Fig. 1: Automotive valve made out of silicon nitride
Fig. 2: Left side : C-scan of the head region of a Si3N4-ceramic valve showing a "natural" defect in the volume (arrow), image size 28 mm x 28 mm. Right side : C-scan of the head region of a test valve containing test defects with about 100 µm diameter (image size 32 mm x 32 mm)
Volume testing of the shaft region is a challenge for ultrasound testing, as refraction of sound at the interfaces is very strong and anisotropic because of the high speed of sound (10.5 mm/µs longitudinal) and the low radii of the shaft. Up to now, only larger defects could be detected in the shaft.
Another application of high-frequency ultrasound is the detection of surface cracks by Rayleigh wave techniques. Using the same probes as for volume testing, the sample is insonified under the critical angle for excitation of a leaky Rayleigh wave. In spite of strong damping, the wave can be traced over a distance of a few mm. This technique is particularly applied to the valve seat surface (Fig. 3). A B-scan with the rotation angle around the long axis of the valve as the spatial coordinate is performed to test this critically loaded area of the valve. In the same way, the shaft surface can be scanned for surface defects. In contrast to dye penetrant testing, cracks not open to the surface can be detected.
Fig. 3: Left: Scheme of leaky Rayleigh wave testing of the valve seat. Right: Test result obtained on a valve. The rotation angle (360°) is the horizontal coordinate, ultrasound time-of-flight is the vertical coordinate. The surface defects appear as nearly vertical lines.
A first technique, which is used to inspect the valve head region, is digital radiography. The X-ray cone-beam, which is subject to different attenuation is first converted into visible light, then video-imaged and digitized. The raw image is evaluated by digital image processing. A typical image resolution is 512 x 512 pixels with a pitch of 10 µm. By subtraction and edge enhancement algorithms small defects are emphasized. In Fig. 4, an image with defects of 200 µm size in the head region of the test valves mentioned above is shown. A complete inspection of the plate requires to measure many of such images in an automated raster scan.
Other X-ray techniques are the 2D- and 3D-computed tomography (CT). They are particularly used for the shaft and the transition region. They also work with pixel sizes of 10 µm. A line detector or a detector array are employed /7/. Due to the large number of projections to be collected, the time consumption of these techniques is relatively high. Figs. 5 and 6 show reconstruction results from the shaft region and the transition region of valves. Other than for digital radiography, the defect position can be reconstructed in three spatial coordinates.
|Fig. 4: Digital radiography showing two test defects in the valve head. Image size 5 mm x 5 mm||Fig. 5:2D-CT image of a defect in the valve shaft (10 µm pixel resolution in four planes with a spacing of 20 µm)||Fig. 6: Visualization of a pore detected by 3D-CT in the transition region between valve head and shaft. .|
Compared with the other techniques described here, a good characterization of the defect is possible. Pores and metallic inclusions (the latter are often surrounded by reaction zones in ceramics) can be well distinguished. Density variations can be detected as well /8/.
Acoustic resonant testing is a well known technique for fast quality control of components and has recently come again into the focus of interest /9,10/. Its basic principle is to induce a vibration of the component in one or several of its eigenmodes by using pulsed or periodic mechanical excitation. The resulting oscillation is analyzed. Defects in the component become visible as a change of the resonant spectrum (position of resonance lines, width and amplitude of the lines, line splitting for components with symmetries).
In the setup used for the present experiments, the valve is supported by piezoelectric transducers at two points (Fig. 7). One transducer is operated as transmitter, the other as receiver. Two approaches were followed for excitation and detection. In the first, the transmitter is excited by a sine wave and the acoustic signal transmitted through the valve is detected using a lock-in amplifier. The frequency of the sine wave is slowly sweeped, allowing to record amplitude and phase of the spectrum over a wide frequency range.
The second approach uses an excitation by a step-function. The resulting resonant signal is recorded over the full decay time using a transient digitizer with large storage depth. The signal is averaged over many repetitions of the experiment and then Fourier transformed over 50000 or more points (Fig. 8).
As expected, the result of both approaches was equivalent, but the time-domain technique turned out to be faster and more flexible. The experiments performed up to now suffered from the poor reproducibility of the mechanical coupling of the vibration to and from the component. In particular the low-frequency bending modes showed significant variations in the resonant frequencies.
Fig. 7: Setup used for resonant testing of ceramic automotive valves.
Fig. 8: Typical result from resonant testing of a valve (top: decay curve over 20 ms, bottom: resonance spectrum from 0 to 100 kHz, logarithmic amplitude scale).
The resonance lines of the valves inspected exhibit Q-values in the range of 1000 to 3000. Therefore, a high detection sensitivity can be expected. On the other hand, it is questionable, if defect sizes of 50 to 100 µm are detectable by this technique. Finite element calculations to study the influence of cracks in disk-shaped samples show, that crack lengths of about 10 % of the radius are required to cause relative frequency changes in the order of 10-4 /11/. Such changes have to be detected with the background of tolerable variations of density, shape and mass already present in the production lot. Experimentally, by weighing a lot of ceramic valves a standard deviation of the mass of 37.4 mg was determined. If this mass is concentrated in a single point of the valve, it is represented by a pore of 2.8 mm in diameter, when one uses the density of silicon nitride.
The optimum technique, which would be able to detect smallest defects in all regions of the sample in a fast way and with highest sensitivity, does not exist. In some cases the numbers for detection limits are still unproved due to lack of suitable test defects.
The present investigations performed in the framework of research programs were aiming at an utmost complete testing for volume and surface defects. Such testing will not be feasible for series components due to economic limitations. From the viewpoint of NDT, an improved specification of the defect size to be detected as a function of the position in the component is necessary. This requires furthermore a close cooperation of developers in construction, production and application.
|Table 1: Summary of main results obtained for ceramic valve testing|
|ultrasound volume testing||ultrasound surface wave testing||resonant testing||X-ray digital radiography||X-ray computed tomography||dye penetrant testing|
|testing time||4 min||4 min||0,5 min||30 min||60 min||10 min|
|detection limit (valve head)||50-100 µm||20 µm||unknown||100 µm||150 µm||50 µm (open crack)|
|valve seat (surface)||-||30 µm (?)||unknown||100 µm||150 µm||50 µm (open crack)|
|detection limit shaft||100-200 µm (?)||cracks 1 mm (?)||100 µm||50 µm||50 µm (open crack)|
|defect characterization||moderate||moderate||poor||good||very good||moderate|
|| UTonline ||