Compensation-based non-destructive automatic quality control using acoustic methods Abstract Keywords Ingolf Hertlin RTE Akustik + Prüftechnik GmbH Gewerbestr. 26, D-76327 Pfinztal, Germany www.rte.de Cracks, porosity and density differences can occur during the manufacturing process. The non-destructive testing method acoustic resonance testing (ART) allows fast quality control of each single part integrated in the production flow. The method evaluates in particular the mechanical stiffness of the test specimen. The temperature of the part, dimensional variations and mass / density variations have an influence on the measurement results (resonant frequencies) in the same way as cracks and porosity and may mask them. Compensation methods allow the - permitted - productspecific variations to be distinguished from defects. In this way the method is able to detect capillary cracks reliably. The presentation describes the procedural method, the type and nature of compensation and the results of industrial applications. Non-destructive testing (NDT), quality control, acoustic, resonance analysis, crack testing, process compensation, acoustic resonance testing Introduction Acoustic sound analysis is a well-established non-destructive method for components that radiate sound, especially in metal and ceramic industry. The method is used in high-volume production to scan the parts fast and cost-effective within the production flow. Not only the raw materials, but also the manufacturing process affects the product characteristics and determines how a specimen sounds. Whereas human testers can easily adapt to this, an automatic inspection system relies on references and compares the measured sound to decide, whether the part is good or bad. In the past, solutions came up which did not cope with the complexity of this task: a) The product material is in general a mixture of different raw materials. Rheological additives and a heating process influence the stiffness of the part. Inoculation during the cast process influences the graphite structure. b) Inhomogeneity of the microstructure or density variation causes an acoustical variation. c) Dimensional variation changes the vibration modes of the part. d) Part temperature affects the elastic modulus and density and thus the resonance frequencies of the part. Many non-destructive inspections for the quality control use surface-related methods and do not assess the inner structure or defects like cracks as volume-related procedures do. These reasons and low performance of test systems at times resulted in low confidence in the assessments of fully automated final inspection systems. To cover the allowed process and part variation it is mandatory to eliminate, i. e. compensate, this impact on the acoustic behaviour of the part. Consecutively powerful classification methods can be used to find defects like capillary cracks or structural defects. Copyright 2009 - RTE Akustik+Prüftechnik GmbH, Germany page 1 of 7
NDT method Acoustic Resonance Testing Dynamic systems can be characterized in terms of one or more natural frequencies [1]. The natural frequency is the frequency at which the system would vibrate if it were given an initial disturbance (stimulus) and then allowed to vibrate freely. The physical model of a natural frequency is a mass hanging from a spring. The mass represents the part being tested and the spring represents its stiffness. The mass m vibrates up and down at its resonant frequency f, which is proportional to the stiffness k divided by the mass (undamped system). k f ~ (1) m The wave propagation in solids consists of longitudinal and transversal waves with the sound velocity cl = E(1 µ ) ρ (1 µ 2µ ²) ct = E 2 ρ (1 + µ ) with E elastic modulus, ρ density and µ Poisson s ratio. Elastic modulus and density are both temperature dependent. Any part has an infinite number of resonances, each determined by a specific combination of material properties and dimensions. Sound testing makes use of this effect: After suitable stimulation, e.g. by striking, the body vibrates in well-defined characteristic patterns and frequencies. A sound is a mixture of many single frequencies (in this case resonances) at a certain level and is not limited to the audible range but also includes the ultrasonic area. The oscillations, registered with a microphone, a laser vibrometer or an accelerometer, are so to speak the "language" of the test object, its distinctive fingerprint. They provide information about the entire test object and not just locally at the site of the sensor. Acoustic resonance testing (ART) [2] is a volume-related, non-destructive testing procedure, which makes the quick and cost-saving 100 % testing of test objects possible (Fig. 1). Sound testing in line production is an approved technology. (2) Fig. 1: Non-destructive testing methods (selection) The principal difference between ART and the classic NDT methods is that ART is sensitive for strength change: a structural defect reduces the stiffness and lowers the resonant frequency. Different defect types can be evaluated with one measurement, e. g. cracks, material deviations and structural deviations (inhomogeneity, stiffness / hardness, porosity). The main objective is that the defect influences the acoustical behaviour of the part and is significant with relation to the production scatter. Copyright 2009 - RTE Akustik+Prüftechnik GmbH, Germany page 2 of 7
The sensitivity (defect size, defect location) is also related to the part size and shape. Experimental and theoretical evaluations show that the size of a defect as well as the defect location has an impact on the frequency values [3]: A larger defect results in the shift to lower peak frequencies (same defect location) and a deeper defect in the shift to higher peak frequencies (same defect size). The first influence is higher than the latter and means that the resonance method is more sensitive for the expansion of a defect than its location. The decision for the applicability of the acoustic method requires systematic engineering and tests prior to automation. Why compensation? Various influences the radiation and measurement of the sound and hence the resonance frequencies (Table 1). In general there are influences from the manufacturing process (called process-related influences) and influences that come from the test bench and / or the test system set-up and measurement conditions (called test-related influences). Especially the excitation to generate the resonance frequencies is important: Position and direction are responsible for the level of a resonance frequency. The strength of the excitation can cause a linear or non-linear response from the specimen resulting in frequency changes and additional frequencies in the spectrum. The contact time is responsible for the frequency range and the damping of frequencies. Influence Effect Process-related influence Compen sation Method Aging hardness change (e. g. in castings) yes calculation Blowholes (minimal) disturbance of wave propagation no test objective Crack size / location wave propagation disturbance, flexural & longitudinal resonances no test objective Density wave propagation no test objective Dimension flexural & longitudinal vibration modes yes characteristics Hardness flexural resonances yes test objective Porosity wave propagation no test objective Raw material e-modulus and density, wave speed yes characteristics Temperature change of e-modulus and density yes measurement Weight see density / raw material yes measurable Test-related influence Human classification set up of characteristics, quality assessment no training strategies Excitation direction resonance stimulation yes engineering, layout Excitation force linear / non-linear part behaviour yes engineering, layout Excitation position resonance stimulation yes engineering, layout Excit. contact time resonance spectrum, damping yes engineering, layout Sensor charact. resonance measurement, resonance spectrum yes engineering, layout Sensor direction resonance measurement yes engineering, layout Sensor position resonance measurement yes engineering, layout Sensor type resonance measurement, resonance spectrum yes engineering, layout Support material wave propagation, damping, degree of freedom yes engineering, layout Support position wave propagation, damping, degree of freedom yes engineering, layout Test characteristics classification power yes engineering Test environment signal-to-noise ratio yes engineering, layout Test parameters measurement resolution, filter parameters and others yes engineering Table 1: Influence on the acoustical behaviour of a specimen Taking all this into account the challenge is to evaluate the impact of these process and test influences according to a specific test object and test objectives. Copyright 2009 - RTE Akustik+Prüftechnik GmbH, Germany page 3 of 7
Compensation techniques As explained above the (accepted) process variation results in the change of specimen behaviour and thus variation of the resonance frequencies To make the test parts comparable it is necessary to eliminate this influence as best as possible by adapting the resonance sprectrum. Within this contribution we explain first some of the process and then some test related influences. Precondition of applying frequency compensation is an extremely accurate measurement. Fig. 2 shows the relative standard deviation related to the vibration mode of a part with the first vibration mode at 8,5 khz and the tenth at 78,3 khz. ART has an excellent measurement precision of about 0.0035 % at the low frequency and 0.0006 % at higher frequencies [4] and depends only on the frequency resolution independent of the specific equipment. Experimental standard deviation (rel.) Rel. deviation 0,0000400 0,0000350 0,0000300 0,0000250 0,0000200 0,0000150 0,0000100 0,0000050 0,0000000 1 2 3 4 5 6 7 8 9 10 Vibration modes (1: 8,5 khz, 10: 78,3 khz) Process-related influence Temperature Fig. 2: Standard deviation dependency Temperature measurement precision depends on a number of factors but can be done with a precision of ± 0,5 C. When the temperature of the work piece changes, all resonance frequencies are affected. A part at lower temperature has a higher stiffness than a part at higher temperature. The result is a negative shift of the resonances. The temperature has a linear influence, but the gradient of the straight-line depends on the frequency (Fig. 3). Fig. 4 shows the effect of temperature compensation. Fig. 3: Temperature compensation (material: steel) Copyright 2009 - RTE Akustik+Prüftechnik GmbH, Germany page 4 of 7
Fig. 4: Frequency temperature compensation (x-axis: frequency, y-axis: level) Dimension To analyse the impact of dimensional variation it is very helpful to know the different vibration modes on basis of a FEM analysis [5]. Through model variation the (theoretical) frequency shift can be calculated. The importance of a dimensional variation depends on the shape of the part and the location of variation, but in general the influence is, e. g. for a bar: 1. The dimensional variation is proportional to the frequency variation, but with opposite sign (shift to lower frequencies when the dimension increases). 2. The variation in radial direction (diameter, transversal wave) has much bigger impact than in axial direction (length, longitudinal wave). A diameter variation directly affects the stiffness of the part! 3. The impact of a length variation dwindles with increasing resonance frequencies. 4. The impact of a diameter variation increases with increasing resonance frequencies. 11800 11760 Influence on heigth frequency [Hz] 11720 11680 11640 11600 11560 11,850 11,900 11,950 12,000 12,050 12,100 12,150 heigth [mm] Fig. 5: Dimensional impact (transversal wave) Fig. 5 shows the influence of a dimensional variation, in this case of a toothed ring with a diameter of 290 mm. The allowed variation of the height of 0,1 mm (0,8 %) causes a frequency variation of 80 Hz (0,67 %). Raw material Concerning acoustic analysis the material properties can simply be characterized by elastic the modulus E (Young s modulus), the density ρ, Poisson s ratio µ, bulk / Copyright 2009 - RTE Akustik+Prüftechnik GmbH, Germany page 5 of 7
compression modulus K and shear modulus G. For homogeneous isotropic materials the relationship between them for a bar is: E = 2G (1+µ) = 3K (1 2µ) (3) [6] A change of raw material can especially happen on a batch change. This influences the acoustic resonance response as follows: 1. A positive change for E leads to a positive deviation of resonance frequencies, weighted by ½. 2. A positive change for ρ leads to a negative deviation of resonance frequencies, weighted by ½. ρ is reciprocal to E (for similar material). 3. The impact by E and ρ is not affected by the value of the resonance frequencies. 4. µ has no influence in the lower frequency range, but increasing influence in increasing frequency. There are different approaches for compensation possible: a) Use of relative rather than absolute resonance frequencies b) Include parts with minimum and maximum characteristics c) Track abnormal / rapid resonance changes and adapt limits d) Pattern recognition (when rapid changes occur due to natural material variation and process conditions [7]). There is no (simple) relationship between E modulus and hardness or tensile strength. Test-related influence As visible in Table 1 there are many influences caused by the testing conditions. The compensation of the different items can be achieved, but experience and engineering is necessary. It is essential for acoustic material testing that the resonance frequencies are induced to be able to measure them. This is not only a matter of the excitation, but also of other parameters and system set-up. Work piece support Fig. 6 shows the influence of the support material and the number of support contact positions (within one circle are measurements with a 3-position support and a 4- position support; the latter has a bigger damping influence and results in a lower amplitude!). Sensor characteristics Fig. 6: Work piece support influence (left: 3,2 khz, right: 13 khz) Fig. 7 shows the influence of different sensor characteristics (in this case microphones from different manufacturers). Reasons for this are different sound field sensitivity and directional characteristics. If you do not use the appropriate sensor you might loose important information. Copyright 2009 - RTE Akustik+Prüftechnik GmbH, Germany page 6 of 7
Sensor type Fig. 7: Sensor characteristic influence (x-axis: frequency, y-axis: level) In addition it is a big difference if you use a microphone as a sensor with its integral capabilities or you use a laser vibrometer or piezoelectric transducer, which measure the part vibrations at a single position. Conclusion Acoustic methods can be applied to the quality control of various kinds of parts. Sometimes is this the only method to meet the requirements for a fast and comprehensive evaluation. Product-specific engineering plays a central role for the setup of characteristics and test procedures and creates the basis for the compensation of process-related influences. The benefit is an objective quality assessment and steadygoing inline high volume production control of high reliability and sensitivity. Guidelines are necessary and will be provided shortly to describe the approach for successful applications under the constraint of process and product related variation [8]. References: [1] Kinsler et. al.: Fundamentals in Acoustics. John Wiley & Sons 2000 [2] Hertlin, I.: Acoustic Resonance Analysis. Informative booklets for non-destructive testing; Vol. 5; Castell-Verlag, Wuppertal 2003 [3] Asano et. al: Impact acoustics methods for defect evaluation in concrete. International symposium on Non-destructive testing in civil engineering 2003 [4] Walte et. al.: Einflüsse von Bauteiltoleranzen und Bauteilfehlern bei der akustischen Resonanzanalyse. Jahrestagung DGZfP, Münster 2009 [5] Hertlin, I.: Acoustic Resonance Analysis Using FEM and Laser Scanning For Defect Characterization in In-Process NDT. Proceedings of the European Conference on NDT, Berlin 2006 [6] Heckl, M., Müller, H. A. et. al.: Taschenbuch der Technischen Akustik, Springer-Verlag, Berlin 1994 [7] Hertlin, I., Rieth-Hoerst, S.: Now Available: Fully Automatic In-line Roofing Tile Inspection. Tile & Brick International Manual 2007, Verlag DVS [8] DGZfP - Richtlinie Zerstörungsfreie Prüfung mittels akustischer Resonanzverfahren Methodik, Einflussgrößen, Voruntersuchungen, Validierung. DGZfP Berlin, 2009 (will be published) Copyright 2009 - RTE Akustik+Prüftechnik GmbH, Germany page 7 of 7