Method for detecting cracks in an aluminum alloy part produced by additive manufacturing.

Abstract

Method for detecting cracks in an aluminum alloy part produced by additive manufacturing. The invention essentially consists of an analysis of the ultrasonic resonance spectra (RUS) acquired on an aluminum alloy part produced by additive manufacturing. By comparing with parts made of the same material but without defects, the inventors have in fact noted that the spectra acquired on defective parts presented resonance peaks with significantly lower quality factors Q. Figure for the abstract: Fig. 9

Description

Title of the invention: Method for detecting cracks in an aluminum alloy part produced by additive manufacturing. Technical field

[0001] The present invention relates to the general field of non-destructive testing (NDT) of parts, in particular those produced using an additive manufacturing technique.

[0002] It relates more particularly to the NDT control of aluminum alloy parts produced by additive manufacturing.

[0003] The method according to the invention is particularly intended to be implemented for aluminum alloy parts manufactured by additive manufacturing technique implementing a preferably complete melting of the powder by means of laser radiation or by means of an electron beam. Prior art

[0004] Additive manufacturing of aluminum alloy is currently at a much less advanced stage of development than other materials, such as polymers, titanium alloys or nickel bases.

[0005] One of the obstacles to overcome for the additive manufacturing of aluminum alloys is to demonstrate that the parts produced can meet high technical constraints, such as for example to meet an aeronautical qualification or in the automotive sector.

[0006] However, it turns out that for certain series of alloys known to those skilled in the art to be difficult to weld (series 2xxx, 6xxx and 7xxx in particular), a particular characteristic of the parts produced by additive manufacturing is that they are likely to contain numerous cracks / microcracks. This is shown in [Fig.l] on which the presence of cracks in an Al 6061 alloy produced by selective laser melting (SLM) can be seen. It should be noted that the scale bar indicated on this optical micrograph represents a length of 200 μm.

[0007] These cracks / microcracks significantly reduce the mechanical performance of the produced parts. Although not fully understood in terms of mechanisms, this crack phenomenon is well known to those skilled in the art, and is associated with the joint presence of columnar solidification structures and thermomechanical stresses.

[0008] Also, to overcome the barrier mentioned above, there arises the unavoidable problem of detecting cracks which form during the production of aluminium alloy parts by additive manufacturing, in particular for materials subject to hot cracking phenomenon.

[0009] The a priori most obvious and simplest method for measuring the density of a part, and in particular checking whether or not it is cracked, consists of making one or more metallographic cuts in the part: [1], With appropriate polishing and chemical attack, possible cracks are easily revealed. However, this method is destructive and local.

[0010] Another technique conventionally used for the characterization of parts produced by additive manufacturing is the measurement of their density by measuring the Archimedes thrust: [1], This method is relatively simple to implement, but it is relatively imprecise and requires a large number of defects in the parts to provide usable results.

[0011] Helium pycnometry also allows the absolute density of a part to be measured. This method is probably more accurate, but it is limited to the characterization of small volume parts. In addition, knowledge of the density does not allow the presence of internal cracks to be detected in all cases.

[0012] A more advanced technique, insofar as it allows the localization of defects, is that of X-ray tomography: [1]. An interesting point is that, due to their low density, aluminum alloys lend themselves well to characterization by X-ray tomography. However, the method is poorly suited to the characterization of large parts, and requires a significant investment for the acquisition of the source and the detector. Furthermore, the acquisition time for the plurality of views necessary for the reconstruction is long.

[0013] Techniques using the propagation of ultrasonic waves (US) also make it possible to characterize the material health of parts and obtain information on the structure of parts produced by fusion / solidification. A known result of the state of the art is that US waves are diffracted on structural defects (grain boundaries, cracks, etc.) of the material, which results in ultrasonic signals that are noisier than on non-defective materials, or even exhibit echoes linked to the location of defects.

[0014] A particularly interesting technique is resonance ultrasonic spectroscopy (RUS). This RUS technique is based on the establishment of standing waves in the part to be tested and on the measurement of a certain number of frequencies or resonance peaks of said part. Used in connection with a numerical simulation of the equations of motion, for example by the finite element method, the technique allows a precise estimation of the elastic constants of the materials [2].

[0015] This RUS technique can also be used for the characterization of defects since, as mentioned above, any structural defect will influence wave propagation. More precisely, on resonance peaks, the influence of defects can result in displacements, splitting, broadening and amplitude variations. In axisymmetric samples of known geometry, defects can even be localized [2].

[0016] Publication [3] addresses the problem of the effect of localized cracks of different lengths on the RUS resonance spectrum of silicon (Si) wafers. More specifically, the treatment is focused on a particular peak, denoted A in the publication, which is seen to shift and widen as the crack length increases. However, this publication [3] does not address the problem of evaluating the defective state of a part where many microscopic defects may be present, as opposed to a localized macroscopic defect. Furthermore, it assumes that the geometry of the part to be tested is precisely known.

[0017] However, as already mentioned, aluminum alloy parts produced by additive manufacturing can have numerous microcracks within them.

[0018] Another characteristic of these parts is that they have a very large number of crystallographic grains, which are generally associated with the extreme solidification conditions (very high thermal gradients, very rapid growth rate) encountered in additive manufacturing processes. In concrete terms, the observed grain width is overwhelmingly less than 30 pm. This is illustrated in [Fig.2], which is an electron backscattered diffraction (EBSD) map: it shows the grain structure in an Al 6061 alloy produced by SLM. It should be noted that the scale bar shown in this optical micrograph represents a length of 200 pm.

[0019] Even in highly cracked samples like those in [Fig.l], due to this crystallographic structure specific to the additive manufacturing process, grain boundaries are much more numerous than cracks in the samples [4].

[0020] There is therefore a need to find a solution for detecting cracks / microcracks in parts produced using an aluminum alloy technique, which in particular makes it possible to reliably distinguish cracks / microcracks and to overcome the drawbacks of the prior art cited above.

[0021] The general aim of the invention is then to respond at least in part to this need. Statement of the invention

[0022] To do this, the invention firstly relates to a method for detecting cracks in an aluminum alloy part produced using an additive manufacturing technique, comprising the following steps:

[0023] a / bringing an ultrasonic transmitter and an ultrasonic receiver into contact with the part;

[0024] b / transmission of signals from the transmitter to the receiver through the room;

[0025] c / measurement of the signal transfer function between transmitter and receiver;

[0026] d / from the measured transfer function, determination of a spectrum at least a resonant frequency peak;

[0027] e / determination of the quality factor (Q) for each resonance frequency peak

[0028] f / comparison of the determined quality factor (Q) or a median of determined quality factors with a reference quality factor (Qref) corresponding to an aluminum alloy part;

[0029] g / if the determined quality factor or the median of determined quality factors is greater than the reference quality factor (Qref), then the part is considered free of unacceptable cracks.

[0030] In step f / , the reference quality factor (Qref) preferably corresponds to an aluminum alloy part known to be of good quality, preferably without cracks.

[0031] Thus, the invention essentially consists of an analysis of the ultrasonic resonance spectra (RUS) acquired on an aluminum alloy part produced by additive manufacturing. By comparing with parts produced from the same material but without defects, the inventors have in fact observed that the spectra acquired on defective parts exhibited resonance peaks with significantly lower quality factors Q.

[0032] This solution is far from being obvious to implement since, as explained in the preamble, the crystallographic study of an aluminum alloy resulting from additive manufacturing clearly shows that the density of grain boundaries is much greater than the density of cracks.

[0033] Also, with a RUS signal analysis, one would have expected that the effect of grain boundaries would be largely dominant in the resonance spectra.

[0034] However, the inventors were able, against all expectations, to observe that in an RUS signal analysis the peak / noise ratio appears mainly linked to the cracks and that the analysis of the spectra according to the invention based on the value of the quality factor of at least one of the peaks of the resonance spectrum allows unambiguously the identification of the cracked parts independently of the grain structure in the analyzed part.

[0035] Contrary to state-of-the-art publications, in particular [3], the method according to the invention is not concerned with the shift of resonance peaks relative to a healthy reference state.

[0036] According to a first advantageous alternative, step d / consists of determining the peak of the fundamental resonance frequency of the part.

[0037] According to a second advantageous alternative, step d / consists of determining of at least 10 resonance peaks by a frequency sweep.

[0038] Advantageously, the reference quality factor (Qref) is at least equal to 800, preferably at least equal to 1000.

[0039] Advantageously again, step d / is carried out by a thresholding algorithm.

[0040] According to an advantageous variant, step e / consists of adjusting at least one peak by means of a Lorentzian function.

[0041] Preferably, step b / is repeated so as to determine at least two, preferably at least three, even more preferably at least five spectra according to step d / . The fact of making several acquisitions (recordings) makes it possible to detect any missing resonance frequencies and to verify that this number of missing frequencies is low.

[0042] According to an advantageous embodiment, the method comprises, before step d, a step dl / of merging the at least two spectra determined by masking. In the case where a missing frequency has been identified, by combining at least two recordings and merging them by masking.

[0043] Beyond the problem of missing frequencies, the procedure of multiple acquisition and summation of the recordings makes it possible to improve the signal-to-noise ratio of the resonance spectra.

[0044] The invention also relates to a system for detecting cracks in an aluminum alloy part produced by an additive manufacturing technique, comprising:

[0045] - an ultrasonic transmitter and an ultrasonic receiver;

[0046] - an amplifier connected to the receiver;

[0047] - a spectrum analyzer connected to both the transmitter and the receiver, for measurement of the transfer function between the transmitter and the receiver;

[0048] - an electronic processing unit connected to the spectrum analyzer, the unit of electronic processing being configured to implement the method as described above.

[0049] The invention also relates to a computer program product for implementing the detection method mentioned above.

[0050] The invention finally relates to a method for non-destructive testing (NDT) of an aluminum alloy part produced by an additive manufacturing technique, comprising a step of measuring the density of the part by an Archimedes thrust measuring technique before or after carrying out the detection method mentioned above.

[0051] The advantages of the invention are numerous, among which we can cite:

[0052] - by means of non-destructive ultrasonic testing, detection to reliably estimate whether aluminum alloy parts produced by manufacturing additive are cracked or not; - quick and easy process to implement; - possibility of implementing the process directly at the output of an additive manufacturing process without prior preparation; - complementary method with another method of non-destructive testing of parts, in particular a technique for measuring Archimedes' thrust.

[0053] The invention can be envisaged for all applications and industrial fields using aluminum alloys, in particular in the aeronautics and automobile sectors. More particularly, it is possible to envisage the non-destructive testing of reactor exchangers produced by additive manufacturing using a method according to the invention.

[0054] Other advantages and characteristics will become more apparent upon reading the detailed description, given for illustrative and non-limiting purposes, with reference to the following figures. Brief description of the drawings

[0055] [Fig. 1] [Fig. 1] is an optical micrograph showing the presence of cracks in an Al 6061 alloy produced by SLM.

[0056] [Fig.2] [Fig.2] is an EBSD electron backscattered diffraction map of the grain structure in an Al 6061 alloy produced by SLM.

[0057] [Fig.3] [Fig.3] is a schematic view of a system for implementing the crack detection method according to the invention.

[0058] [Fig.4] [Fig.4] is a photographic reproduction of part of the assembly of the sensors of the system according to the invention on an aluminum alloy sample to be analyzed.

[0059] [Fig.5] [Fig.5] a reproduction of a resonance peak spectrum obtained after extraction of peaks as a function of frequencies, using the detection method according to the invention, the spectrum obtained being superimposed on that of the original measurement.

[0060] [Fig.6] [Fig.6] is a reproduction of the quality factors of the peaks according to [Fig.5]

[0061] [Fig.7] [Fig.7] is a reproduction of the peak amplitudes according to [Fig.5].

[0062] [Fig.8] [Fig.8] shows the spectra of two separate recordings and their fusion by masking.

[0063] [Fig.9] [Fig.9] shows the distribution curves of quality factors for different aluminum alloy samples analyzed by the method according to the invention. Detailed description

[0064] Figures 1 and 2 have already been commented on in the preamble and are therefore not commented on further below.

[0065] [Fig.3] shows a crack detection system 1 implemented according to the invention.

[0066] The system 1 comprises an ultrasonic transmitter 2 and an ultrasonic receiver 3. These sensors are preferably piezoelectric.

[0067] As shown in [Fig. 3], the transmitter 2 and the receiver 3 are in direct contact with a part P made of aluminum alloy produced by additive manufacturing to be analyzed. Preferably, the transmitter 2 and the receiver 3 are each in direct contact with a corner of the part P. As shown in [Fig. 4], in particular depending on the shape of the part P and / or the sensors 2, 3, one or more point supports A can be considered to facilitate the measurement of the signals.

[0068] An amplifier 4 is connected to the receiver 3.

[0069] A spectrum analyzer 5 is connected both to the transmitter 2 and optionally to the amplifier 4 or otherwise to the receiver 3.

[0070] Finally, an electronic processing unit 6, for example integrated into a laptop computer, is connected to the spectrum analyzer 5.

[0071] The spectrum analyzer 5 makes it possible to measure the transfer function of the signals in transmission-reception.

[0072] The method according to the invention implemented by the system 1 is as follows.

[0073] The transfer function exhibits a series of peaks which are characteristic of the modes vibration characteristics of the part P, depending on its geometry, its elasticity, its internal defects

[0074] The adjustment of the spectrum analyzer is advantageously made according to the orders of magnitude of the known resonance frequencies of the part P in order to observe at least one resonance frequency.

[0075] Different methods exist for estimating the order of magnitude of the fundamental resonance frequency.

[0076] For example, to determine a search domain for the fundamental frequency of a part P made of aluminum alloy, the inventors chose to search for the fundamental resonance peak around the frequency determined by the relationship:

[0077] [Math 1]:

[0078] f0 = ^

[0079] in which c is an estimate of the smallest speed of elastic waves in the object considered and 1 an estimate of its largest dimension.

[0080] From a recorded peak spectrum, various quantities can be used to classify the tested parts into defective or healthy categories, i.e. acceptable in relation to technical specifications depending on the applications.

[0081] The signal processing carried out by the processing unit from the analyzer of

[0082]

[0083]

[0084]

[0085]

[0086]

[0087]

[0088]

[0089]

[0090]

[0091]

[0092] spectra is as follows. Resonance peak extraction is performed on the signal previously subjected to a thresholding algorithm. This may involve, for example, thresholding from portions of the signal above a certain amplitude level, typically using the "find jieaks" function: see for example [5], [6]. On each of the peaks extracted in the previous step, we determine by a least squares method ("least square fit" in Anglo-Saxon terminology) the three parameters (central frequency co0, amplitude A, quality factor Q) of a Lo-rentzian function given by the following formula: [Math 2]: As an example, Figures 5 to 7 show the data obtained for a spectrum of a cracked aluminum alloy sample. Specifically, [Fig.5] shows the spectrum obtained after peak extraction (vertical bars on the graph), [Fig.6] shows the adjusted Q factors over the frequency range, and [Fig.7] shows the peak amplitude. It is evident from [Fig.6] that the quality factors Q take values ​​between 200 and 700, which are indicative of a cracked sample for the aluminum alloy material evaluated. It may happen that resonance frequencies are invisible in the measured spectrum. This is due to the positioning of the sample on the emitter 2 and receiver 3. As a precaution, the inventors believe that it is preferable to make several recordings of the spectrum in order to verify that the number of missing frequencies is low. In the case where a missing frequency has been identified, it is proposed to create a new spectrum by combining two recordings and merging, using a masking window, the data from the two recordings. This procedure is illustrated in [Fig.8] from two recordings. The inventors carried out various tests with samples of Al 6061 and 7075 alloys. For these samples, quality factors Q lower than a reference value Q ref of 800 are indicative of cracked samples, while quality factors Q higher than this value, at the median, indicate samples without the presence of cracks. [Fig.9] reports this observation showing the distribution of quality factors for four samples in the form of cubes of Al 6061 alloy, two of which are free of cracks and the other two have cracks. The calculated median of the quality factors Q for each cube is indicated by the horizontal bar at center of the “violin” shaped curves.

[0093] Thus, in this [Fig.9], the median of the quality factors obtained is greater than the reference value equal to 1000 for cubes 22 and 4 and less than this value for cubes 21 and 2. And therefore, according to the method of the invention, the cubes referenced 22 and 4 are without cracks, while cubes 21 and 2 are cracked, which has been verified experimentally by metallography. It is also interesting to point out that cube 22 is not dense in the sense of measurement by Archimedes' thrust, with a relative density of the order of 97%, which confirms the absence of correlation between the values ​​of the quality and density factors.

[0094] In this sense, the method according to the invention is a complementary method to density measurement, since a good part must be both dense and free of cracks. For example, cube 22 is free of cracks but must be rejected due to its poor density, whereas cube 21 must be rejected despite its good density because it has many cracks.

[0095] The invention is not limited to the examples which have just been described; it is possible in particular to combine characteristics of the examples illustrated within non-illustrated variants.

[0096] Other variants and improvements may be envisaged without departing from the scope of the invention.

[0097] For example, if in the system envisaged, as illustrated, the contacting of the transmitter-receiver is carried out on corners of the part to be analyzed, one can also envisage other locations, or even add point supports A as shown in [Fig.4] to facilitate the measurement. List of cited references:

[0098] [1]: “Comparison of density measurement techniques for additive manufactured metallic parts”, AB Spierings, M. Schneider, R. Eggenberger, Rapid Prototyping Journal 17 / 5 (2011) 380-386.

[0099] [2]: “Evaluation of defects in materials using resonant ultrasound spectroscopy”, K. Flynn, M. Radovic, J. Mater. Sci. 46 (2011) 2548-2556.

[0100] [3]: “Crack detection and analyses using resonance ultrasonic vibrations in full-size crystalline Silicon wafers", A. Belyaev, O. Polupan, W. Dallas, S. Ostapenko, D. Hess, J. Wohlgemuth, Appl. Phys. Lett. 88 (2006) 111907.

[0101] [4]: « A solution to the hot cracking problemfor aluminium alloys manufactured by laser beam melting », M. Opprecht, J.P. Garandet, G. Roux, C. Flament, M. Soulier, Acta Materialia 197 (2020) 40-5.

[0102] [5]: https: / / docs.scipy.org / doc / scipy / reference / generated / scipy.signal.find_peaks.html

[0103] [6]: « Improvedpeak détection in mass spectrum by incorporating continuons wavelet transform-basedpattern matching», P. Du, W. Kibbe and S. Lin, Vol. 22 no. 17 2006, pages 2059-2065 doi:10.1093 / bioinformatics / btl355.

Claims

Claims

1. Method for detecting cracks in an aluminum alloy part produced by an additive manufacturing technique, comprising the following steps: a / bringing an ultrasonic transmitter and an ultrasonic receiver into contact with the part; b / emitting signals from the transmitter to the receiver through the part; c / measuring the signal transfer function between transmitter and receiver; d / from the measured transfer function, determining a spectrum at at least one resonance frequency peak; e / determining the quality factor (Q) for each resonance frequency peak; f / comparing the determined quality factor (Q) or a median of determined quality factors with a reference quality factor (Q ref ) corresponding to an aluminum alloy part;g / if the determined quality factor or the median of determined quality factors is greater than the reference quality factor (Qref), then the part is considered free of unacceptable cracks.;

2. Method according to claim 1, step d / consisting of determining the peak of the fundamental resonance frequency of the part.

3. Method according to claim 1, step d / consisting of determining at least 10 resonance peaks by a frequency sweep.

4. Method according to claim 2 or 3, the reference quality factor (Qref) being at least equal to 800, preferably at least equal to 1000.

5. Method according to one of claims 2 to 4, step d / being carried out by a thresholding algorithm.

6. Method according to one of the preceding claims, step e / consisting of adjusting the at least one peak by means of a Lo-Rentzian function.

7. Method according to one of the preceding claims, step b / being repeated so as to determine at least two, preferably at least three, even more preferably at least five spectra according to step d / .

8. Method according to claim 7, comprising before step e / , a step dl / of merging the at least two determined spectra

9. System (1) for detecting cracks in an alloy part of aluminum produced by an additive manufacturing technique, comprising: - an ultrasonic transmitter (2) and an ultrasonic receiver (3); - an amplifier (4) connected to the receiver; - a spectrum analyzer (5) connected to both the transmitter and the receiver for measuring the transfer function between the transmitter and the receiver; - an electronic processing unit (6) connected to the spectrum analyzer, the electronic processing unit being configured to implement the method according to one of claims 1 to 8.

10. Computer program product for implementing the method according to one of claims 1 to 8 when said program is executed on a processor / computer.

11. Method for non-destructive testing (NDT) of an aluminum alloy part produced by an additive manufacturing technique, comprising a step of measuring the density of the part by an Archimedes thrust measuring technique before or after carrying out the method according to one of claims 1 to 8.

Images