EWGAE 2010 Vienna, 8th to 10th September

Transcription

1 EWGAE 21 Vienna, 8th to 1th September Validation of AE-signals recorded with conventional Equipment using 3D-Scanning-Laser-Vibrometer Albert ALBERS 1, Jochen SCHELL 2, Markus DICKERHOF 1, Benjamin HESSENAUER 1 1 IPEK - Institute of Product Engineering, Karlsruhe Institute of Technology, D Karlsruhe, Germany; albers@ipek.uka.de, dickerhof@ipek.uka.de, hessenauer@ipek.uka.de, tel. +49 () Polytec GmbH, Polytec-Platz 1-7, D Waldbronn, Germany, tel. +49 () Keywords:, frequency analysis, modal analysis, simulation Abstract The work presented in this paper focuses on interpretation and explanation of acoustic emissions (AE) recorded by means of s. It is a well known fact that AE signals are changing on their way from the source to the sensor. Especially the spectral contents of the recorded signal and of the source-signal are generally not the same. Among the main influencing factors are the mechanical characteristics of the solids the sound waves travel through, of the sensor coupling and of the sensor itself as a somewhat resonant system. For a better understanding of the impact of these factors on a recorded signal a Polytec threedimensional scanning for non-contact and non-reactive acquisition of the vibration and surface wave data has been used on a geometrically simple structure (steel cuboid) in combination with conventional AE measurement equipment (i.e. a ). The simple geometry of the chosen structure also allows for a relatively reliable numerical calculation of its natural frequencies and the associated eigenmodes at least in the lower frequency spectrum by means of a modal analysis using finite element methods. Introduction Online monitoring of critical system components such as bearings is crucial to minimize the probability of unnoticed degradation leading to disastrous breakdown and probably severe subsequent damages. Additionally, if knowledge e. g. about the actual wear condition of the system can be derived, the opportunity of condition-based maintenance arises and offers further considerable economical advantages. Previous own works (e.g. [1, 2]) as well as numerous other sources (e.g. [3-8]) document the applicability of methods based on ultrasonic structure-born sound (acoustic emissions) for continous monitoring of both rolling and plain bearings. In contrast to the acoustic emissions caused by internal sources (dislocations, grain boundary sliding, phase transformation, crack formation and growth, plastic deformation or fracture) [9] which are of primary interest in the field of strucural health monitoring, here external sources of acoustic emissions (especially impact and friction) [9] are of particular relevance: in rolling element bearings one of the most typical damage symptoms is pitting of the raceways causing impacts when overrolled while hydrodynamic plain bearings tend to wear as a result of mixed friction. For tracking the occurrence of alle the named emissions, piezoelctric sensors are used in most applications. Objectives The examinations presented here aim on building deeper understanding of acoustic emissions caused especially by friction processes and to validate the measurement results gained by use of s.

2 It is well known from literature that structure-born sound signals are influenced by the mechanical properties of the sensor, its coupling to the structure and of course the source of emission itself [1]. As a result, e. g. the spectral content of measured signals will usually not be identical to that of the original source and depends significantly on the measurement equipment used. This applies particularly to s, as they are mechanical oscillation systems and hence show more or less pronounced resonance behaviour. Furthermore, significant scatter may occour even among specimen of the same type that are identical in construction [1, 4]. However, little effort has been taken on measuring the above mentioned influences so far. Here, a measurement shall be contrasted with a reference measurement system which is free of resonances and able to deliver results contactless and hence without reaction. These requirements are generally fulfilled by s. The additional benefit of three-dimensional mode shape information and full field results can be gained by the application of 3D-scanning vibrometers like the one used in this study (Polytec PSV 4-3D). Procedure As laser measurement technology requires direct optical access to the target of measurement, an application e.g. to bearing test benches without special design will inevitably involve considerable effort. Thus, first measurements are conducted on a steel cuboid in laboratory environment. Moreover, the uncomplicated geometry of the chosen structure allows for a rather reliable numerical calculation of its natural frequencies and the associated eigenmodes by means of a modal analysis using finite element methods (FEM). Figure 1 shows a photograph of the experimental set-up on the left and a corresponding signal flow diagram on the right. A piezoelectric actuator and sensor are mounted on the frontside of the steel cuboid which is at the same time scanned by three laser heads, acquiring the 3-dimensional vibration information. The actuator converts given signals into mechanical motion. Therefore, a force is applied on the structure s surface to simulate acoustic emissions. This translation of information into motion is influenced by the dynamic behaviour of the actuator and its coupling to the structure (resonances, etc.). As mentioned above, the same applies to measurements by means of mechanical sensors. signal excitation laser head 1 test item AE sensor coupling piezoactuator excitation by friction laser head 2 piezoactuator coupling structure AE-sensor laser system laser head 3 signal signal measurement Fig. 1: Photograph of the experimental set-up (left), zoom of the test-item front view (middle) and signal flow diagram (right) In acoustic emission analysis very high frequencies (up to several MHz, [1]) are commonly regarded. Unfortunately deterministic excitation and optical measuring of mechanical vibrations

3 are more difficult in higher frequency ranges. Similarly, in numerical analysis, such as FEM, the significance of results in higher frequency ranges is limited. The optical measurement system used here is capable of measuring vibrations up to 2 MHz in 3D mode, up to 24 MHz in 1D mode. The resolution depends on the chosen frequency range. The vibrometer is more sensitive for lower ultrasound frequencies. The measurement range of 5mm/s/V yields is limited to a max frequency of 1 khz, but yields the best resolution in the ultrasound frequency domain, of typically.2 µm/s/sqrt (Hz) [11]. The actuator from Physik Instrumente (model P-842.2) contains a preloaded piezo stack [12]. As the signals are very small and the best resolution possible is mandatory, the evaluations presented in this paper are limited to a maximum frequency of 1 khz. In many references [BARANOV et al.] friction is described as a source of very broadband emissions. So it is intended to generate a reference signal with flat spectrum and then compare the system s response to that of a real friction process. In principal, three types of signals offer a flat frequency spectrum and are commonly applied: impulse (i.e. striking the structure with a hammer), sinus chirp (sweeping a single frequency with constant amplitude through the bandwidth of interest) and white noise (random signal with flat spectrum). Scanning the test item s surface with the can be quite quick, but takes a relatively long time if heavy averaging is needed, as it is the case here for resolution purposes. During this time a constant excitation is needed (steady state conditions). Hence, it is practical to use a periodic sinus chirp in this setup with a frequency range from to 1 khz. Practiacally this means exciting all frequency lines of an FFT analysis within one time block with constant amplitude. The boundaries of the regarded system are defined by known inputs and outputs. In this case, only the input to the actuator and the output signal of the AE sensor and the respectively are known. Therefore, it is not possible, but also not needed, to derive the transfer function of the regarded structure itself, but only of the combination of structure, actuator and sensor where applicable and coupling as shown in fig. 1. The paramount subject of this study is the comparison of both measurement signals derived by the experimental setup on the one hand and comparison of two different excitation types (actuator vs. friction) on the other hand. These results are compared to results of a numerical modal analysis of the structure, i.e. the numerical mode shapes and eigenfrequencies. It turns out that it is rather difficult to generate a real friction process which is causing acoustic emissions strong enough to be measurable by the in this setup. Sufficient surface velocities can be achieved by using a double-cut file acting on the upper side of the test item. Using the approach described above, the key questions of this study can be summed up as: - Are the signal spectra measured by and comparable? - Are the signal spectra measured under frictional and piezoelectric excitement comparable? - Are the structure s Eigen frequencies visible in the spectra? This applies especially to the excitation by friction. Results From the measured signals, spectra are derived using a Fast-Fourier-Transformation (FFT). Fig. 2 shows a comparison of and AE sensor signal spectra with piezo-electric excitation (periodic chirp -1 khz). In case of the laser, the presented spectrum is averaged over all scan points on the surface while the AE sensor measures surface velocity only at one location. The positions of peaks noticeable in this plot are in principle a superposition of the resonance frequencies of the regarded system while their height is mainly influenced by damping phenomena. It should be noted that a 3 rd order Butterworth High-pass filter with a cut-off frequency of 5 khz is integrated in the amplifier of the AE sensor causing a suppression of low frequencies to

4 prevent overdriving of the subsequent measurement chain. The frequency response of the laser vibrometer is constant up to a limiting frequency of 1 khz for the chosen measurement range. Fig. 2 shows a good congruence of both spectra concerning peak localisation while the peak heights are different for the reasons mentioned above. It is obvious that the is less sensible than the at frequencies above 7 khz while the reverse is true at frequencies below approx. 45 khz. This is mainly due to the characteristics of the Piezo sensor, the sensitivity of the is very largely constant.the peaks at low frequency have higher amplitudes and dominate therefore the response of the vibrometer, while they are largely suppressed by the filter of the Piezo sensor. U / mv () v z f / khz Fig. 2: comparison of spectra of laser and AE sensor at periodic chirp excitation In Fig. 3 the calculated eigenfrequencies are taken into account. The vertical black lines denote the eigenfrequencies derived from the numerical modal analysis. These fit very well for frequencies below 3 khz. Possible discrepancies result e.g. from scatter of material parameters as well as a different system boundary (excluded actuator and sensor mass) and boundary conditions not taken into account in the numerical analysis (reaction force on contact area). As mentioned above, another source of discrepancies is the measurement of only one point with the AE sensor and a whole surface with the. The density of eigenfrequencies gets higher as frequency increases. Therefore, mapping of experimental and numerical results proves to be difficult at higher frequencies. U / mv () v z f / khz Fig. 3: comparison of spectra of laser and AE sensor at periodic chirp excitation with numerical eigenfrequencies (zoom: frequency range -3kHz) Table 1 shows some examples of calculated mode shapes (left column) mapped to measured vibration modes (right column) in combination with the corresponding frequencies. A good

5 agreement between measured and calculated deflection shapes can be observed. The slight asymmetric behaviour of the measured shapes is probably a result of the piezoelectric actor s mass which is excluded in the simulation as mentioned above. The areas where the piezo actuator and sensor are applied are not accessible for optical measurement (Note the white area in the top and top left region). Not all calculated mode shapes can be found in the measured data. One reason is that mode shapes with a deflection normal to the direction of excitation are (almost) not excited in the experiment. Table 1: Examples of calculated mode shapes and measured vibration modes (periodic sinus chirp excitation) Mode Nr. Frequency (Simulation) Mode Shape (Simulation) Frequency (Measurement) Mode Shape (Measurement) Hz 2219 Hz Hz Hz Hz 3263 Hz U / mv () v x f / khz Fig. 4: comparison of spectra of laser and AE sensor at friction

6 Regarding the excitation by friction, results of laser and AE sensor signals are less congruent as compared to the periodic chirp excitation. Fig. 4 shows an overall view of the derived spectra. The higher sensibility of the at frequencies above 7 khz is very pronounced in this case. A closer look at the lower frequency range shows some similarities (note the left ordinate s offset for better comparability): U / mv () v x f / khz Fig. 5: comparison of spectra of laser and AE sensor at friction (zoom: frequency range -55kHz) Including the calculated eigenfrequencies, we find some congruities, again particularly in the range up to 3 khz. However, the results are less conclusive than before. U / mv () v x f / khz Fig. 6: comparison of spectra of laser and AE sensor at friction and numerical eigenfrequencies (zoom: frequency range -3 khz) Fig. 7 shows a comparison of the signal spectra of AE sensor measurements under frictional and piezoelectric excitation. Obviously, higher frequencies are relatively pronounced under frictional excitation compared to broadband periodic chirp excitation. This suggests a characteristric spectral content of friction different to that of the periodic chirp.

7 U / mv friction) friction periodic sinus-chirp f / khz Fig. 7: comparison of spectra of AE sensor at friction and at periodic chirp (note the different scaling of the axes) U / mv piezoelectric excitation) As it is not possible to keep up a constant frictional excitement for a sufficient time interval to scan the whole surface in the described setup, only few points can be scanned by the laser. This only allows for a comparison in the frequency range and not the mode shapes themselves. Conclusions For piezo-electric as well as for frictional excitation the eigenfrequencies are noticeable in the measured signal spectra of both AE sensor and. The spectral results derived by means of AE sensor and show quite good congruence for broadband periodic chirp excitation. The similarities under frictional excitation are limited mostly to lower frequencies which are usually rather uninteresting for typical acoustic emission applications. This is mainly attributable to the higher sensibility of the at frequencies above 7 khz while the sensitivity of the is constant in the effective range chosen. The direct comparison of measured AE sensor signals at friction and at periodic chirp excitation shows a clear domination of higher frequencies in case of friction. Application Transferring this knowledge to bearing monitoring, there is one key issue of practical relevance: in rolling element bearings one of the most typical damage symptoms is pitting of the raceways causing impacts when overrolled while hydrodynamic plain bearings tend to wear as a result of mixed friction. An impact, simplified as perfect impulse, has the same frequency spectrum as the periodic chirp applied in this study. Therefore, in mixed bearing systems monitored by means of acoustic emission analysis, the differentiation of damage signals should be possible not only in the time but also in the frequency domain. This allows for a better distinction of both signals. Tests carried out on a test bench as shown in Fig. 8 (left) confirm this. A plain bearing and a rolling-element bearing can be operated simultaneously. Fig. 8 (middle, right) shows exemplary measurement results as spectrograms derived by an AE sensor at the point of measurement depicted in Fig. 8 (left). The middle diagram shows the signal measured during operation of a

8 rolling-element bearing with pitting in the outer raceway (undamaged plain bearing). The right diagram shows the signal measured at mixed friction in the plain bearing (undamaged rollingelement bearing). The different spectral contents are clearly visible and substantiate the results obtained in the presented study f / khz t / s t / s Fig. 8: spectrograms of damaged rolling-element bearing (left) and plain bearing under mixed friction (right), amplitude axis (color) in db ref. 1V 2-6 References 1. Albers A., Burger W., Scovino R., and Dickerhof M., 26, "Monitoring Lubrication Regimes in Sliding Bearings," Practicing Oil Analysis Magazine, 9(3), pp Albers, A., Dickerhof, M., 29, Interpretation von Körperschallemissionen bei der simultanen Überwachung von Gleit- und Wälzlagern, Dresdner Maschinenelemente Kolloquium (DMK) 29, Dresden, Schwalbe H., 1978, "Schallemissionsmessung an Gleitlagern zur Schadensfrüherkennung," Fachtagung Instandhaltung 1978 Inspektion, 17./18. Oktober 1978, Deutsches Komitee Instandhaltung, Wiesbaden, pp. VIII/1-VIII/8. 4. Sturm A., and Uhlemann S., 1985, "Diagnosis of plain bearings by acoustic emission analysis," Measurement, 3(4), pp Ziegler B., Schwalbe H., and Miszczak A., 27, "Nachweis des Reibzustandes in Gleitlagern mit Hilfe der Schallemission," 16. Kolloquium Schallemission Ellmer G., 1991, "Anwendung des Kurzzeiteffektivwertes der Schallemission bei der Gleitlagerdiagnose." Dissertation, TH Zittau 7. Al-Ghamd A. M., and Mba D., 26, "A comparative experimental study on the use of acoustic emission and vibration analysis for bearing defect identification and estimation of defect size," Mechanical Systems and Signal Processing, 2(7), pp Mba D., 23, "Acoustic Emissions and Monitoring Bearing Health," Tribology Transactions, 46(3), pp Sturm A., and Förster R., 199, Maschinen- und Anlagendiagnostik für die zustandsbezogene Instandhaltung, Teubner, Stuttgart, Chapter 7, p. 252 ff. 1. Kolerus J., 198, " Acoustic emission analysis Part 1 - Acoustic emission: Nature of origin, propagation and application.," Technisches Messen, 47(11), pp Data Sheet of Polytec PSV 4-3D, available at Data Sheets of Physik Instrumente Preloaded Piezo Actors, available at