Infrared Physics & Technology 43 (2002) 139–143 www.elsevier.com/locate/infrared
Porosity detection in composite aeronautical structures A. Ciliberto a, G. Cavaccini a, O. Salvetti b, M. Chimenti b, L. Azzarelli b, P.G. Bison c, S. Marinetti c, A. Freda c, E. Grinzato c,* a b
Alenia S.p.a., Area Aeronautica, 80083 Pomigliano d’Arco (NA), Italy CNR-IEI, Area della Ricerca CNR, Via G. Moruzzi 1, 56124 Pisa, Italy c CNR-ITEF, Corso Stati Uniti 4, 37127 Padova, Italy
Abstract The paper is devoted to the identification of anomalous porosity level on composite structures by thermography. The thermal properties mapping of local diffusivity is used for this purpose. Experimental results obtained in different experimental conditions are reported. These results are compared with findings given by Ultrasounds. A procedure for porosity classification on carbon fibre reinforced plastics by means of two-dimensional wavelet transform and statistical analysis is presented. Ó 2002 Elsevier Science B.V. All rights reserved. Keywords: IR thermography; Porosity; CFRP; Thermal diffusivity; Wavelet transform; PCA; NDE
1. Introduction One of the main and peculiar objectives in aerospace field is to guarantee the reliability, safety and durability of aeronautic structures. In particular, for this specific aspect, NDT plays an essential and fundamental role: NDT guarantee that the structure is free of flaws and defects detrimental to the safety of the aircraft. Such a verification shall be performed without damaging the structure. To this purpose, a lot of NDT methods are employed: Ultrasound, X-ray, eddy current, magnetic particles, penetrant, thermography, shearography, etc. These methods, and corresponding techniques provide with different sensitivity and resolution
*
Corresponding author. Tel.: +39-49-829-5722; fax: +39-49829-5728. E-mail address:
[email protected] (E. Grinzato).
levels, information about the presence of defects and their influence on the structure performances. For this reason, it is necessary that the component, along all the different phases (design, manufacturing development, production, in-service) is non-destructively tested, analysed and evaluated to verify its compliance to design, manufacturing, production and durability requirements. Non-destructive testing on aerospace components is continuously changing due to the developments in materials and manufacturing processes. Thermography is gaining a role that mainly depends on the reliability of findings, compared with other methods. Furthermore, dedicated procedures to particular component-defects couples are devised using advanced tools as numerical modelling or artificial intelligence based algorithms. Porosity evaluation is an area of great interest in the aerospace industry, due to the increasing production of structural and secondary parts in
1350-4495/02/$ - see front matter Ó 2002 Elsevier Science B.V. All rights reserved. PII: S 1 3 5 0 - 4 4 9 5 ( 0 2 ) 0 0 1 3 2 - 9
140
A. Ciliberto et al. / Infrared Physics & Technology 43 (2002) 139–143
carbon fibre reinforced plastics (CFRP) and to the important role played by the presence of defects on the interlaminar shear strength of composite laminates. In fact, that can be detrimental for some mechanical properties, e.g. lowering fatigue resistance, increasing the susceptibility to the water intrusion and influencing strength features. Porosity is normally investigated by the Ultrasound method, but under usual conditions, voids are not always distinguished from resin rich zones, that are themselves defects under some circumstances. Furthermore, the distribution of pores has a continuous behaviour into the bulk, making the ordinary direct control by an expert operator difficult. The sample used is representative of a fuselage panel of typhoon. It is a composite panel with CFRP skin and honeycomb nomex manufactured by bonding process.
2. Thermal NDE The presence of a higher percentage of microscopic air bubbles decreases both the thermal conductivity and the material average density. Con1 sequently, the thermal diffusivity ða ¼ kðqcÞ Þ, that is the ratio between conductivity (k) and heat capacity (qc) is also altered. The effective specific heat (c) and density (q) of composite derives from their basic components using the law of mixtures: qc ¼ qa ca u þ qm cm ð1 uÞ; where u is the volume fraction of the air bubbles (subscript a). The effective conductivity of a heterogeneous material can be roughly estimate by Eq. (1) [1], in which a large number of particles is randomly distributed in a homogeneous matrix (subscript m). k ¼ km
ka ð1 þ 2uÞ þ 2km ð1 uÞ ka ð1 uÞ þ km ð2 þ uÞ
ð1Þ
Eq. (1) has been evaluated for spherical particles attributing null conductivity to the alveoli [2]. Theoretically, temperature dependence of thermophysical properties should also be considered, but this does not seem necessary in the temperature ranges [3] used in TNDE inspection [4].
In our case, the effective conductivity decreases more than the heat capacity, resulting in an inverse proportionality of the local diffusivity on the po0:5 rosity. Notice that effusivity ðb ¼ ðkqcÞ Þ is even more correlated to porosity, but the measure is strongly affected by the absorbed energy. The schema for the testing procedure is very close to that used in the standardised Parker’s method [5]. The adopted procedure includes the heating of the rear surface of the sample with three photographic lamps of 1 kW each, set on a row and synchronized with the acquisition of a thermographic sequence of the other side of the sample. The heating lasts for 1.5 s, while the data recording, having a frequency of 10 Hz, lasts for 30 s. After the scanning of the whole sequence, the maximum temperature is fetched for each pixel (see Fig. 2a). Then, the time when half the maximum values are reached (s0:5 ) is mapped in a synthetic image. Finally, any pixel is processed using 1 the classical formula a ¼ 0:139ðl2 Þðs0:5 Þ , where l is the sample thickness. Fig. 1b shows the diffusivity map inside the region selected in Fig. 1a according to the left side scale. The lightest areas have not to be considered because corresponding to different materials. Here, regions of lower diffusivity (about 5 107 m2 s1 ) are clearly seen, indicating hidden structures because of their regular geometric shape. Furthermore, other regions are suspected to have an anomalous porosity, but the noise level does not allow at this level, a sure rejection. Tests have been repeated in different conditions. Fig. 2a and b clearly shows the same lower diffusivity and therefore higher porous areas in the same position, even if the specimen was rotated horizontally and rear-front. Testing the sample on both sides is studied the capability to evaluate porosity, independently from the depth of the defected volume. Nevertheless, a small difference in the absolute values is seen acting as an offset.
3. Ultrasonic NDE of porosity and classification by 2D wavelet Fig. 3 shows in the background an Ultrasonic C-scan of the sample obtained by through trans-
A. Ciliberto et al. / Infrared Physics & Technology 43 (2002) 139–143
141
Fig. 1. (a) Maximum temperature distribution along the sequence; the marked region has been considered for the diffusivity map. (b) Diffusivity map of the sample as shown on the left scale; lighter regions corresponds to different materials or depth; defected and suspected porous regions are darker.
Fig. 2. (a) Diffusivity map of the upper-front part of the panel; the lower part received a scarce heating level. (b) Diffusivity map of the upper-rear part; as in Fig. 3a the same areas indicates a higher diffusivity.
Fig. 3. C-san of the sample by Ultrasounds; learning regions are marked (Ln inside a box) and test regions (Tn), as used for the porosity classification procedure.
mission technique using 2.25 MHz search unit and a 1 mm index. The C-scan shows some porosity areas and some region with artefacts due to the surface bad preparation. The tool used for the enhancement of the porosity evaluation is based on a two-dimensional (2D) wavelet transform, applied on the results given by the time domain processing, previously performed. Briefly, the procedure computes the standard deviations of the histograms of the details images produced by the wavelet transform and applies the principal components analysis technique (PCA) to the set of computed values. In such a way, the input image is mapped onto the plane defined by the first two principal components [6,7].
142
A. Ciliberto et al. / Infrared Physics & Technology 43 (2002) 139–143
The procedure is composed of a learning phase, where images obtained from sound samples are processed and a measurement phase applied to the area under test. The conformity of each tested sample (and in particular the presence of porosity defects) is determined according to the distance in the PCA plane between the point mapped from the test image and the centroid of the cluster determined by the mapping of the reference images. In this example the procedure has been applied to a single thermal image, by using in the learning phase an adequate set of windows selected in zones of the image without defects. For each region processed in the measurement phase it is possible to estimate the presence of defects, and to rank them. In this test, 10 learning windows (L0–L9) have been chosen in homogeneous areas without evident defects, and 12 test regions (T0–T11) have been subsequently considered (see Fig. 3), all regions have 20 20 pixels size. Fig. 4 shows the results obtained using the wavelet function coif 5 with decomposition level 4, that is the mapping points of the learning regions with the associated cluster inside the circle and the projection of the test regions. Points T8 and T9, derived from regions very similar to the reference ones, lying inside the cluster. All other points are mapped outside, so indicating some difference with respect to the reference regions, like the presence
of porosity. In particular, possible defects are detected in regions T3, T10, T11, and in regions T0, T1 and T2. Actually, these last regions correspond to a hidden structure buried in the sample. 4. Conclusions The measure of the thermal diffusivity map has proven to be effective for the porosity evaluation of components made by CFRP. In particular three regions have been identified as highly porous. Additional information about hidden structures or the materials and thickness changing of the part are useful by-products of the test. From a technical point of view, the used equipment is simple and the heating evenness is not a critical point. Therefore, the only drawback of the procedure is the need to see both the sides of the part. A mathematical evaluation about the thermal properties varying indicates that the highest porosity gives the lowest diffusivity values, much lower than a rich resin area. This is an interesting future field of investigation. However, this achievement requires a very good data filtration procedure because of the very low signal level processed (a few mK). The comparison with findings given by Ultrasound and IR thermography is satisfying, even if the US map has not a clear definition of the porous areas and artefacts are present. The porosity evaluation given by the statistical Analysis could fully automate the NDE process. Acknowledgements This work has been partially developed in the mainframe of the BRITE-Euram project INDUCE; Authors would like to thank Eng. Marco Bozzi for his valuable contribution in the preparation of this paper and M. Agresti from Alenia. References
Fig. 4. Porosity classification given by 2D wavelet trasform and PCA; learning regions (cross) and sound areas are inside the circle, porosity of tested regions is proportional to the distance from the circle centre.
[1] J.F. Kerrisk, Thermal diffusivity of heterogeneous materials. II. Limits of the steady-state approximation, J. Appl. Phys. 43 (1) (1972) 112–117. [2] J.F. Kerrisk, Thermal diffusivity of heterogeneous materials, J. Appl. Phys. 42 (1) (1971) 267–271.
A. Ciliberto et al. / Infrared Physics & Technology 43 (2002) 139–143 [3] NASA, TPSX––thermal protection systems expert and material properties database, Web Edition (Unrestricted) @http://asm.arc.nasa.gov/tpsx/tpsxhome.shtml, 1999. [4] Xavier P.V. Maldague, Nondestructive Evaluation of Materials by Infrared Thermography, Springer-Verlag, Berlin, 1993. [5] W.J. Parker, R.J. Jenkins, C.P. Butler, G.L. Abbot, Flash method of determining thermal diffusivity, heat capacity, and thermal conductivity, J. Appl. Phys. 32 (1961) 1679.
143
[6] G. Cavaccini, M. Agresti, G. Borzacchiello, E. Bozzi, M. Chimenti, O. Salvetti, An evaluation approach to NDT Ultrasound processes by wavelet transform, in: Fifteenth World Conference on Nondestructive Testing, Roma (Italy), October 15–21, 2000. [7] E. Bozzi, M. Chimenti, O. Salvetti, Standard wavelet processing, BRITE Euram Consortium Contract No. BRPR-CT98-805, Technical Report TR-4130, February 2001.
