Keywords

1 Introduction

Carbon composites are high-strength, low-weight materials which find increasing use in the aviation and automobile industry. Their mechanical properties can exceed those of metallic materials while being much lighter. However, even small defects, which might be within the material and invisible from the outside, can weaken composite parts significantly. Therefore, there is a need for reliable inspection methods that can detect defects within the material without destroying the specimen. Eddy current sensors have been proven to be able to detect various defects in composites. This work proposes an integration of such a sensor in an inspection system based on a 6-axis robotic arm. This setup allows for performing two different inspection methods: the polar method, where the eddy current probe inspects a single position while rotating, and the scanning method, where the sensor scans the whole specimen. The research question is: Can the polar and the scanning eddy current inspection methods be combined in a single system to detect the most common defects in dry fibre layups? Furthermore, the challenges during the implementation are discussed in this paper.

2 Literature Review

2.1 Carbon Composite Manufacturing

Production processes for composite materials can be divided into two categories: prepreg (short for pre-impregnated) methods, where the fibres have been mixed with the resin before they have been brought into the final shape, and dry fibre processes, where the resin is added at the end of the process in an infusion step. Prepreg processes, such as Automated Fibre Placement, require expensive material handling under freezing temperatures. Dry fibre materials can be stored and processed at room temperature [13]. However, even small defects in the dry fibre layups can have significant effects on the final part’s properties. Some of the most common defects in dry fibre processes and their influence on the final material’s properties have been investigated in the literature. These are: Misaligned fibres [12], gaps in between the fibres [5], and foreign objects [10].

2.2 Eddy Current Sensors

On the left-hand side, Fig. 1 illustrates the basic principle of an eddy current sensor on a conductive specimen. An emitting coil creates an alternating primary magnetic field, which penetrates the specimen. According to Faraday’s law of inductivity, a changing magnetic field induces an eddy current in the specimen. This creates a secondary magnetic field which can be picked up by either a second coil or by measuring the change of impedance in the emitting coil. The sensor’s response is a signal in the complex plane with the coil’s resistance on one axis and the inductive reactance on the other.

The result depends on different aspects - two of the most sensitive are the lift-off (distance between the probe and the specimen) and the specimen’s electrical resistance. Figure 2 shows the loci of the results when these two variables are changed. The exact shape of the loci depends on the used probe and the material of the specimen [7]. In addition, tilting of the probe influences the results significantly [4].

Fig. 1.
figure 1

Principle of the polar measuring method of uni-directional carbon fibres with a half-transmission Eddy current probe

Full size image
Fig. 2.
figure 2

Eddy current sensor responses for specimen with different conductivity \(\sigma \) and different lift-offs (adapted from [7]). The shape of the loci depend on the material of the specimen and the specific probe.

Full size image

2.3 Measuring Methods

Polar Method

The polar measurement is a method to determine the local fibre orientation at a specific location. It makes use of the elliptical shape of the eddy current, which is caused by the non-isotropic electrical properties of carbon fibre material. This aspect can be measured by a probe with spatially separated transmission and receiving coils, such as the half-transmission design. While this end-effector is rotated over the point of interest, the measured current in the receiving coil changes in correspondence to the shape of the eddy current. The principle is illustrated in Fig. 1 on the left. The complex response in a 2D plane first needs to be reduced to a 1D signal and then plotted against the rotation angle of the probe (illustrated in Fig. 1 on the right-hand side). The peaks of the lobes in this graph align with the fibre orientation [3]. The signal is symmetric. Therefore, it is sufficient to evaluate only 180\(^{\circ }\) of the whole rotation [8].

The exact response of the coil is highly dependent on the exact probe design and the material of the specimen. Therefore, there are different approaches to reducing the complex signal to a 1D signal. Lange and Mook assessed the magnitude of the 2D signal [8]. Another method is to apply a vector rotation of the complex signal around the origin. Subsequently, the x-value of the data can be evaluated [9].

Using the polar method, Lange and Mook determined the fibre orientation of a single ply with an accuracy of less than 1\(^{\circ }\). However, they found that the results are less accurate for two plies orientated at a small angle relative to each other. Therefore, they proposed a reverse engineering approach based on a synthesis of two signals followed by a correlation. This approach showed an improved accuracy [8].

All the systems used in the literature to perform a polar measurement had the probe fixed to one position - either in a tool or as a rig set-up that always measures the same spot. For the inspection of complex 3D parts, a system is needed that is able to approach different positions for inspection.

Scanning Method

While the polar method can be used to inspect a single spot, the scanning method is able to provide a global view of the whole specimen. The analysis of this image can reveal various defects, such as the fibre angle deviations, gaps, overlaps, or foreign objects [11]. For this, the probe is guided along a grid that covers the specimen’s entire surface. Along the path, at certain intervals, the sensor’s response is measured. These data points are again reduced to a 1D signal and mapped on a grey scale. The lateral information and the grey scale value can be combined into a grey scale image of the entire specimen.

The resolution of the image depends on the size of the grid. The smaller the grid cells, the higher the resolution of the image. For the evaluation of the fibre angle, the texture of the textile must be visible. Even with a very small grid size, the measurement will not resolve the individual fibres but rather a macroscopic pattern of the fibre ply [11]. A woven material has a very distinct pattern. In non-crimp fabrics, using a very small grid size, the stitching pattern can be visualised. Unidirectional tapes, which are placed without gaps, lack a regular pattern and appear as a homogeneous surface in the grey-scale images. For this reason, Bühlow states that this method is unusable for unidirectional tapes [2].

With a visible pattern, image processing algorithms can be applied to the image to detect the fibre orientation. Bühlow determined it with a precision of 0.1\(^{\circ }\) [2]. Heuer et al. demonstrated how to identify defects, such as foreign objects, in-plane, and out-of-plane waviness in cured composite parts. Copper injections were detected under up to 12 layers. The detection depth of the tape, polyethylene foil, and paper was 9 layers [6]. Mook et al. scanned a specimen with changing fibre orientation in a stair-like manner and were able to illustrate the discontinuities in the results [9]. Gaps in dry fibre layups have not been investigated with an eddy current sensor.

3 Method

3.1 Hardware Selection

In order to realise a system which is capable of performing both measuring methods, the eddy current probe is attached to a 6-axis robot arm. CIKONI GmbH supplied the system. This allows not only the necessary movements for the measurements of a flat specimen but can also follow the surface of a draped, 3D specimen. A similar set-up has been demonstrated by Bardl et al. who investigated a woven material while draping it into the third dimension. They, however, only applied the scanning method [1].

The eddy current sensor system is based on Suragus’ EddyCus Sensor Kit. It can be equipped with different probes. Bühlow found that differential probes are not suitable for carbon fibre inspection [2]. For the scanning method, the absolute probe design (one coil design - response is measured as a change in impedance in the excitation coil) is chosen because it is the smallest possible design, improving the resolution of the measurement. For the polar method, a half-transmission probe (separate excitation and receiving coils) is used because a spatial separation between the transmission and receiving coil is needed.

As mentioned above, the lift-off is a relevant factor during the measurement and should be kept constant. For the polar method, a constant distance can be achieved by rotating the probe perpendicularly to the surface. A perfect normal alignment is also important to reduce tilt effects. During the scanning method, a constant lift-off is harder to maintain, as dry fibre layups do not have a uniform surface but deviate in height. This out-of-plane waviness leads to bright spots on the image where the fibres were closer to the probe. These effects dominate the results and make it impossible to evaluate the other areas. To tackle this issue, a non-conductive, incompressible foil of constant thickness was placed on the fibres. The robot was programmed to put the sensor in light contact with the additional layer - this way, a constant distance was maintained during the measurement.

3.2 Specimen and Experimental Plan

For all experiments, a unidirectional non-crimp fabric out of dry carbon fibres, with an areal weight of 80g/m2 is used. On the left-hand side, Fig. 3 shows the back of the fabric with a check-pattern of scrim fibres that keeps the carbon fibres in place. The fabric was produced by epo GmbH.

A separate piece of this fabric is prepared for gap examinations. Gaps of different sizes are cut into the fabric. In addition, some fibres are displaced to create a different kind of gap. Figure 3 shows this sample on the right-hand side.

For the scanning method, two setups are measured. For the first setup, the fabric with the induced gaps is covered by an defect-free layer and scanned. For the other specimen, different foreign objects of different sizes (1–5 mm) are placed under one defect-free layer of the fabric. This specimen is scanned to determine whether the particles can be detected.

For the polar measuring method, a device that rotates the fabric is built. It allows to set different fibre orientations with a tolerance of <0.5\(^{\circ }\). To determine the accuracy and precision of the eddy current measurement, two layers were stacked with an angle of 90\(^{\circ }\) and placed in the rotation device. It is set to 11 different angles between 0 and 90\(^{\circ }\). For each angle, 3 measurements are conducted.

Fig. 3.
figure 3

Left: The non-crimp fabric material which is used for the specimen. Right: Specimen with introduced gaps.

Full size image

4 Results and Discussion

4.1 Fibre Orientation

As an example of the polar measurement, Fig. 4 (left) illustrates the results of the 3 repetitions where the fixture was set to 45\(^{\circ }\). The 3 polar plots show a good agreement, and the red lines that indicate the maxima of the graph (which correlate with the detected fibre orientations) are well aligned. In the entire experiment, the angle difference between the two layers was measured to an accuracy of 0.9\(^{\circ }\). It should be mentioned that the specimen was handled manually. Therefore, the accuracy might be limited by slight deviations introduced by the operator. To eliminate this factor, the standard deviation of all measurements was calculated to be 0.16\(^{\circ }\). This is well below the precision of the fixture.

Fig. 4.
figure 4

Results of the eddy current measurement. Left: Result of the polar measurement. Middle: Photo of the specimen before it is covered with an additional ply. Right: Scan of the same specimen.

Full size image

4.2 Gap Detection

The results of the scanning measurement with the gap specimen are illustrated in Fig. 4 on the right-hand side. The image of the specimen before it is covered with an additional ply is compared with the scan result. The gaps are clearly visible as brighter spots on the image. Even the small gaps, which are just displaced fibres, show up on the measurement. The vertical lines in the image are along the scanning direction of the sensor. The used eddy current instrument tends to drift over time, even if it is not moved. This effect is compensated by an automatic drift compensation which is applied frequently. However, the drift can still be seen in the image. The lines which run diagonally from the top left corner to the bottom right are the fibres of the layer covering the defective layer.

4.3 Foreign Objects

In Fig. 5, the fibre orientation of the covering layer and the scanning direction are both horizontal. The three white spots on the grey scale image are metal pieces of different sizes that have been induced. Even the small piece of 1\(\,\times \,\)1mm2 can be identified. The signals are clear enough to be easily thresholded for image segmentation. Objects made out of plastic do not show up in the scan. This result is reasonable, as they are not conductive and can therefore not create an eddy current.

Fig. 5.
figure 5

Results of the eddy current measurement of a specimen with different inclusions. Left: grey scale image. Right: Segmented result with a threshold applied to the gery scale. (Color figure online)

Full size image

4.4 Discussion

The set-up of an eddy-current sensor attached to a 6-axis robotic arm allows for the execution of both measuring methods - the polar and the scanning method. The main challenge is to position the sensor precisely. For the polar method, the probe must be kept perpendicular to the specimen’s surface. Otherwise, tilting effects might falsify the results. Therefore, polar measurements on a flat specimen can easily be performed. However, for 3D measurements the surface must be known in detail in advance to calculate the surface normals. For the scanning method, it is very critical to keep a constant distance between the probe and the fibres. The surface of dry fibre layups tends to be uneven, which leads to disturbing spots on the scan due to the lift-off effect. For this investigation, the issue was solved by flattening the fabrics with a non-conductive foil in between the probe and the fibres. However, this procedure turns this originally non-contact inspection into a contact method.

The data captured by the rotating method can be evaluated by finding the maximum of the response. A specimen with fibres in 0 and 90\(^{\circ }\) orientation can be measured with a precision of 0.16\(^{\circ }\). Smaller angle differences between the layers require a more complex evaluation, as described in the literature [8].

The scanning method can detect defects like gaps and conductive foreign objects. Pieces from non-conductive materials, such as plastic foils, cannot be seen on the scans. Heuer et al. showed that even non-conductive insertions in cured carbon composite parts can be visualised [6]. The reason why these object showed up in infused parts, but not in dry layups is that during infusion the fibres drape tightly around the insertion. Therefore, the eddy current path in the fibres is influenced by the object which is detectable in the scan. On the other hand, the fibres of dry layups rest loosely on thin objects. This does not change the eddy current path in a detectable manner.

The research question can be answered positively. However, different probes are required for the two methods, and the limitations, which are mentioned above, must be addressed.

5 Conclusion

This research demonstrated that the most common defects in a dry fibre layup (fibre angle deviations, gaps and foreign objects) can be detected with a single eddy current inspection system. To do so, a 6-axis robot was programmed to position the eddy current sensor. Two measuring methods were investigated: the polar and the scanning method. The former is capable of detecting fibre angles with a precision of 0.16\(^{\circ }\). This was proven for a 2 ply layup with perpendicular fibre orientation. For this method, a probe with a half-transmission set-up was used. The scanning approach, using an absolute probe design, proved to be useful for inspecting gaps and foreign objects. Metal insertions and gaps were easily detected, while non-conductive objects could not be identified.

The main challenge was to position the probe, as it needs to be perpendicular and within a small, constant distance to the fibres. Dry fibre layups tend to have an inhomogeneous surface which makes a contact-less scanning inspection very difficult. In the future, this issue could be solved by either a spring-loaded mechanism that keeps the probe at a constant distance from the fibres or a prior laser scan inspection to determine the exact surface of the specimen.

Furthermore, the reliability of these measuring methods should be investigated for different materials and multiple layers.