Determining Paint Adhesion to Wood Using a Uniform Double ...

Mark Knaebe

1

and R. Sam Williams

1

Determining Paint Adhesion to Wood Using a Uniform Double-Cantilever Beam Technique

REFERENCE: Knaebe, M. and Williams, R. S., “Determining Paint

Adhesion to Wood Using a Uniform Double-Cantilever Beam Technique,” Journal of Testing and Evaluation, JTEVA, Vol. 21, No. 4, July 1993, pp. 272-279. ABSTRACT: Accurately predicting paint adhesion without the lengthy time required for typical exposure and evaluation is advantageous. Mechanical tests that use shear block. lap shear, or tensile specimens to measure paint adhesion result in large data variability, making it difficult to differentiate subtle changes in adhesive strength. The objective of this study was to decrease the variability of measured paint adhesion in fracture toughness tests. Preliminary and computer-aided experiments were conducted using uniform double-cantilever beam (UDCB) and shear block specimens. The coefficient of variation (COV) for the shear block tests was about 15%. The UDCB specimens cut from the same panels as the shear block specimens had a COV of 10%. Computer-aided tests of UDCB specimens gave a COV of less than 5%. Results show that using UDCB specimens with the aid of a computer is an excellent technique for determining paint adhesion to wood. KEYWORDS: paint, uniform double-cantilever beam (UDCB), fracture toughness, adhesion, adhesive strength, bond strength

Paint adhesion is quantified as the work necessary to separate the paint from its substrate by propagating a crack through the paint/wood interphase. In typical adhesive tests, paint adhesion is determined by a second adherend that is bonded to the paint or a reinforcing material that is included in the film. This second adherend could simply be adhesive tape as used in the cross hatch method (which was written for metal substrates but used in our study for wood) (see ASTM D 3359, Test Methods for Measuring Adhesion by Tape Test). However, a load cell must be used for analytical measurements. The second adherend can be another wooden adherend as in shear tests or a wooden or steel rod as in a torsion test. Paint adhesion to wood can also be determined by embedding cloth or a similar reinforcing material in the paint when it is applied. After the paint cures, one end of the piece of cloth extending out of the film is attached to the test machine to measure paint adhesion. Regardless of the method used to measure paint adhesion to wood, data variability Manuscript received 8/10/92; accepted for publication 12/1/92. 1 Chemist and supervisory research chemist, respectively, USDA Forest Service, Forest Products Laboratory, Madison, WI 53705-2398. The Forest Products Laboratory is maintained in cooperation with the University of Wisconsin. This article was written and prepared by U.S. Government employees on official time, and is therefore in the public domain and not subject to copyright.

tends to be large, thus making subtle strength changes difficult to differentiate. In previous research work at the USDA Forest Service, Forest Products Laboratory (FPL), paint adhesion to wood was determined using shear block and tensile specimens [1,2]. In the first of these studies, the coefficient of variation (COV) of the data was high. However, the decrease in paint adhesion was also high. Results clearly showed that if the wood substrate was exposed to summer sun for up to 16 weeks prior to being painted, adhesive strength of primer paint could be decreased up to 50%. In the second of these studies, COVs as high as 25% were observed. When the wood was exposed for only 4 or 8 weeks, the decrease in adhesive strength of paint was less obvious as a result of the large variability in the data. The large COVs were caused by inconsistent loci of failure. In many specimens, the failure occurred partially or totally in the wood substrate. These results reflect the highly variable cohesive strength of wood, not the paint adhesion to wood. The control specimens and those with strong wood-paint bonds had the highest data variability. Paint adhesion to wood was also measured by Mahlberg [3 ] and Ahola [4] using the torsion method. They reported COVs of the same magnitude that were found in previous studies [1,2] , generally 10% and higher. The variability in data from adhesion tests of coatings on wood was similar to that found in testing of wood adhesives. In addition to the variability of data, the effect of specimen geometry makes it difficult to compare paint adhesion results. Strozier and others [5] attributed inconsistent data from shear tests to induced secondary Mode I stress located near the edge of the adhesive. Okkonen and River [6] and references therein reported that specimen size, shape, bond area, and tool type could affect paint adhesion. In tests using full- or reduced-size specimens, tangential- or radial-glued surfaces, and tools with or without offset (see ASTM D 143, Method of Testing Small Clear Specimens of Timber or ASTM D 905, Test Method for Strength Properties of Adhesive Bonds in Shear by Compression Loading), COVs varied from about 4 to 20%. No trend seemed to exist between COV and type of test. Although the intent of these studies was to accurately determine paint adhesion, lack of precision was also apparent. In this study, our objective was to decrease the variability of measured paint adhesion in fracture toughness tests using the uniform double-cantilever beam (UDCB) specimen. Determining the fracture toughness of materials on the basis of the UDCB specimen or the contoured double-cantilever beam (CDCB) 1993 by the American Society for Testing and Materials

272

KNAEBE AND WILLIAMS ON PAINT ADHESION TO WOOD

273

specimen is described in ASTM D 3433, Test Method for Fracture Strength in Cleavage of Adhesives in Bonded Joints. Both UDCB and CDCB specimens are currently used to test a wide range of materials. The CDCB specimens are currently used to test wood adhesives at the Forest Products Laboratory [7,8]. In this study, we compared COVs of paint adhesion using shear block tests to those obtained using a fracture mechanics test. The test was based on the UDCB specimen (see ASTM D 3433). The use of UDCB specimens for evaluating paint adhesion to wood is a new application. Experimental Two experiments were conducted. The first was a preliminary comparison of shear block and UDCB specimens cut from selected boards having similar grain angle; crack propagation was visually observed. In the second, specimens with consistent grain angle were cut from large timbers; crack propagation was determined with the aid of a computer. The techniques and analyses of ASTM D 3433 were modified to determine the location of crack propagation.

FIG. 1—Relationship of grain angle to modulus of elasticity.

Preliminary Comparison Experiment To determine the effectiveness of UDCB specimens compared to shear block specimens, western redcedar (Thuja plicata) with grain angles between 3° and 9° were cut into panels 6 mm by 102 mm by 330 mm (0.25 in. by 4 in. by 13 in.). These panels of vertical grain bevel siding were weathered for 0, 1, 2, or 4 weeks in an accelerated weathering chamber. Then, the panels were painted with two coats of alkyd primer or latex primer and cured. A piece of hard maple was glued with epoxy to the surface of the paint, and the panels were equilibrated to 12% moisture content. Each panel was then cut to yield seven shear block and three UDCB test specimens. The shear block specimens were tested using a modification [1 ] of ASTM D 905. Before testing, white typing correction fluid was applied to the edge of the UDCB specimens (15.9 mm by 304.8 mm (0.625 in. by 12 in.)) to aid visual determination of crack growth. Crosshead speed of the test instrument was 3.6 mm/min (0.14 in./min), and the load was manually recorded at each data point (25.4mm intervals) as the crack propagated along the paint/wood interphase.

a.

b.

c.

Computer-Aided Experiment Grain angle greatly affects modulus of elasticity (MOE) of the specimen and thus can affect the results of the test (Fig. 1). Grain angle must be controlled to direct the crack along the bond line and to maintain a consistent MOE throughout all tests. Although the fracture toughness equation takes the MOE into consideration, similar MOE values can simplify the calculation. To obtain consistent grain angle, 19-mm- (0.75-in.-) thick boards were cut from 152-mm by 305-mm by 3.66-m (6-in. by 12-in. by 12-ft) western redcedar timbers. Grain angle was determined on the tangential surface, and parallel lines for cutting boards were drawn at a 5.7° angle (10:1 1ength:width) to the grain angle. The 152-mm-wide by 19-mm-thick (6-in.-wide by 0.75-in.-thick) boards were sawn from five different timbers (Fig. 2a). The boards were brought to equilibrium moisture content and planed on one side,

d. FIG. 2—For consistent grain angle, 19-mm- (0.75-in.-) thick boards were cut from five 152-mm by 305-mm by 3.66-m- (6-in. by 12-in. by 12ft) western redcedar timbers (a). This board was cut into two edge-matched panels (b). These panels were planed and cut to 324 mm (c); paint was removed from edge of panel, and panel was glued to its mate (edge-match pair). Four days later, three replicate specimens (d) were cut from each panel.

274

JOURNAL OF TESTING AND EVALUATION variable displacement transformer (LVDT) connected to the moving beam of the testing machine. The load was about 20 kg (44 lb) for crack initiation and about 18 kg (40 lb) when the first electrical contact opened. As the crack propagated, the load decreased. At a load displacement of about 23 mm (0.9 in.) and a load of about 3 kg (6 lb), the last contact opened. Results and Discussion Wood, being an anisotropic material, differs from other materials used in fracture toughness tests. The MOE of wood is about ten times greater in the longitudinal (parallel to the grain) than in the radial direction, and about 20 times greater than in the tangential direction [9]. The MOE for any grain angle (Eα ) can be estimated using the Hankinson equation (1)

FIG. 3—Specimen after notch was cut in the end of the specimen so that it could be attached to test machine.

ripped down the middle to give two 70-mm- (2.75-in.-) wide panels. and labeled to show edge-matched pairs (Fig. 26). Of the two 70-mm panels, one was saved and planed just prior to being glued to the painted surface of its mate. The other 70mm panel was planed to give a thickness of 12.7 mm (0.5 in.) and cut to 356 mm (14 in.). An alkyd primer and exterior flat latex topcoat were brush applied, and the panel was exposed to accelerated weathering. These panels were not weathered before painting as were the panels in the preliminary experiment. Following weathering, the panels were cut to 324 mm (12.75 in.), cutting off both ends. A small strip of paint was removed (shallow rabbet) from the starting end of each panel and replaced with common clear tape to assist in the crack initiation. The edge-matched panels previously ripped from the original board were planed to 12.7 mm (0.5 in.) and cut to 324 mm (12.75 in.) (Fig. 2c). The freshly planed surfaces were glued to the paint surface using Epon 8282 and curing agent Epon V-40. Three hours after the epoxy was mixed, the panels were placed in a press at 0.14 MPa (20 lb/in.2). Four days later, three replicate specimens 16 mm by 305 mm (0.625 in. by 12 in.) were cut from each panel (Fig. 2d). A 3-mm- (0.125 in.-) wide by 13-mm- (0.5in.-) deep notch was cut in the starting end of the specimen so that it could be attached to the test machine (Fig. 3). Narrow strips of conducting paint (Nickel Print. GC Electronics, Rockford, Illinois’) were painted on the sides of the specimens across the wood/paint/wood interphase at 25.4-mm (1-in.) intervals so that the crack propagation could be monitored electronically. Electrical contacts were attached to the sample (Fig. 4) and connected to a computer. Test Machine

and (2) where EL is MOE in the longitudinal direction, Er is MOE in the radial direction, Et is MOE in the tangential direction, and α is angle. Figure 1 shows the relationship of grain angle to MOE. A painted, flat-grain surface is represented by points on the tangential curve and those on a vertical grain surface on the radial curve. It was necessary to choose a grain angle large enough to carry a crack to the interphase but not large enough to weaken the sample. Grain angles of 3° to 10° appeared to be reasonable choices. Angles greater than 10° would weaken the specimen too much. A simple 10:1 length to width ratio. 5.7°, was chosen. The MOE was on or between the two curves at 5.7°. Vertical and flat grain can also affect MOE. Vertical-grained wood, which has a high MOE, is low in tension perpendicular to the grain strength. Wood splits more easily in the radial direction than in the tangential direction. Excessive strain resulted in wood failure in several of the purely vertical-grained specimens. Flat-grained specimens have a surface less suitable for painting than do vertical specimens; therefore, best results were obtained from specimens slightly off the vertical grain. The test results for shear block were derived from the dimensions of the specimen and the maximum stress needed to break the bond. The test results for UDCB specimens were more complex. Besides dimension and stress, crack length and test machine displacement were needed for the derivation. Fracture toughness (G 1 c ) for Mode I failure was calculated using the formula

The specimen was attached to a test machine using specially prepared stirrups (Fig. 5) and pulled in tension at 3.6 mm/min (0.14 in./min). Load displacement was measured using a linear where -The use of trade or firm names in this publication is for reader information and does not imply endorsement by the U.S. Department of Agriculture of any product or service.

P is force, a is crack propagation distance, h is thickness of the specimen,


FIG. 4—Electrical

contacts

rest

on

conducting

275

paint.

fracture toughness formula can be reduced to (6) The opening mode fracture toughness was calculated for each of the ten points along the specimen. The moment (M ) for each point was also calculated and used to refine the data (explained later). M = Pa

(7)

Preliminary Comparison Experiment F I G . 5 —Stirrup used to connect specimen to test machine.

b is width of the specimen, and E is MOE. Substituting (4) where ∆ = 1/2 load displacement Then, (5) Equation 3 has a shear component that adds 8% to the data at 25.4 mm (1 in.), decreases geometrically to 2% at 50.8 mm (2 in.), and is insignificant beyond 50.8 mm (2 in.), so that the

Results of the shear block tests are given in Table 1 and Fig. 6; results from the UDCB are given in Table 2. The COV for specimens using the shear block method was about 15%, and the COV for matching UDCB specimens was about 10%. Results for the UDCB specimens are a measure of the fracture toughness of the wood/paint interphase. Although better than the COV from the shear block tests, the COV of 10% for the UDCB test was high. This is because the grain angles of the UDCB specimens were somewhat random and propagation was determined visually. The preliminary test clearly showed better precision with UDCB specimens than with shear block specimens, even when random grain angles were chosen for the UDCB specimens. Both tests showed a dramatic effect between the control and short exposure. Longer exposure caused only a slight increase in ease of paint removal. The 15% COV for the shear block specimens may be acceptable; however, determining minor damage would be difficult. Data from the preliminary UDCB test are given in Fig. 7. Two corrections should be factored into the fracture toughness Eq 6. First, the measured a distance in fracture toughness and moment was smaller than the actual a distance. This distance is from the load center to the axis of rotation (beyond measured crack front). Also, any flexibility in the material (white typing correction fluid or conducting paint) used to aid in determining crack front position would indicate the crack front short of the actual

276

JOURNAL OF TESTING AND EVALUATION

TABLE 1—Shear block data from the preliminary experiment. a

Exposure (week)

Mean (MPa)

COV (%)

0

809 787 798 805 805 721 599 599 544 539 518 564 392 374 403 319 433 420 270 279 334 236 281 267

4.1 6.7 14.0 8.2 13.2 6.6 8.2 7.3 17.3 28.0 12.6 11.0 13.1 13.0 17.2 4.5 24.1 16.1 12.5 15.4 11.6 23.9 11.6 22.5

1

2

b

c

TABLE 2—UDCB data from the preliminary experiment. d

Mean (MPa)

COV (%)

787.3

10.4

Exposure (week) 0 1

560.9

15.8

2

390.3

19.2

4

277.7

19.4

a

Mean of seven replicates cut from the same panel. b COV of seven replicates cut from the same panel. c Mean of all specimens with the same exposure. d COV of all specimens with the same exposure.

crack front. In any case, the measured a distance was too small. As explained later, the distance to the apparent axis of rotation was found to be 20 mm (0.8 in.) beyond the measured crack front in the preliminary test. Second, zero point for load displacement is not known until all machine slack and flexibility are considered. Zero point was

COV (%) for data points (in.) 4 to 9

5 to 9

6 to 9

7 to 9

9.7 5.2 12.6 11.8 10.4 19.6 14.3 10.4 10.0 21.2 13.6 8.6 10.4 13.2 8.3 6.3 17.5 10.5 10.4

9.0 4.2 12.9 10.4 12.1 21.7 10.5 9.9 8.3 16.1 13.7 7.8 10.0 9.4 8.9 5.6 15.0 10.9 10.1

7.1 3.0 13.0 7.9 10.7 20.8 10.7 10.6 9.2 13.2 12.8 7.5 8.9 10.1 8.4 5.8 15.9 10.5 10.8

7.4 3.5 9.3 7.1 2.7 20.4 10.2 8.8 8.8 12.9 14.4 6.9 10.1 7.3 5.8 5.5 17.3 7.7 9.4

arbitrarily set at a 4.5-kg (10-lb) load. We found empirically that adding only 1.27 mm (0.05 in.) to the load displacement in the preliminary UDCB experiment, as shown in the following paragraph, improved the data accuracy. The raw data for the preliminary test included the moment, MOE of the adherend, and fracture toughness (G 1 c ) of the paint wood bond (Fig. 7, three left graphs). These values were calculated from Eqs 7, 4, and 6, respectively. First. the a distance is corrected by substituting successive approximations for a (+ 20 mm) in Eq 7 to produce the most level plot for the moment. The MOE values of the adherends and G 1 c were recalculated and are given in Fig. 7 (three middle graphs). The MOE and

FIG. 6—AN data points from preliminary shear block test with mean (M), median (m), and boundaries for 50 and 75% of the data.


277

278

JOURNAL OF TESTING AND EVALUATION

G 1 c should be the same at any point on the specimen; therefore, the plot of MOE should be level. (Note that the MOE and G 1 c are only partially improved.) In a similar fashion as the a correction, the load displacement was then corrected using Eq 4 to give the most similar MOE values throughout the specimen. Using the corrected a and corrected load displacement, G1 c was recalculated using Eq 6. Computer-Aided

length and 0.6 mm (0.025 in.) to correct for load displacement. The corrections were determined in the same manner that we followed in the preliminary comparison experiment. The effects of this corrective procedure (partially shown in Fig. 8) were more apparent than in the preliminary experiment (Fig. 7). The COV would be further decreased by including only data from the center of each specimen (generally points 7, 8, and 9). The COV of data points from different groupings was as follows:

Experiment

Propagation Data Points 4 to 9 5 to 9 6 to 9 7 to 9

The MOE and fracture toughness results for one exposure for the UDCB specimens having consistent grain angle are given in Table 3 and Fig. 8. A computer was used to collect the data. The COV of the fracture toughness data for these tests was decreased to 3% by adding 13 mm (0.5 in.) to correct for crack

COV % 3.2 3.3 3.4 1.7

TABLE 3—UDCB data from the computer-aided experiment. Fracture Toughness (J/m2) at 25.4-mm (1-in.) Increments Along Specimen Replicate

1

2

3

4

5

6

7

8

9

10

1 2 3

0 70.37 — 24.5

216.3 240.1 235.3

240.8 237.8 234.2

225.2 236.1 238.8

230.0 242.8 240.1

232.0 248.3 240.0

230.2 240.4 239.1

226.6 226.4 228.5

230.4 218.5 228.0

213.3 218.3 210.8

F I G . 8 —UDCB

data

from

computer-aided

experiment.


In the computer-aided experiment, the load displacement of the specimen was limited by the LVDT. It may be possible to further improve the precision by making the specimens thinner. Larger range LVDTs would permit thinner specimens and possibly decrease the COV by increasing the accuracy of the load displacement measurement. The preparation of specimens from panels having consistent grain angle and the improved procedure to measure crack length using a computer were the main factors that contributed to improving the precision of the test. When MOE was calculated for the specimens with consistent grain angle (all western redcedar) using the data at each point with both corrections, the result was a MOE of 6 GPa (0.9 × 109 lb/in.2) (Fig. 8). This agrees with the estimated MOE derived from the Hankinson equation. The fracture toughness results for one exposure after corrections of the experiment with specimens cut from timbers with a consistent grain angle are shown in Fig. 8 and Table 3. Conclusions

Determining fracture toughness using the uniform doublecantilever beam (UDCB) specimen with the aid of a computer is an excellent technique for determining paint adhesion to wood. The COV for the shear block test was about 15%. The data from UDCB specimens cut from the same panels have COVs of 10%. When the grain angle was closely controlled and the crack propagation monitored by computer, a COV of less than 5% was obtained. This successive approximation technique mathematically solves the problems involved with cantilever tests with little difficulty. Other techniques using contoured or tapered cantilever specimens may introduce error by the addition of other materials. Using reference beams is not advisable with wood because of the inconsistency of the substrate.

279

Acknowledgments

We thank Gary Larson for writing the computer program, Steve Hankel for making the electrical contacts and stirrups. and Earl Geske for making the computer hardware. References