) (a)
(b) (b)
(c) (c)
Figure 1. Novel flexible arrayed sensor (a) Bottom view of the unfolded sensor; (b) Top view of the Figure 1. 1. Novel flexible view of of the theunfolded unfoldedsensor; sensor;(b)(b) Top view Figure Novel flexiblearrayed arrayedsensor sensor (a) (a) Bottom Bottom view Top view of of thethe unfolded sensor; (c) Actual sensor; 1—Differential sensing pair, 2—Sensor holder. unfolded sensor; (c)(c) Actual 2—Sensorholder. holder. unfolded sensor; Actualsensor; sensor;1—Differential 1—Differential sensing sensing pair, 2—Sensor
Figure2. Schematic diagram of the sensor. A—Top layer; B—Mid-layer 1; C—Mid-layer 2; D—Bottom Figure2. Schematic diagram of the sensor. A—Top layer; B—Mid-layer 1; C—Mid-layer 2; D—Bottom Figure 2. 1—Sensing Schematic diagram of the sensor. A—Top layer; B—Mid-layer 1; C—Mid-layer 2; D—Bottom layer; traces; 2—Excitation traces; 3—Terminals of sensing traces; 4—Terminals of layer; 1—Sensing traces; 2—Excitation traces; 3—Terminals of sensing traces; 4—Terminals of layer; 1—Sensing traces; 2—Excitation traces; 3—Terminals of sensing traces; 4—Terminals of excitation excitation traces; 5—via hole. excitation traces; 5—via hole. traces; 5—via hole.
When an alternate current I flows in excitation traces, an alternating magnetic field B is When an alternate current I flows in excitation traces, an alternating magnetic field B is generated, which induces a voltage on each of the traces, sensingan traces given by: When anwhich alternate current I flowson ineach excitation alternating magnetic field B is generated, generated, induces a voltage of the sensing traces given by: which induces a voltage on each of the sensing traces by: N dgiven N d t ,Wi (1) V t ,Wi (1) Vtt iN 1 dt ÿ d t i 1 dφt,W i Vt “ ´ (1) where subscript t denotes the sensing trace number, Ndtis the number of windings, and t ,Wi is the where subscript t denotes the sensing trace number, i “ 1 N is the number of windings, and t ,W is the i magnetic field flux depending on the area defined by each winding Wi . For each of the differential W magnetic field flux depending on the area defined by each winding . For each of the differential where subscript t denotes the sensing trace number, N is the numberi of windings, and φt,Wi is the sensing pairs, the output voltage V out can be expressed as: V outarea can defined be expressed as: winding Wi . For each of the differential sensing field pairs,flux the depending output voltage magnetic on the by each sensing pairs, the output voltage as: Nout N d N d V dcan be expressed 1,Wi N d 1,Wi N d 2,Wi N d 2,2,WWi i V W 1, (2) W W 1, 2, i i ˜V out ¸ i ˜ ˘ (2) dt ¸ d`t N N N i 1 dt ÿ i 1 i 1 ÿ outÿ dφ dφ d φ ` φ t t t d d d 1,W 2,W 1,W 2,W 1 i i i 1 i ` ´ i1 i “i ´ Vout “ ´ (2) dt dt dt where subscripts 1 and i2“denote two sensing of a differential sensing pair, respectively. If 1 i “ traces 1 i“1 where subscripts 1 and 2 denote two sensing traces of a differential sensing pair, respectively. If sensing traces of a differential sensing pair are under the same circumstance, the induced current on sensing traces of a differential sensing pair are underathe same circumstance, therespectively. induced current on where subscripts 1 and denote twomagnitude, sensing traces differential sensing pair, sensing sensing traces will be2of the same but of the output voltage will remain close to zeroIfdue to sensing will besensing of the same butsame the output voltage will remain close to zero to traces a traces differential pair magnitude, aretraces under the induced current ondue sensing the of winding direction of the sensing and the flowcircumstance, direction of thethe excitation current. Otherwise, the winding direction of the sensing traces and the flow direction of the excitation current. Otherwise, traces will be contribution of the same butwill thenot output voltage remain close to the the voltage ofmagnitude, a sensing trace be canceled bywill the contribution ofto thezero otherdue sensing the voltage contribution of a sensing trace will not be canceled by the contribution of the other sensingthe winding direction of the sensing traces and the flow direction of the excitation current. Otherwise, trace leading to nonzero output voltages. trace leading to nonzero output voltages. voltage contribution of a sensing trace will not be canceled by the contribution of the other sensing trace leading to nonzero output voltages.
Sensors 2016, 16, 952
Sensors 2016, 16, 952 3. Finite Element Modeling
4 of 9
4 of 9
3.The Finite Elementsensor Modeling proposed was simulated using a finite element modeling (FEM) software, COMSOL Multiphysics, to validate the sensor behaviors and optimize the modeling excitation(FEM) frequencies. ToCOMSOL simplify the The proposed sensor was simulated using a finite element software, simulation model, excitation traces and a differential pair offrequencies. the sensor To were included Multiphysics, toonly validate the sensor behaviors and optimizesensing the excitation simplify in the with the hollow the artificial defect and the airsensing as shown 3. Thewere air gap themodel simulation model, onlyaxle, excitation traces and a differential pairinofFigure the sensor between the sensing traces and the hollow axle inner surface was 0.5 mm. A Lumped Element included in the model with the hollow axle, the artificial defect and the air as shown in Figure 3. Thenode gapto between traces the hollow axle inner surface was 0.5 mm. Element wasair used mimicthe thesensing insertion ofand a resistor between output boundaries of Aa Lumped differential sensing was usedthe to mimic insertion of a resistor output boundaries a differential sensing as pairnode to measure outputthe voltage generated bybetween eddy current. Excitationoftraces were modeled pair to measure the output voltage by eddy current. traces were modeled asaxle Single-Turn Coil nodes subjected to agenerated current excitation. Only Excitation a small segment of the hollow Single-Turn Coil nodes subjected to a current excitation. Only a small segment of the hollow axle was was simulated at the frequency domain with the consideration of time consumption and hardware simulated at the frequency domain with the consideration of time consumption and hardware requirement. The Magnetic Insulation interface was applied at all of the outer boundaries of the model, requirement. The Magnetic Insulation interface was applied at all of the outer boundaries of the which set the tangential components of the magnetic potential to zero at these boundaries. The Free model, which set the tangential components of the magnetic potential to zero at these boundaries. Tetrahedral node was added to create an unstructured tetrahedral mesh with a minimum element The Free Tetrahedral node was added to create an unstructured tetrahedral mesh with a minimum size element of 0.048size mm maximum elementelement growthgrowth rate of 1.35. TheThe hollow axle the air of and 0.048amm and a maximum rate of 1.35. hollow axlesteel steel and and the wereairapplied in the domain 3 and34,and respectively. The The electrical conductivity of of thethe hollow were applied in the domain 4, respectively. electrical conductivity hollowaxle axlesteel 6 S/m.6 Both excitation traces and sensing traces were defined as copper. The (ASTM 1050) is 5.655 ˆ 10 steel (ASTM 1050) is 5.655 × 10 S/m. Both excitation traces and sensing traces were defined as copper. domain of the artificial defect defect was considered as the The domain of the artificial was considered asair. the air.
Figure 3. Simulation model of the sensor. 1—Excitation traces; 2—Sensing traces; 3—Segment of
Figure 3. Simulation model of the sensor. 1—Excitation traces; 2—Sensing traces; 3—Segment of hollow axle; 4—Air; 5—Artificial Defect; 6—Lumped element. hollow axle; 4—Air; 5—Artificial Defect; 6—Lumped element.
At first, the model was simulated at 1 MHz with a defect located under one of the sensing traces. At first, the model was simulated at 1 MHz with a defect underin one of the traces. The distribution of the eddy current density in the hollow axle located steel is shown Figure 4a.sensing As can be eddy current at the inner surface of the hollow are similar to the4a. shape of be The seen, distribution of theloops, eddygenerated current density in the hollow axle steel axle, is shown in Figure As can traces.loops, The current density regions two segments of excitation is seen,excitation eddy current generated at of the inner below surface of adjacent the hollow axle, are similar totraces the shape much higher since thecurrent excitation current of these below adjacent segments is segments of the same of excitation traces. The density of regions two adjacent ofdirection. excitationThe traces existence of defect the distribution of the eddy current,segments making the is much higher since changes the excitation current of these adjacent is eddy of thecurrent same concentrate direction. The on two ends of the defect. These features can also be observed by the distribution of the magnetic flux existence of defect changes the distribution of the eddy current, making the eddy current concentrate density as shown in Figure 4b. on two ends of the defect. These features can also be observed by the distribution of the magnetic flux To obtain the sensor response at different excitation frequencies, the defect was positioned at density as shown in Figure 4b. different locations relative to the sensor. Figure 5a,b shows the change of the output voltage of a To obtain sensing the sensor at different was defect positioned differential pairresponse for different excitation excitation frequenciesfrequencies, and positionsthe of adefect transverse and at different locationsdefect, relative to the sensor. Figure 5a,b of shows the change offor thethese output a longitudinal respectively. Limited difference the shape of curves two voltage kinds of of a differential sensing pair for different excitation frequencies and positions of a transverse defect and a defects with the same defect depth was observed because of the different defect orientations. The longitudinal defect, respectively. Limited difference of the shape of curves for these two kinds of defects peak-to-peak values of the output voltages were measured to determine the optimal excitation withfrequency the same as defect depth was observed the the different defect orientations. The peak-to-peak shown in Figure 6. Figure because 6 shows of when excitation frequency is 0.9 MHz, the best response be obtained all ofmeasured the simulated defects. the optimal excitation frequency as shown values of the can output voltagesfor were to determine
in Figure 6. Figure 6 shows when the excitation frequency is 0.9 MHz, the best response can be obtained for all of the simulated defects.
Sensors 2016, 16, 952 Sensors 2016, 16, 952 Sensors 2016, 16, 952 Sensors 2016, 16, 952
5 of 9 5 of 9 5 of 9 5 of 9
(a) (b) (a) (b) (a) (b) Figure 4. (a) Distribution of the eddy current density (A/m22), f = 1MHz; (b) Distribution of the Figure ), ff = Figure 4. 4. (a) (a) Distribution Distribution of of the the eddy eddy current current density density (A/m (A/m2), = 1MHz; 1MHz; (b) (b) Distribution Distribution of of the the 2
magnetic density (T),ffof = 1the MHz. Figure 4. flux (a) Distribution eddy current density (A/m ), f = 1MHz; (b) Distribution of the magnetic magnetic flux flux density density (T), (T), f == 11 MHz. MHz. magnetic flux density (T), f = 1 MHz.
(a) (b) (a) (b) (a) (b) Figure 5. Output voltage of a differential sensing pair with the position sweep of defects at different
Figure 5. Output voltage of a differential sensing pair with the position sweep of defects at different Figure 5. Output voltage(a) of Transverse differential sensing pair with the position sweep of defects defects at at different excitation frequencies. defect with 0.5with mmthe in position depth; (b) Longitudinal defect with Figure 5. Output voltage of aa differential sensing pair sweep of different excitation frequencies.(a)(a)Transverse Transverse defect with 0.5 in mm in depth; (b) Longitudinal defect with excitation frequencies. defect with 0.5 mm depth; (b) Longitudinal defect with 0.5with mm 0.5 mm in frequencies. depth. excitation (a) Transverse defect with 0.5 mm in depth; (b) Longitudinal defect 0.5 mm in depth. in depth. 0.5 mm in depth.
Figure 6. Peak-to-peak voltage for different kinds of defects at different frequencies from simulation. Figure 6. Peak-to-peak voltage for different kinds of defects at different frequencies from simulation. (“0.5 mm, Longitudinal” denotes longitudinal defect with 0.5 in depth). Figure 6. voltage for different of at mm different frequencies Figure 6. Peak-to-peak Peak-to-peak voltage forlongitudinal different kinds kinds of defects defects different frequencies from from simulation. simulation. (“0.5 mm, Longitudinal” denotes defect with 0.5atmm in depth). (“0.5 mm, Longitudinal” denotes longitudinal defect with 0.5 mm in depth). (“0.5 mm, Longitudinal” denotes longitudinal defect with 0.5 mm in depth).
Sensors 2016, 16, 952
6 of 9
Sensors 2016, 16, 952
6 of 9
4. Experimental Validation 4. Experimental Validation
4.1. Experimental Set-up
4.1. Experimental Set-up
An experimental system was set up to validate the simulation results and the feasibility of the An experimental system was set up to validate the simulation results and the feasibility of the sensor as shown in Figure 7. The motion control platform consisted of stepper motors, stepper motor sensor as shown in Figure 7. The motion control platform consisted of stepper motors, stepper motor drivers and and a corresponding moduletoto implement the linear and rotary drivers a correspondingtransmitting transmitting module implement the linear and rotary motionsmotions of the of the sensor. generationofof excitation signal and detection of the differential output voltage sensor. The The generation thethe excitation signal and detection of the differential output voltage were were realized realizedby byaa RITEC RITECRAM-5000 RAM-5000measurement measurement system, which featured a broadband gated system, which featured a broadband gated RF RF amplifier, a unique tracking receiver, sensitivedetectors, detectors, gated integrators amplifier, a unique tracking receiver,quadrature quadrature phase phase sensitive gated integrators and and multiple frequency synthesizers. Sincethe themeasurement measurement system only two reception channels, a multiple frequency synthesizers. Since systemhad had only two reception channels, a multiplexer module designed forthe thereception reception of of of 14 differential sensing pairs.pairs. multiplexer module waswas designed for of output outputvoltages voltages 14 differential sensing the excitation andreception the reception channels configured the impedance matching Both Both the excitation and the channels were were configured with with the impedance matching network network to make the status of excitation and reception at their best. Processed data was transmitted to make the status of excitation and reception at their best. Processed data was transmitted to a PC to a PC with a LabVIEW interface for inspection configuration, motion control and results with a LabVIEW interface for inspection configuration, motion control and results visualization. visualization.
Figure Diagramblock block of of the system. Figure 7. 7. Diagram the experimental experimental system.
Experiments were conducted by using a hollow axle specimen made of ASTM 1050 steel. Both
Experiments were conducted by using hollow axleofspecimen made 1050 steel. longitudinal and transverse artificial defectsa with depths 0.5 mm, 1.0 mm of andASTM 2.0 mm studied in Both longitudinal transverse artificial defects depths of 0.5 mm, 1.0All mm and 2.0 studied the FEM and simulation were reproduced on the with specimen by electro-erosion. defects are mm 0.35 mm in in the FEM simulation reproduced on the specimen by activated electro-erosion. Allthe defects 0.35 mm in width and 10 mmwere in length. A differential sensing pair was to measure sensorare responses at different excitation frequencies for verification of simulation results. The inspection the width and 10 mm in length. A differential sensing pair was activated to measure the sensorofresponses specimen was carried out by using the differential sensing pairs with the and circumferential at different excitation frequencies for all verification of simulation results. Theaxial inspection of the specimen sampling intervals of 1 mm and 1.29°, respectively. In order to achieve the resolution, the sensor was carried out by using all the differential sensing pairs with the axial and circumferential sampling needed to rotate 20 times with a rotation angle of 1.29° after each axial step with a step distance of ˝ intervals of 1 mm and 1.29 , respectively. In order to achieve the resolution, the sensor needed to rotate 1 mm and then rotated reversely after next axial step with the same motion parameters. The excitation 20 times with a rotation angle of 1.29˝ after each axial step with a step distance of 1 mm and then and reception of the sensor were implemented after each rotation. The gain of the RITEC RAM-5000 rotated reversely after next step with the same motion parameters. The excitation and reception measurement system wasaxial set as 13 dB. of the sensor were implemented after each rotation. The gain of the RITEC RAM-5000 measurement system set as 13Results dB. 4.2.was Experimental Figure 8 shows the change of the output voltage for a linear sweep of the sensor position through the whole specimen. When the first sensing trace of the differential sensing pair approaches a defect, the output voltage from to itsvoltage minimum. it starts to of increase and approximates Figure 8 shows thedecreases change of thezero output for aThen linear sweep the sensor position through to zerospecimen. when the center thefirst pairsensing is aligned withofthe center of the defect. Subsequently, the voltage the whole Whenofthe trace the differential sensing pair approaches a defect, increases and then decreases as the sensor leaves the defect. The shapes of the curves for longitudinal the output voltage decreases from zero to its minimum. Then it starts to increase and approximates and transverse defects are different, which is consistent with the simulation results. The peak-to-peak to zero when the center of the pair is aligned with the center of the defect. Subsequently, the voltage values of the output voltages for different kinds of defects and different excitation frequencies were increases and then decreases as the sensor leaves the defect. The shapes of the curves for longitudinal obtained experimentally as shown in Figure 9. The peak-to-peak voltage achieves the maximum
4.2. Experimental Results
and transverse defects are different, which is consistent with the simulation results. The peak-to-peak values of the output voltages for different kinds of defects and different excitation frequencies were obtained experimentally as shown in Figure 9. The peak-to-peak voltage achieves the maximum value
Sensors 2016, 16, 952 Sensors 2016, 16, 952 Sensors 2016, 16, 952
7 of 9 7 of 9 7 of 9
value the when the excitation excitation frequency is 0.9 0.9which MHz is which is the theassame same as the the one mentioned mentioned in simulation simulation when excitation frequency is 0.9 MHz the same the one mentioned in simulation section value when the frequency is MHz which is as one in section above. The corresponding values from experiments and simulations are different because the above. The corresponding values from experiments and simulations are different because the voltage section above. The corresponding values from experiments and simulations are different because the voltage values from experiments were pre-amplified by the RITEC RAM-5000 measurement system values experiments were pre-amplified by the RITEC RAM-5000 measurement system while the voltagefrom values from experiments were pre-amplified by the RITEC RAM-5000 measurement system while the the voltage values from simulations simulations were measured measured directly from two ends of the thetrace. sensing trace. voltage values from simulations were measured directly from two from ends two of the sensing In order while voltage values from were directly ends of sensing trace. In quantitatively order to to quantitatively quantitatively compare results fromand experiments and simulations, the results results were from to compare results from results experiments simulations, thesimulations, results from simulations In order compare from experiments and the from simulations were multiplied by the gain used in the experiments and then the relative errors were multiplied the gain used inby thethe experiments then the relativeand errors were as shown simulationsbywere multiplied gain used and in the experiments then the calculated relative errors were calculated as shown in the Table 1. The maximum relative error is less than 4%. in the Table 1. The maximum relative error is less than 4%. calculated as shown in the Table 1. The maximum relative error is less than 4%.
Figure 8. 8. Output Output voltage voltage of of aa differential differential sensing sensing pair pair along along aa linear linear sweep sweep of of the the sensor sensor position position at at Figure 8. linear sweep the sensor position 0.9 MHz. 1—2.0 mm, longitudinal defect; 2—1.0 mm, longitudinal defect; 3—0.5 mm, longitudinal 0.9 MHz. MHz. 1—2.0 mm, longitudinal defect; 2—1.0 mm, longitudinal longitudinal defect; defect; 3—0.5 3—0.5 mm, mm, longitudinal longitudinal defect; 4—0.5 4—0.5 mm, mm, transverse transverse defect; defect; 5—1.0 5—1.0 mm, mm, transverse transverse defect; defect; 6—2.0 6—2.0mm, mm,transverse transversedefect. defect. defect; transverse defect; 5—1.0 mm, transverse defect; 6—2.0 mm, transverse defect.
Figure 9. 9. Peak-to-peak Peak-to-peak voltage voltage for for different different kinds kinds of of defects at at different frequencies frequencies from experiment. experiment. Figure Figure 9. Peak-to-peak voltage for different kinds of defects defects at different different frequenciesfrom from experiment.
relative errors errors of of results results from from experiment experiment and and simulation simulation (%). (%). Table 1. 1. The relative Table Table 1. The The relative errors of results from experiment and simulation (%). Frequency (MHz) (MHz) Frequency 0.6 Type 0.7 0.7 0.8 0.9 1.0 1.1 1.2 Defect 0.6 0.8 0.9 1.0 1.1 1.2 Defect Type 0.6 0.7 0.8 0.9 1.0 1.1 1.2 Defect Type Frequency (MHz) 2.0 mm, mm, longitudinal longitudinal 1.79 11 0.95 1.58 3.79 2.06 1.35 2.64 2.0 1.79 0.95 1.58 3.79 2.06 1.35 2.64 2.0 mm, longitudinal 1.79 1 0.75 0.95 1.58 3.79 2.17 2.06 1.352.232.64 2.71 1.0 mm, longitudinal 2.14 2.23 2.20 1.0 mm, longitudinal 2.14 2.23 0.75 2.20 2.17 2.23 2.71 1.0 mm, longitudinal 2.23 0.75 2.17 2.232.842.71 1.05 0.5 mm, mm, longitudinal longitudinal 2.72 0.76 2.14 0.53 0.53 2.59 2.20 0.07 0.07 0.5 2.72 0.76 2.59 2.84 0.5 mm, longitudinal 0.76 0.53 0.07 2.842.381.05 1.05 2.0 mm, mm, transverse transverse 1.28 1.54 2.72 1.43 1.43 3.14 2.59 1.35 1.35 0.76 2.0 1.28 1.54 3.14 2.0 mm, transverse 1.28 1.54 1.43 3.14 1.35 2.382.380.76 0.76 1.0 mm, transverse 0.23 0.11 3.84 3.95 2.15 0.23 2.56 1.0 mm, transverse 0.23 0.11 0.23 3.84 3.95 3.95 2.15 1.0 mm, transverse 0.11 3.84 2.15 0.230.232.56 2.56 0.5 mm, transverse 1.59 2.29 0.31 3.44 2.83 2.84 2.06 0.5 mm, transverse 1.59 2.29 1.59 0.31 3.44 3.44 2.83 0.5 mm, transverse 2.29 0.31 2.83 2.842.842.06 2.06 1 The relative errors are calculated using ˇ = V Vpp ,Exp denotes where V denotes V Vppˇˇ,Sim V Vpp ,Exp 100% 100% ,, where 1 The 1 relative errors are calculated using ˇ = Vpp , Exp The relative errors are calculated using δ= Vpp,Exppp ´ {Vpp,Exp the pp , Exp , ExpVpp,Sim pp ,Sim pp ,ˆ Exp100%, where Vpp,Exp denotes peak-to-peak voltage for different kinds of defects at different frequencies from experiments, V the pp,Sim the peak-to-peak peak-to-peak voltage voltage for for different different kinds kinds of of defects defects at different different frequencies frequencies from from experiments, experiments, the peak-to-peak voltage for different kinds of defects at differentatfrequencies from simulations.
V pp ,Sim the the peak-to-peak peak-to-peak voltage voltage for for different different kinds kinds of of defects defects at at different different frequencies frequencies from from V pp ,Sim
simulations. simulations.
Sensors 2016, 16, 952 Sensors 2016, 16, 952
8 of 9 8 of 9
All of of the 6 artificial defects were The inspection of the specimen was conducted at 0.9 MHz. All detected as shown in Figure 10, where the maximum noise amplitude is 0.012 V and the response longitudinal defect defect with with 0.5 0.5 mm mm depth depth is is 0.038 0.038 V. V. The defect signal amplitude is three amplitude of longitudinal times higher than the noise amplitude, which proves the feasibility of the proposed sensor. The noise mainly comes from electronic noises of the the RITEC RITEC RAM-5000 RAM-5000 measurement measurement system system and and lift-off lift-off effect. effect. The latter one is the dominant factor resulting from the processing error of the hollow axle inner be noticed, noticed, the the sensor sensor is ismore moresensitive sensitive to tothe thetransverse transversedefects. defects. surface and the sensor holder. As can be
Figure 10. Inspection result of the specimen. 1—2.0 mm, longitudinal defect; 2—1.0 mm, longitudinal Figure 10. Inspection result of the specimen. 1—2.0 mm, longitudinal defect; 2—1.0 mm, longitudinal defect; defect; 3—0.5 3—0.5 mm, mm, longitudinal longitudinal defect; defect; 4—0.5 4—0.5 mm, mm, transverse transverse defect; defect; 5—1.0 5—1.0 mm, mm, transverse transverse defect; defect; 6—2.0 transverse defect. defect. 6—2.0 mm, mm, transverse
5. Conclusions 5. Conclusions A flexible arrayed eddy current sensor has been presented. This new design is suitable for the A flexible arrayed eddy current sensor has been presented. This new design is suitable for the inspection of the hollow axle inner surfaces. The optimal excitation frequency, which is 0.9 MHz, is inspection of the hollow axle inner surfaces. The optimal excitation frequency, which is 0.9 MHz, is determined by FEM simulation, whose results are in good agreement with the experimental results. determined by FEM simulation, whose results are in good agreement with the experimental results. The maximum relative error between simulations and experiments is less than 4%. Results from The maximum relative error between simulations and experiments is less than 4%. Results from simulations and experiments show the sensor is capable of detecting both longitudinal and transverse simulations and experiments show the sensor is capable of detecting both longitudinal and transverse defects with depths as small as 0.5 mm. The sensor is more sensitive to the transverse defects, defects with depths as small as 0.5 mm. The sensor is more sensitive to the transverse defects, therefore therefore future work is required to increase the sensibility of the sensor to the longitudinal defects. future work is required to increase the sensibility of the sensor to the longitudinal defects. Author Contributions: Contributions: Z.S. Author Z.S. and and D.C. D.C. proposed proposed the the idea; idea; Z.S. Z.S. and and D.C. D.C. conceived conceived and and designed designed the the experiments; experiments; D.C. and C.Z. performed the experiments; D.C. also contributed to the writing and editing the manuscript. D.C. and C.Z. performed the experiments; D.C. also contributed to the writing and editing the manuscript. Z.S., Z.S., research andand were charged withwith the critical revisions of the of manuscript; all of the Q.C and and W.Z. W.Z.guided guidedthe the research were charged the critical revisions the manuscript; allauthors of the analyzedanalyzed the results. authors the results. Conflicts of Conflicts of Interest: Interest: The The authors authors declare declare no no conflict conflict of of interest. interest.
References References 1. 1.
2. 2. 3. 3. 4.
4. 5.
Marty, P.N.; Engl, G.; Krafft, S.; Spinelli, J.M. Latest development in the UT inspection of train wheels and Marty, P.N.; Engl, G.; Krafft, S.; Spinelli, J.M. Latest development in the UT inspection of train wheels and axles. In Proceedings of the 18th World Conference on Nondestructive Testing, Durban, South Africa, 16–20 axles. In Proceedings of the 18th World Conference on Nondestructive Testing, Durban, South Africa, April 2012. 16–20 April 2012. Carboni, M.; Cantini, S. Advanced ultrasonic “Probability of Detection” curves for designing in-service Carboni, M.; Cantini, S. Advanced ultrasonic “Probability of Detection” curves for designing in-service inspection intervals. Int. J. Fatigue 2016, 86, 77–87. [CrossRef] inspection intervals. Int. J. Fatigue 2016, 86, 77–87. Rudlin, J.; Raude, A.; Völz, U.; Conte, A.L. New methods of rail axle inspection and assessment. In Proceedings Rudlin, J.; Raude, A.; Völz, U.; Conte, A.L. New methods of rail axle inspection and assessment. of the 18th World Conference on Nondestructive Testing, Durban, South Africa, 16–20 April 2012. In Proceedings of the 18th World Conference on Nondestructive Testing, Durban, South Africa, 16–20 Kappes, W.; Hentschel, D.; Oelschlägel, T. Potential improvements of the presently applied in-service April 2012. inspection of wheelset axles. Int. J. Fatigue 2016, 86, 64–76. [CrossRef] Kappes, W.; Hentschel, D.; Oelschlägel, T. Potential improvements of the presently applied in-service inspection of wheelset axles. Int. J. Fatigue 2016, 86, 64–76. Li, F.; Sun, X.; Qiu, J.; Zhou, L.; Li, H.; Meng, G. Guided wave propagation in high-speed train axle and damage detection based on wave mode conversion. Struct. Control. Health Monit. 2015, 22, 1133–1147.
Sensors 2016, 16, 952
5.
6.
7.
8. 9.
10. 11. 12.
13. 14.
15.
16. 17.
9 of 9
Li, F.; Sun, X.; Qiu, J.; Zhou, L.; Li, H.; Meng, G. Guided wave propagation in high-speed train axle and damage detection based on wave mode conversion. Struct. Control. Health Monit. 2015, 22, 1133–1147. [CrossRef] Ziaja, A.; Cheng, L.; Radecki, R.; Packo, P.; Staszewski, W. Cylindrical guided wave approach for damage detection in hollow train axles. In Proceedings of the 10th International Workshop on Structural Health Monitoring, Palo Alto, CA, USA, 1–3 September 2015; pp. 2054–2061. Lin, J.; Lin, F.; Lin, C.; Zhang, K.; Yu, X. Hollow shaft array electromagnetic testing system based on integrated nondestructive testing techniques. In Electromagnetic Nondestructive Evaluation (XVIII); IOS Press: Washington, DC, USA, 2015; Volume 40, pp. 296–302. Cavuto, A.; Martarelli, M.; Pandarese, G.; Revel, G.M.; Tomasini, E.P. Experimental investigation by laser ultrasonics for high speed train axle diagnostics. Ultrasonics 2015, 55, 48–57. [CrossRef] [PubMed] Chady, T.; Psuj, G.; Kowalczyk, J.; Spychalski, I. Electromagnetic system for nondestructive evaluation of train hollow axles. In Proceedings of the 2013 Far East Forum on Nondestructive Evaluation/Testing: New Technology & Application, Szczecin, Poland, 17–20 June 2013. Bureau, J.; Ward, R.C.; Jenstead, W. Advances in eddy current array sensor technology. In Proceedings of the 18th World Conference on Nondestructive Testing, Shanghai, China, 25–28 October 2008. Lepage, B. Development of a flexible cross-wound eddy current array probe. Int. J. Appl. Electromagn. Mech. 2014, 45, 633–638. Marchand, B.; Decitre, J.M.; Sergeeva-Chollet, N.; Skarlatos, A. Development of flexible array eddy current probes for complex geometries and inspection of magnetic parts using magnetic sensors. In Proceedings of the Conference on Review of Progress in Quantitative Nondestructive Evaluation, Denver, CO, USA, 15–20 July 2012. Crouch, A.; Goyen, T.; Porter, P. New method uses conformable array to map external pipeline corrosion. Oil Gas. J. 2004, 102, 55–59. Sheiretov, Y.; Evans, L.; Schlicker, D.; Zilberstein, V.; Goldfine, N.; Sikorski, R. TBC characterization using magnetic and electric field sensors. In Proceedings of the ASME Turbo Expo 2007: Power for Land, Sea, and Air, Montreal, QC, Canada, 14–17 May 2007; pp. 97–103. Endo, H.; Nishimizu, A.; Tooma, M.; Ouchi, H.; Yoshida, I.; Nonaka, Y.; Otani, K. Signal evaluation system of flexible array ECT probes for inspecting complexly shaped surfaces. In Proceedings of the Conference on Review of Progress in Quantitative Nondestructive Evaluation, San Diego, CA, USA, 18–23 July 2010. Xie, R.; Chen, D.; Pan, M.; Tian, W.; Wu, X.; Zhou, W.; Tang, Y. Fatigue crack length sizing using a novel flexible eddy current sensor array. Sensors 2015, 15, 32138–32151. [CrossRef] [PubMed] García-Martín, J.; Gómez-Gil, J.; Vázquez-Sánchez, E. Non-destructive techniques based on eddy current testing. Sensors 2011, 11, 2525–2565. [CrossRef] [PubMed] © 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC-BY) license (http://creativecommons.org/licenses/by/4.0/).
