Non-Destructive Component Testing by Acoustic Resonance Analysis
Acoustic resonance testing (ART) is a relatively new Non-Destructive Technology (NDT) technology for many types of components: sintered, cast, forged, ceramic, etc. It is especially well suited for economic 100% testing in mass production. ART is a comparative method for quality assessment, which allows testing both the surface near region and the volume of a part within a short cycle time (Figure 1).
Figure 1: Acoustic Resonance Non DestructiveTest
Because of the sound made during the inspection of the part, this type of test is sometimes referred to as "ping testing". The process is also sometimes referred to as "crack testing" or "crack detection testing".
The following article explains the preconditions of the method, its functional principle and its implementation into hardware and software, and the important influence factors on practical test setup.
Contents: 1. History of Acoustic Resonance Testing 2. Basic Principles 3. General Requirements 3.1 Impact Device 3.2 Slide Device 3.3 Comparison of Devices 4. Sensor and Excitation Locations 4.1 Microphone 4.2 Accelerometer 4.3 Laser 4.4 Locations 5. Calculation of Frequency Spectrum 6. Further Data Evaluation and Part Assessment 7. Application Examples
1. History of Acoustic Resonance Testing
Acoustic resonance testing (ART), although being one of the oldest quality testing methods of the world in handcraft like pottery or glass production, is a relatively new method for industrial series production. For the industrial use of automated ART solutions, the testing roof tiles in the 1990's can be named as a starting point (Figure 2).
Figure 2: Crack in Ceramic Roof Tile.
Significant enhancement and optimization work was necessary to transfer automated ART solutions to successful testing of metal parts. Metal part defect detection requires significantly higher quality standards compared to coarse ceramic parts.
This has led to the development of different ART testing systems:
Basic systems which simply check if certain resonance frequencies appear or if they are within a defined frequency range. This is still a simple and rather cheap approach for easy to answer questions (e.g. Simcenter Anovis Resonance Finder).
Neural network approaches (now often called “artificial intelligence” or short “AI”) which currently are in fashion for a lot of applications. AI based systems need pre-sorted references for good parts and all the different types of bad parts to be found. Thus, they cannot deal with differences between production lots or with drift of the resonance behavior within a production lot. Such were the above-mentioned testing systems for roof tiles in the 1990's.
Statistic-based ART systems (Figure 3) which only need good parts from current production for adjustment. Such systems can deal with differences between production lots, as well as drift within a production lot. Statistic based ART systems offer the possibility to sort out small defects as well as defects that were not previously known.
The rest of this article concentrates on the characteristics of such a statistic-based ART systems, such as Simcenter Anovis Non-Destructive Testing (NDT) system.
2. Basic Principles
ART is a comparative method for quality assessment, which allows testing both the surface-near region and volume of a part within short cycle times. It assesses the resonant frequencies of nominally identical parts in series production. The resonance frequencies of parts are defined by internal characteristics like mass, density, geometry, elastic modulus and stiffness. Thus, the results of the method are directly connected with characteristics reflecting the stability or endurance of the tested parts, and not with optical surface characteristics.
But how does ART work in practice? Generally, there are three steps:
Excite the resonance frequencies of each single part.
Record the sound or vibration signal with suitable sensors.
Finally, the signal needs to be processed to come to a quality assessment.
The sound excitation should not cause damage or destruction of the object under test. However, the sound must be sufficiently intensive so that non-linear behavior (caused e.g. by cracks) can occur, and has to be as short-pulsed as possible. This is achieved either by hitting the part with an (electromagnetically driven) impact device or by use of a slide device, i.e. by letting the part drop onto an impingement plate. Their advantages compared to piezoelectric actuators or pneumatic impulse devices are not discussed here.
3. General Requirements
The selection of a suitable Non-Destructive Testing (NDT) method (or a combination of different NDT methods) which is able to find all quality-relevant issues must be tuned to the particular part production. Answers to questions like the following need to be found:
Which types of flaws need to be detected?
What is the minimum defect size?
What material is the part made from?
Which geometry and at which locations?
What is the depth of the component?
What is the allowable cycle time for testing during manufacturing?
All such questions help to find an answer to the basic question: Which NDT method fits all these requirements?
For acoustic resonance testing as a comparative NDT method, a preliminary study needs to be performed in order to give reliable answers to questions like the ones above. The main reason is that the minimum detectable defect size strongly depends on the scatter of normal series production: ART sorts out parts with significantly other resonance behavior than normal production. But at the beginning of a project the extent of normal production scatter for the respective component is yet known.
One of the first considerations is the type of device to be used to excite the part: Impact or Slide Device (Figure 4).
Figure 4: Impact device (left) and slide device (right) for excitation of resonant frequencies.
An impact device hits the component with an electromagnetically driven hammer, while a slide device uses an impingement plate made of hard metal from a short free fall.
3.1 Impact Device
An impact device hits the component with an electromagnetically driven hammer.
A part support (“part settlement”) is the key component for excellent reproducibility of the results. It provides reproducible positioning of the component and, at the same time, avoids clamping of the component. Clamping would cause damping of the vibrations of the object under test. Its resonance frequencies shall not be influenced by more than what is needed to hold it in a stable, defined position. Hitting components that lie – more or less even – on a moving conveyor belt results in much lower quality and reproducibility of the obtained spectra. It therefore offers only an appropriate solution for severe defects like completely or almost completely cracked rings. In each case, point and intensity of excitation needs to be selected.
In order to decrease the duration of vibration excitation, the impact hammer is accelerated to rather high speed, depending on the selectable acceleration voltage. The impact hammer itself has a relatively small mass m compared to that of components suitable for this kind of excitation, because its kinetic energy E = 0.5 × m × v2 must stay below the damage level of the respective component. The lower limit of impact energy is given by the necessity to measure sufficiently high frequency amplitudes for a stable automatic evaluation that allows detecting the quality-relevant issues. After hitting the component, the impact hammer, which is very near to the microphone, (Figure 5) must not make any sound during measurement time, i.e. immediately after hitting the part. Furthermore, its tip needs to be separated immediately and steadily from the component’s surface .
Figure 5: Impact device (top center) used to test a pulley (middle center). A microphone (top right) records the acoustic response to determine if there is a fault. Elastomer inserts (bottom, green) avoid that the part springs back.
This function is achieved by the low mass of the hammer part that gets reflected very well after hitting a component placed in a stable part settlement) in connection with a fine-tuned distance between the component’s surface and the ultimate position of the hammer tip, and it is supported by an additional spring mechanism.
At the same time the component must not spring back after being hit, which is achieved by appropriate design of the part support and its elastomer inserts.
3.2 Slide Device
A slide device uses an impingement plate made of hard metal from a short free fall to test a part to excite the part. The resulting impact is measured with a microphone.
The orientation of the component at the start, the angle of inclination of rail and impingement plate, and the distance between rail and impingement plate need to be defined according to the geometry of the part under test. Design goal here is that the component must not tumble during free fall or after hitting the impingement plate. Sound excitation by a slide device enables extremely short testing cycles down to less than two seconds.
Excitation intensity can be varied in some extent by changing the height of drop. But it is not possible to freely select the point of excitation because it mainly depends on the orientation of the part at the slide’s release pin, on its interaction with the guide rail, and on the part’s free fall in connection with its center of gravity. A good design of the slide rail (Figure 6) and an appropriate choice of the part’s orientation are indispensable for achieving good reproducibility of the measurement results.
Figure 6: Slide device used to test shift rings for gearboxes.
The risk of damage is strongly dependent on the individual component’s material and geometry and on the point where it hits the impingement plate. Excitation via slide device is – as general rule – well-suited for components lighter than 50 grams. One reason for this is that testing of heavier components would increase the abrasive wear of the impingement plate considerably. Another reason is that there is a minimum drop height for the component mainly determined by its diameter, because the component must have left the rail and fall freely before hitting the impingement plate. Bigger (and therefore often heavier) components also require a bigger free-fall height, and the resulting impact energy might be too high to guarantee non-destructiveness.
3.3 Comparison of Devices
In the following, the two concepts of sound excitation are compared regarding vibrational isolation, restrictions of the respective concept, and consequences on the cycle time for testing.
Both devices the part support for the impact device and the impingement plate of the slide device, need vibrational isolation from the environment, so that almost exclusively the vibrational response of the object under test is recorded. This is achieved by elastomer inserts and the selection of appropriate mass and material for some functional parts of the system.
Both concepts of sound excitation are restricted regarding size and weight of the components to be tested. Sound excitation via impact device is well suited to mid and large sized components, but very small and/or light components might be expelled from the part support when hit by the impact hammer. Sound excitation via slide device is restricted to small and lightweight components.
Sound excitation via impact device enables two additional measurement modes compared to excitation via slide device: First, multiple impacts can be performed on different well-defined points of the same part, which might be necessary for testing or components that show rotational symmetry. Second, a measurement can be repeated. This is sometimes done in the NOK case, if pseudo rejects due to reasons like dirt on the point of excitation or rare cases of disturbing noise shall be excluded.
If the impact device is used for sound excitation, the part handling (loading and unloading the sample support) usually takes more time than the testing itself. Nevertheless, testing cycles of three seconds are normal. Significantly lower testing cycles can be achieved by parallelization of loading, testing and unloading via using several identical part settlements mounted on a turntable. For sound excitation via slide device, testing cycles below two seconds are feasible if all components fit well together: smart communication between handling equipment, testing equipment and sorting devices, supported by a performant PC and well-configured data analysis and storage functions.
4. Sensor and Excitation Locations
Guidance on possible sensors used in Acoustic Resonance Testing for Non-Destructive Tests, as well as locations for the their placement:
The standard sensor for acoustic resonance testing is a measurement microphone (Figure 7) recording airborne noise that is emitted from all regions of the part under test.
Figure 7: Broadband microphone used in Acoustic Resonance Testing (ART).
It depends on the defect specifications whether a bandwidth of 20 kHz is enough or a broadband microphone with a bandwidth of up to 75 kHz is required.
Structure-borne noise sensors like accelerometers (Figure 8) record the signal at only one point of the part’s surface and with limited bandwidth compared to broadband microphones.
Figure 8: Simcenter Anovis industrial accelerometer is applied in automated fashion to part.
One advantage of accelerometers is in noisy environments where factory sounds interfere with microphone measurements. The surface vibration of a part during a impact test is proportional to the emitted sound, but less susceptible to background noises. The use of structure borne noise sensors in acoustic resonance testing is common for car body parts or cast parts.
Contacting sensors can hardly be applied automatically to parts of medium and small size without strongly influencing the result. Non-contacting sensors such as laser Doppler vibrometers enable higher bandwidths but are relatively expensive.
4.4 Location Considerations
Sound excitation with the impact device is usually carried out from the side or from above, hitting the component perpendicular to the surface at the favored point.
Excitation point and microphone position need to be selected according to the component’s characteristics and the geometry of the part settlement:
Any highly symmetric microphone position with respect to the part should be avoided in order to prevent destructive interference of sound waves.
The same applies to the point where the impact device hits the part’s surface: Any resonance mode cannot be excited in its nodes, and resonance modes often show nodes at highly symmetric points (Figure 9).
Figure 9: Symmetric structures, like brake rotors, have areas of no response (nodal lines) that need to be avoided for successful ART tests.
The signal of interest starts after sound excitation has occurred. There is no benefit in using signals much longer than the period showing considerable sound level (see Figure 10).
Figure 10: Left - Time domain of excitation signal is short pulse. Right - Short time domain pulse creates broad frequency input. The recording time should match the excitation pulse to avoid extraneous signals and create pseudo-rejects.
It is a major task of the ART software to provide appropriate trigger functions to cut out the respective signal. Any further hit of the part during the signal of interest results in a significantly higher noise level which spoils the spectrum and may lead to pseudo-rejects:
Windows and Leakage: Applying a Fast-Fourier-Transform (FFT) on the time signal is a fast and applicable way to get the frequency spectrum. Since our signal of interest has of course limited length, a windowing function which has the value zero at the beginning and end of the sound signal must be used to avoid the appearance of leakage effects.
Spectral Resolution: The number of calculated signal lines strongly corresponds to the duration of the signal of interest and to the shape of the resonance peaks. Well-shaped resonance peaks support functions that allow to determine the frequency maximum positions with a higher resolution than given by the FFT spectral resolution.
The cycle times for in-line testing must keep pace with the production line: It takes about 1 second to perform a single sound excitation, to calculate the spectrum, to automatically determine the defined feature values and to evaluate them in the sense of a pass/fail decision. From this lower limit on the minimum cycle time it is rather a question of an optimized handling of the components to be tested.
The knowledge about signal processing and the experience in optimizing individually adapted part settlements or sound excitation parameters is one reason why the market for ART systems is dominated by solution providers instead of equipment providers. This knowledge needs to be transferred to the customer in any industry to the extent in which it is necessary and desired, e.g. during commissioning or by operator training.
All three categories as well as combinations of maximum positions (ratios, differences) form metrics for part assessment.
The challenge is to transform these metrics into an assessment of part quality. Experience shows that in real industrial production - except for some trivial tasks - this is not possible by an assessment via fixed limits. In fact, an adaption to the vibrational behavior of the whole production is necessary as shown in Figure 11.
Figure 11: Three scenarios for application of ART (see captions) and their consequences upon the width of the tolerance band for any assessment feature (graphs in the middle). Symbolic illustrations of the defect detection limit (red) that can be achieved within the respective scenarios are shown. The scenario on the right detects the smallest deviations, while the scenario on the left only detects larger deviations.
Figure 10 describes three scenarios with different defect detection limits:
Left Figure: Case where unsorted parts coming from different production lots (blue versus magenta) are tested: The production scatters of each single production lots adds together, creating a wider variation band. An individual component must show significantly different behavior compared to all production lots in order to be sorted out and identified as a bad part.
Middle Figure: All parts coming from one production lot (blue only) are compared to each other. As a result, smaller deviations can be identified versus evaluating multiple lots.
Right Figure: In the scenario on the right side, parts are tested in production order (first-in, first-out, FIFO). Since production scatter is a combination of current production scatter and drift over time, these two contributions can now be separated, and a narrow tolerance band shifts with the drift of the part characteristics. This scenario clearly has influence on production process and logistics, but it will lead to the best possible fault detection at minimum pseudo-rejects.
Since acoustic resonance analysis is a comparative evaluation differentiating between suspicious bad parts and the majority of good parts, the normal production scatter of the good parts defines the detection limit for what can be sorted out as significantly different.
7. Application Examples
Pictures of different applications of Acoustic Resonance Testing:
Figure 12: A 150 micron crack detected in sintered metal part.
Figure 13: Sound excitation with impact device for testing steering box pulley.
Figure 14: Car body part testing at the end of press line with accelerometer acting as vibration sensor.
Figure 15: Sound excitation with slide device to test shift ring of gearbox.