|NDT.net - March 2001, Vol. 6 No. 03|
12th International Symposium
on Nondestructive Testing of Wood
Detection of glue deficiency in laminated wood with thermographyBerglind, Henrik
Trätek - The Swedish Institute for Wood Technology Research
Skeria 2, S-931 77 Skellefteĺ
IKP - Institute for Polymer Testing and Polymer Science
Universitätsbereich Vaihingen, D-70550 Stuttgart
Corresponding Author Contact:
In the production process of gluing wooden layers, adhesion problems sometimes occur (Meinlschmidt et al., 1998). One explanation can be that the amount of glue is not sufficient or that glue is even missing in some areas. In order to minimise such quality problems, a non-destructive test method, which can give a continuous control of the process and/or the product, is needed.
This paper presents results from measurements performed in order to evaluate the potential of heating up thermography as a method to detect glue deficiency in laminated wood. Defect depth, defect size, degree of glue deficiency and wooden species of the surface layer have been varied.
Lack of glue with a thermal defect size, i.e. the quotient of defect size and defect depth, higher than 4 is detectable. Highest contrast is achieved for the largest defects (24 mm) below the thinnest (0,5 mm) surface layer made of the dark wooden species Merbau. Bright wooden species results in lower contrast. Starved glue joints show about half the contrast compared to areas with total lack of glue.
Thermography is a non-contact and non-destructive test method, which with the aid of a heat camera can give information about the thermal structure beneath the surface of an object down to a limited depth. Compared to many other non-destructive test methods, thermography is relatively fast and manages to examine large areas with relatively high velocity, which makes it suitable for industrial quality inspection purposes. At thermography testing, a heat flow inside the test object is often produced. When a defect is present, the heat flow is disturbed, which can affect the temperature of the object surface. The heat camera detects such a temperature contrast and the information is thereafter processed with suitable computer software. The main thermography methods are called pulse (Reynolds et al., 1984, Lau et al., 1991, Thomas et al., 1995), heating up and lock-in (Busse et al., 1992) thermography. Differences between the thermography methods consist in the type of heating and evaluation methods.
During the past 15 years, a considerable work has been done about detecting defects between a surface layer and a substrate with thermography. It has especially dealt about materials connected to the field of aerospace industry, such as carbon-reinforced polymers (Krapez et al., 1994), thermally sprayed coatings (Ahmed et al., 1987) and aluminium laminates (Maldague et al., 1987).
Some research has also been performed on glued wood products. Sembach et al. (1997) have detected 19 mm wide drilled slots 4-5 mm beneath the surface of medium density fibre boards and chipboards respectively using lock-in thermography. The same authors have also separated badly laminated from correctly laminated chipboards. Wu et al. (1995) have also used lock-in thermography. They detected holes with a diameter of 4 mm, differences in wooden species of the substrate and knots in the substrate beneath a 2mm thick veneer. Xu et al. (1993) have used heating up thermography, where 10-50 mm large areas without glue was detected with a contrast up to 0,7 °C beneath 1,3-3 mm thick surface layers of white Seraya.
The goal of the measurements presented in this paper is to study penetration depth for surface layers made of different wooden species and resolution capability of heating up thermography.
A test piece consisted of a surface layer glued on a 9,5-mm thick substrate made out of Scots pine. The surface layers consisted of four different wooden species with thicknesses between 0,5-4,0 mm. Merbau (Intsia bijuga) and Oak (Querqus robur) are dark, while Pine (Pinus sylvestris) and Alder (Alnus glutinosa) are bright. The surface layers were practically without knots, but oak and especially pine had marked colour differences between earlywood and latewood. Glue defects were simulated with different wide bands (3, 6, 12 and 24 mm), which alternately were covered with glue on both the surface layer and the substrate, just on the substrate and finally bands without glue (see fig. 1). Glue on both sides resulted in the recommended glue spread of 140 g/m2. One layer of glue resulted in 70 g/m2.
|Fig 1: Schematical sketch over the distribution of glue over the test piece. (Length: 155 mm. Width: 65 mm).|
Surface layers and substrates were glued with urea-formaldehyde-glue and hardening took place at room temperature during an hour with a clamping pressure of 0,03 MPa. The wooden species had a moisture content around 8 %. Each test piece was produced in three replicates.
Experimental set-up and measurement method
Figure 2.2 shows the experimental set-up during the measurements with heating up thermography. Three test pieces were mounted in a fixture made out of a heat insulating material. The heat camera was of the type AGEMA 900, which was placed in front of the test pieces at a distance of 1 m. The heat camera detects infrared radiation in the wavelength interval 8-12 mm and has a temperature resolution of 0,08 K. The heat source consisted of two halogen lamps with a power of 1000 W each. The lamps were placed 1.7 m from the test pieces and the optical radiation had an angle of incidence of 30o.
At the measurements, test pieces were illuminated from room temperature up to 40oC with a constant intensity and at the same time the heat camera sampled pictures with a frequency of 15 Hz. The images had a size of 300 x 150 mm2 and as images from the heat camera contains 272 x 136 pixels, the spatial resolution of the images was around 1 x 1 mm2/ pixel.
|Fig 2: Experimental set-up.|
In order to calculate contrast between bands with correctly glued areas and areas with glue deficiency, the average of the grey scale values, sij (i = 1 55), for each pixel column was calculated (see fig. 3a). These average values were plotted in a diagram with the length of the test piece as x-axis (j = 1 130), which resulted in signal profiles (se fig. 3b). The contrast was thereafter calculated as the difference in the signal value between a top above a glue defect and the average of the signal from the neighbouring correctly glued areas.
|Fig 3: a) A test piece seen from above with its signal matrix with 55 x 130 pixels, which were used to calculate average values of the signals from each pixel column. b) Signal profile above an area with lack of glue and illustration of the procedure for determination of the contrast between correctly glued areas and areas with glue deficiency.|
Maximal contrast and inspection time
During heating of the test pieces, a contrast between correctly glued areas and areas with glue deficiency is developed. In figure 4a, an example of three test pieces with 0,5 mm thick surface layers of Merbau is shown 11 s after the start of the heating. A similar image can be obtained already after 2 s, but it then contains more noise. In figure 4b, three diagrams are shown with profiles over the vertical average grey scale value plotted against the length of each test piece.
|Fig 4: a) Infrared image of three test pieces with 0,5 mm thick surface layers of Merbau. Areas with glue deficiency are bright. b) Three diagrams with signal profiles in the length direction of the test pieces. High values indicate glue deficiency.|
According to the infrared image and the signal profiles in figure 4, it is possible to separate between an area without glue and correctly glued areas for all defect sizes, i.e. 24, 12, 6 and 3 mm. For areas with one layer of glue, the signal varies mainly linearly between areas with zero and two layers of glue respectively, which indicates that the glue had spread somewhat before the hardening. For the 24 mm wide area with one layer of glue at the test piece at the top of figure 4a, a plateau has been formed. Here the glue had become enough viscous before the manual spreading of the glue on the wood, so the glue did not spread at all. The contrasts from the defects are superpositioned by temperature gradients whose forms vary from test piece to test piece. These gradients were caused by an uneven intensity distribution of the optical radiation from the two lamps.
In figure 5 it is shown how the contrast between a 24 mm wide band without glue and neighbouring areas with two layers of glue varies with heating time for different defect depths.
|Fig 5: Contrast between a 24 mm wide area without glue and neighbouring areas with two layers of glue beneath 0,5-2,0 mm thick surface layers of merbau. Each value represents the average of three replicates..|
Inspection time, i.e. heating time until maximal or enough contrast has been achieved, increases with defect depth like an ordinary echo. The maximal contrast decreases with defect depth, which is due to absorption and lateral spreading of the heat flow. The contrast between areas without glue and areas with two layers of glue and its variation with time is shown in figure 6 for different defect sizes.
|Fig 6: The contrast between an area without glue and neighbouring correctly glued areas and its variation with different defect sizes below a 0,5 mm thick surface layer of merbau. Each value represents the average of three replicates|
The contrast increase is slower and does not reach as high values for small defects as for big defects. The dependence on the defect size for the contrast has been explained by Almond et al. (1994). All heat energy that meets a defect is not reflected. Apart from some transmitted energy through the defect, some heat will flow around the edges of the defect. The influence of this effect decreases with defect size, which explains why contrast increases with defect size.
Penetration depth and resolution capability
The contrast has been measured on 0,5-4,0 mm thick surface layers of Merbau. The variation of how the measured contrast, C, varies with the defect size, D, and the defect depth, L, was curve fitted with multiple regression in the software Statgraphics and in Simca. The following equation was obtained with a coefficient of determination, R2, of 90 %, a coefficient of prediction, Q2, of 88 % and a standard error of 0,9 contrast units (D and L in mm, C is expressed in grey scale values according to the definition in fig. 3):
|C = 1,3+0,28D-0,91L|
|Fig 7: Contrast between areas without glue and neighbouring correctly glued areas and its variation with defect depth (0,5-4 mm) and defect size (3-24 mm) below surface layers of Merbau. The inspection times were 15 s. for 0,5 mm, 30 s. for 1,0 mm, 45 s. for 1,5 mm, 60 s. for 2,0 mm, 75 s. for 3,0 mm and 90 s. for 4,0 mm.|
The contrast decreases, as shown in figure 5, 6, and 7 with defect depth and increases with defect size. In order to be able to define the resolution capability of a thermography method independent of a certain defect depth or a certain defect size, the notion thermal defect size, i.e. the quotient of defect size and defect depth, is used. In figure 8 measured values of the contrast are plotted against thermal defect size.
Contrast increases with the logarithm of thermal defect size for Merbau. Apparently larger defects result in a higher contrast compared to smaller defects at the same thermal defect size. Below a thermal defect size of 3-4 units, the defect ceases to be detectable, which is comparable with delaminations in carbon reinforced polymers (Wyss et al., 1996). This means for example that the heating up thermography method should be able to detect 1,5-2 mm big defects below a surface layer of Merbau with a thickness of 0,5 mm or 9-12 mm big defects below 3 mm thick surface layers of Merbau.
|Fig 8: Contrast between areas without glue and neighbouring correctly glued areas plotted against the logarithm of the thermal defect size for four different defect sizes (3, 6, 12 and 24 mm) below 0,5-2,0 mm thick surface layers of Merbau. Each value represents the average of three replicates.|
Influence of the wooden species for the surface layer
Figure 9 shows how contrast varies with thermal defect size for different wooden species of the surface layer.
The penetration depth for a 24 mm wide area without glue is according to figure 9 above 2 mm. Merbau gives the highest contrast followed by Oak, Pine and Alder. The order depends probably on that dark wooden species absorbs more heat than bright wooden species. A higher absorption gives a greater heat flow, which spreads into the test object, reflects against the defect, returns to the surface of the test object and is detected as a higher contrast by the heat camera. The non-linear behaviour regarding Oak, Pine and Alder can be attributed to the error sources described below.
|Fig 9: Contrast between areas without glue and neighbouring correctly glued areas plotted against the logarithm of the thermal defect size for different wooden species. Defect size: 24 mm. Defect depth: 0,5-2,0 mm.|
There are a couple of effects that can influence the surface temperature of the test object and which therefore can be misinterpreted to be a sign of a glue deficiency. They are inhomogeneous lighting of the test object and variabilities of the wood, like occurrence of knots or occurrence of earlywood and latewood respectively. Image processing could possibly decrease their influence.
Temperature contrasts between the bright earlywood and the darker latewood disturbed the contrast determination to a higher extent for Pine compared to the other wooden species. In this experiment, well-defined borders between areas with different amount of glue were strived to attain. Because of the low viscosity of the urea-formaldehyde glue, it occurred that the glue spread somewhat after application or during pressing. This resulted in smaller defects and as a small defect gives lower contrast than a bigger one, some of the evaluated values could be misleading.
Heating up thermography works as a non-destructive test method to detect glue deficiency beneath a layer of solid wood.
In this experiment, the highest contrast was obtained for 24 mm wide areas without glue below a 0,5 mm thick surface layer made out of Merbau after an inspection time of 40 s. The contrast decreased to the half of that value if either defect size was decreased to 10 mm, if defect depth was increased to 2,5 mm, if brighter wooden species as Pine or Alder was used or if inspection time was decreased to 5 s.
The penetration depth was between 2 and 4 mm, depending on the wooden species. Below a thermal defect size of 3-4 units, the defects ceased to be detectable.
This work was performed with the support of the KK-foundation, NUTEK (the Swedish National Board for Industrial and Technical Development), Casco products, AB Gustav Kährs and Tarkett-Sommer.
|© NDT.net - firstname.lastname@example.org|||Top||