Abstract
This paper presents an eddy current (EC) probe with orthogonal excitation coils that can change the exciting current direction electrically and tunnel magnetoresistive (TMR) sensor for inspection of stacked CFRP. As CFRP is electrically anisotropic, eddy current excited along the fiber direction is stronger than in other directions. A high sensitivity TMR sensor located at the center of the two orthogonal coils measures the z component of induced magnetic field (Bz) which constitutes the probe signal. The output of the sensor is 0 if the induced currents are symmetric about the sensor. Signals of defects in different orientations depend on the direction of the exciting current. The operation of the probe is validated using a finite element model. A prototype was built and experimental results from CFRP sample with defects in different orientations are presented and analyzed.
Keywords
Introduction
Carbon fiber reinforced polymers (CFRP) are used widely because of their excellent mechanical properties. The use of CFRP also presents major challenges for reliable and accurate evaluation of their structural integrity. Several types of anomalies can occur during the manufacturing and service and thereby degrade the integrity and safety of the component. Some of these defects are disbonds, delaminations, fiber breaks, fiber waviness, missing plies, etc. Traditional nondestructive evaluation (NDE) techniques, such as ultrasonic techniques need couplant to inject energy into the sample, making the inspection slow and complex [1,2]. In contrast eddy current techniques make non contact measurements allowing rapid inspection of large parts. Eddy current inspection systems have shown promising results for CFRP. A number of defect modes such as fiber breakage and waviness have been detected [3–5]. The feasibility of these sensor types were also demonstrated on glass fiber reinforced polymers (GFRP). However, sensitivity of these probes is rather low and an optimized design of the sensor together with robust signal conditioning are crucial for good performance.
Spintronic sensors based on the tunnel magnetoresistive (TMR) effect have been widely explored over recent years for many applications from biomedical to industrial devices [6,7]. As the TMR sensors can be integrated on CMOS wafers which include the readout electronics, the MR technology can be made very compact and versatile. Reducing the sensor area increases the resolution of imaging that change over small distances. Some other figures of merit are the low power requirements, reduced cost and high magnetic field sensitivity at room temperature. MR sensors are thereby suitable for array applications with fine spatial resolution and high sensitivity when compared to coils, fluxgates and hall sensors.
Probe design
CFRP is generally made of several thin unidirectional plies stacked with different fiber orientations. CFRP structures are therefore anisotropic in that the conductivities in different directions are different. Assuming that CFRP fibers are parallel to the x-y plane, the conductivity tensor of CFRP is given by equation (1), [8].
When the excitation current is along the fiber direction, the induced current is stronger along the fiber than in other directions and so will the signal generated by a defect in the fiber. In order to vary the excitation current direction electrically, an excitation method using orthogonal coils as shown in Fig. 1(a) is used in this work. Two planar coils that are perpendicular to each other and excited by current sources that are in phase but with different amplitude generates eddy currents in any desired direction, as shown in Fig. 1(b). A high sensitivity TMR sensor, located at the center of the two orthogonal coils measures the z component of induced magnetic field (B z ) to pick up defect signal.
(a) Schematic of EC probe with orthogonal coils and TMR sensor located at the center of the coils. (b) Orthogonal excitation coils to generate currents along controlled directions.
A 3D finite element model (FEM) is used to validate the principle of the proposed approach for detecting defects in a multi-layer CFRP sample. Commercial FEM computation software COMSOL is used to model the geometry. The material conductivity is [σ L σ T σ cp ] = [2 × 104,100,40] S/m in the simulation. The conductivity tensor is calculated according equation (1) in the model.
Induced eddy current
First, the induced eddy current in CFRP by a single line current and by a single turn circular coil (diameter 3 mm) are simulated. A 6 layer CFRP is considered where each layer has a thickness of 0.5 mm. The orientations of the layers are [0°∕90°]3. The excitation current frequency is 100 kHz, current amplitude is 1 A and current lift-off from the top surface of CFRP is 0.5 mm. The simulation results are presented in Fig. 2, where the red arrows indicate the current direction. It is seen that the maximum eddy current density induced by the linear excitation along the fiber direction is about 3 times stronger than the maximum eddy current density induced by the circular coil.
Simulation result of induced eddy current in CFRP (a) y = 0 plane, linear excitation; (b) z = 0 plane, linear excitation; (c) y = 0 plane, circular single turn coil excitation; (d) z = 0 plane, circular single turn coil excitation.
Next, defect in single fiber layer is simulated in the 6 layer CFRP described in the previous section. A cuboid defect (dimensions 3 × 3 × 0.5 mm) is introduced in the second layer from top. The excitation coil is approximated by an infinite current sheet which produces a uniform magnetic field. The error due to this approximation is small when compared to that of a multiline linear coil, while eliminating the need for scanning. A C-scan image is thereby obtained from a single FEM calculation. The excitation current frequency is 100 kHz and current density is 1000 A/m. Two cases, namely, excitation current along x- and y- directions are studied. The simulation results of defect located in the 2nd layer are presented in Fig. 3, where the real component of B z 0.5 mm above the top surface of the CFRP sample is plotted. Figure 3 shows results with x- and y- direction excitation current. As the defect is in the 2nd layer where the fibers are along y- direction, the defect signal is much stronger when the excitation current is along y- direction than in x- direction. Similarly, when the defect is located in the 3rd layer, signal of the defect is much stronger when the excitation current is in x- direction than in y- direction, as shown in Fig. 3. This property can be used to determine which ply has defect if the defect appears only in single ply.
Simulation result of real component of B z with excitation current applied along x- and y axis respectively with (a) defect in the 2nd layer that fibers along y- direction, (b) defect located in the 3rd layer that fibers along x- direction.
Figure 4 presents the defect signals at different excitation frequencies. It is seen that for defect in the 2nd layer, the amplitude of the signal increases as the excitation frequency increases in the frequency range from 100 kHz to 5 MHz. However, for defect located in the 4th layer, the signal increases first than decreases as frequency increase. So for defects embedded in different depths, proper frequency should be used to achieve best performance. Multi-frequency measurements can be used to determine depth locations of defects in CFRP samples.
Simulation result: line plot of real component of B z vs. x along y = 0, z = 0.5 mm line with defect located in the 2nd layer (a) and 4th layer (b). (c) Shows the peak value vs. frequency for defect located in the 2nd layer (n = 2) and 4th layer (n = 4).
A second simulation model was setup to correspond to the experimental investigation with the prototype probe described above. Signal response due to a rectangular notch through different numbers of layers is simulated. The model contains a 16 ply CFRP with [0∕90]8 layup sequence following the same conductivity properties as described previously in equation (1). The thickness of each ply is 0.2 mm giving a total sample thickness of 3.2 mm. Three defects which are all rectangular notches of dimension 10 mm length along x, 1 mm width along y and depths of 4, 8 and 12 plies, are located on the back side of the sample. The model uses an infinite current sheet as excitation source with amplitude 1000 A/m and frequency 100 kHz. The sensor measuring the z-component of the magnetic field is located 3.2 mm above the sample surface. The simulation results are as shown in Fig. 5. Excitation current along x- and y- directions are studied, firstly for different notch depth, and secondly for different defect orientations.
Figure 5 shows the effect of notch depth with excitation current flowing parallel and perpendicular to the defect. The defect is 10 mm along the x-axis and only 1 mm along y. The response due to currents flowing along the y-axis is therefore dominant. Figure 5 shows correlation between the defect orientation and the directions of the applied excitation current as well as to the size of the defect.
Signals due to a notch which is 12 ply (2.4 mm) deep and has different orientations relative to the x-axis are presented in Fig. 6. It is clear that the signal response is influenced by the defect orientation relative to excitation current direction.
Experiment validation
A 3.2 mm thick CFRP sample was produced with 16 woven 0/90 plies. A number of machined rectangular notches and flat bottom holes were manufactured in the sample. The angle of the rectangular notches are varied and so also the depth of notches and holes. The top view of the sample is presented in Fig. 7(a). Properties of all defects are given in Table 1.
Imaginary part of B z 3.2 mm above the sample due to a 10 × 1 mm notch; top row - current along the defect (Jx), bottom row - current perpendicular to defect (Jy).
Imaginary part of B z 3.2 mm above the sample due to a 10 × 1 mm defected with different orientation relative x-axis, top row with current along the defect (Jx), bottom row current perpendicular to defect (Jy).
(a) CFRP sample with machined defects. (b) Experiment setup.
A prototype probe was built with two coils and a TMR sensor placed in the center between the coils close to the sample surface. The lift-off distance for the TMR sensor is 3.2 mm and the sensing axis is normal to the sample surface. The coils are wound so that linear currents along the x and y axis can be produced in proximity to the TMR sensor. The sensor is placed in the symmetry point of the two coils in order to reduce background fields from excitation sources. A sinusoidal current of frequency 100 kHz is excited using an Agilent 35500B function generator with applied peak to peak amplitude of 10 V. The output of the sensor is connected to a lock-in amplifier model 844 from Stanford Research Systems. The experimental setup is presented in Fig. 7(b).
Properties of manufactured defects
The probe is placed directly on the sample surface using a spring loaded fixture to ensure robust scanning with minimal lift-off variations. A scanning speed of 10 mm/s used and the distance between adjacent scanning lines is 1 mm. In-phase and quadrature components from the lock-in amplifier are collected with a sampling frequency of 1000 data points/s.
Figure 8 shows the experiment results with the excitation current along x- and y- direction separately. It is seen that the defect A1 has strong signal in Fig. 8(a) but it has little indication in Fig. 8(b). This is due to the fact that the defect A1 is oriented along y direction that has a little impact on the eddy current in y- direction. On the other hand, when the induced eddy current is along x- direction, it is disturbed by the A1 defect resulting in a strong signal. Similarly, the defects B1, B2 and B3 have much stronger signal in (b) than in (a). The flat bottom holes, which are not orientated in a specified direction, have comparable signals in (a) and (b).
To further quantitatively analyze the effect of the current direction, the experiment results are rotated as
Experiment result with excitation current along x-direction (a) and y-direction (b).
Signal energy E vs. rotating angle.
EC probe with TMR sensor and orthogonal linear excitation, which can induce eddy current in a chosen direction without rotating the probe mechanically, was presented for detecting flaws in CFRP laminates. Due to conductivity anisotropy of CFRP, it is possible to detect defects in different layers by exciting current along different directions. Simulation results show that linear excitation along fiber induces higher eddy current density than a circular coil in the case of CFRP laminates. A high sensitivity TMR sensor is located at the center of the two orthogonal coils to measure the z component of induced magnetic field (Bz). TMR sensor provides high sensitivity, fine spatial resolution and wide bandwidth. It was demonstrated that the signals of defects in different orientations depend on the direction of exciting current. A prototype probe was built and a CFRP sample with different kinds of defects was tested in lab. Experimental results are presented in this paper showing the feasibility of this novel probe design. However, a significant amount of development and testing work remains to be done, e.g. optimize coil design for different flaw mechanisms in CFRP samples.