The specimen

The specimen V2 (Figure 9.22.1) has the outline of the tensile bar used for material testing on the same machine, cut in two. Each half ends in a tongue half as wide as the bar, and the tongues overlap side by side. Only the face between the two fillets is glued, 24 mm long and as high as the print is thick, so the glue is loaded along its plane, in shear. The fillets and the end faces stay dry.

Download the geometry, Provstav_V2.step, from the Lim-labb page on Canvas, and print both halves flat on the bed with 100 % infill. The halves fit together without a gap in CAD, so the glue line is as thin as the print lets it be.

Glues and safety

Two glues are approved for the workshop: a cyanoacrylate (Biltema superlim) and a construction adhesive (Biltema bygglim). Construction adhesives of modified silicone or MS polymer, such as Tec7 or Biltema Multibond, are interesting too. Epoxy needs the thermoset training and is not supported in the lab. Victor knows which glues the workshop supports and where to glue; ask him, or Mirza, when in doubt or for help with the gluing.

Gluing

1. Measure the bond face of every half with a caliper: its length \(L\) and its height \(b\).
2. Wipe the bond faces clean; the glue holds only to a surface free from grease [1]. Sanding the faces is a variable to test: on printed ABS it gave stronger joints than cleaning alone [2].
3. Put glue on the bond face of one half only, not on the fillets or the end faces.
4. Press the halves together and hold them aligned until the glue has fixed, a few seconds for cyanoacrylate [1] and hours for a construction adhesive [3]. Wipe off glue that squeezes out onto the fillets.
5. Let the glue cure as its maker says before pulling: 24 to 72 h for cyanoacrylate [1]; a construction adhesive cures from the air inwards, about 6 mm a day for Tec7 [3]. How long the glue needs to reach its full strength is itself a question the machine can answer.
6. Write down for every specimen: the glue, the surface preparation, the date and time of gluing and of testing, and the printer and filament.

A butt joint, glued end to end, is a different load case and can be tested as well: which is stronger?

The tensile test machine

Book the machine in the booking schedule. Ask Mirza for a walkthrough of the machine; groups are gathered for it, and before it you watch the recording of Figure 9.22.2.

The machine is run from the program UTM Control on your own laptop (Windows 10 or 11). To install it, open PowerShell (Start menu, type powershell) and paste this line:

irm https://raw.githubusercontent.com/cenmir/UTM/main/deploy/get.ps1 | iex

It downloads the program, sets up Python and puts an icon, UTM Control, on the desktop. Two USB cables go to the laptop, one for the camera and one for the machine. The camera needs a USB 3 port, the blue one, and does not work through a hub. Double-click UTM Control to start.

The program saves each test as a CSV file whose comment lines start with # (Table 9.22.2; UTM repository).

Procedure

1. Clamp the grip ends of the glued specimen in the machine, the specimen straight.
2. Enter the bonded area you measured, \(bL\), in the program.
3. To measure how far the joint slides before it lets go, put two small marker dots on the specimen, one on each side of the overlap, and calibrate their distance in the program. Without the dots the crosshead position measures the joint, the specimen and the machine together.
4. Pull to failure at a constant speed, and note the speed.
5. Photograph both fracture faces and note where the joint failed: in the interface (bare plastic on one face), in the glue (glue left on both faces), or in the plastic, in a tongue or at a fillet.
6. Save the CSV with the glue and the specimen number in its name.

Test at least three specimens per glue: printed parts scatter, and one curve says little.

What the theory predicts

Divided by the bonded area, the largest force gives the nominal shear strength of the joint,

\[ \tau_\text{nom} = \frac{F_\max}{bL} \tag{9.22.1}\]

the value a data sheet reports from a lap-shear test. Two things make the true load case differ from that uniform shear.

The first is the offset. Near the end of the glue line one tongue carries all of the force, along its own centre line, \(e = t/2 = 2.5\) mm from the line of the pull (Figure 9.22.1 (b)). The tongue is then bent by \(Fe\) as well as pulled, and the same moment opens the end of the glue line in peel, as in a single lap joint (Adhesive joints). In the tongue’s section \(t \times b\) the largest stress is

\[ \sigma_\max = \frac{F}{bt} + \frac{6Fe}{bt^2} = \frac{F}{bt}\left(1 + \frac{6e}{t}\right) \tag{9.22.2}\]

The clamped grips hold back part of the moment, so this is an upper bound. With the yield strength of PLA printed flat, 51 MPa [4], the tongue yields at

b, L, t = 8.0, 24.0, 5.0        # print thickness, length of the bond face, tongue width [mm]
A_bond = b*L                    # bonded area [mm^2]
e = t/2                         # offset of a tongue's centre line from the line of the pull [mm]
k_tongue = 1 + 6*e/t            # peak stress in the tongue in units of F/(b t), @eq-gl-tongue
R_e_PLA = 51.0                  # yield strength of PLA printed flat [MPa], Prusament data sheet
F_tongue = R_e_PLA*b*t/k_tongue # force at which the tongue yields next to the glue line [N]
tau_tongue = F_tongue/A_bond    # the nominal shear stress at that force [MPa]

A glue line that holds more than about 2.7 MPa nominal shear therefore does not break first: the tongue does, and the test then measures the PLA and the offset, not the glue. Where the specimen broke tells which of the two we measured.

The second is shear lag. By the shear-lag model, the glue line carries load over a length of about

\[ \frac{1}{\omega} = \sqrt{\frac{h\,E\,t}{2G}} \]

at each end, with the tongue width \(t\) as the adherend thickness, the modulus \(E\) of the print, and the thickness \(h\) and shear modulus \(G\) of the glue line (9.20.8). Neither glue states its shear modulus, so we compare a rigid glue, with the \(G = 1\) GPa assumed for the epoxy of Example 4, with the construction adhesive, \(G \approx 0.67\) MPa from the modulus of Tec7 [3]. The glue line thicknesses are guesses; measure yours on a broken specimen.

E_PLA = 2300.0                  # modulus of PLA printed flat [MPa], Prusament data sheet
glues = {                       # G [MPa], h [mm]: assumed, replace with your own values
    'rigid glue': (1000.0, 0.1),
    'construction adhesive': (2.0/3, 0.5),
}
lag = {}
for name, (G, h) in glues.items():
    w = np.sqrt(2*G/(h*E_PLA*t))           # @eq-pj-volkersen-ode [1/mm]
    u = w*L/2
    lag[name] = dict(inv_w=1/w, peak=u/np.tanh(u), used=np.tanh(u)/u)

For the 24 mm bond face of the specimen the two glues give

A rigid glue loads the first millimetre at each end of the face, peaks at about sixteen times the nominal stress there, and fails from the ends; its \(\tau_\text{nom}\) belongs to this specimen and not to the glue. The construction adhesive loads all of the face evenly, so its \(\tau_\text{nom}\), peel at the ends aside, comes close to the shear strength of the glue.

Analysis with Python

The files in assets/gluelab are fictive recordings in the format of UTM Control, three specimens glued with cyanoacrylate (ca_*.csv) and three with construction adhesive (construction_*.csv), made so that the cells below run before you have data. Replace them with your own files. We read every file in the folder, skipping the comment lines:

folder = pathlib.Path('assets/gluelab')
runs = {f.stem: pd.read_csv(f, comment='#') for f in sorted(folder.glob('*.csv'))}
colour = {'ca': RED, 'construction': BLUE}   # one colour per glue, from the file name

The force against the crosshead position shows the two kinds of joint at a glance (Figure 9.22.3).

From each curve we take the largest force, the nominal shear stress by 9.22.1 with the area of Table 9.22.1 (use your measured \(bL\)), and the position at the largest force:

res = pd.DataFrame([{'specimen': name, 'glue': name.split('_')[0],
                     'F_max': d['Force_N'].max(),
                     'u_at_F_max': d['Position_mm'][d['Force_N'].idxmax()]}
                    for name, d in runs.items()])
res['tau_nom'] = res['F_max']/A_bond         # @eq-gl-tau-nom [MPa]
summary = res.groupby('glue')['tau_nom'].agg(['mean', 'std', 'count'])

The fictive cyanoacrylate joints reach about 3.2 MPa nominal shear, above the 2.7 MPa at which 9.22.2 lets the tongue yield, so in a real test of such a glue look for a broken tongue before believing the number.

Questions

Work as a group, and let curiosity steer what you use the machine for. Answer with your measurements, your photographs and the theory above.

1. Where did each specimen fail: in the interface, in the glue, or in the plastic? Do we measure the glue?
2. What nominal shear strength does each glue reach, and how large is the spread between specimens?
3. How does the real load case differ from pure shear? Compare \(F_\max\) with the force at which 9.22.2 lets the tongue yield, and look for signs of peel at the ends of the glue line.
4. Estimate \(1/\omega\) for your glue with the thickness of the glue line you measured on a broken specimen. Would a 48 mm long bond face carry twice the load?
5. How far does the joint slide before it lets go? Use the marker dots on each side of the overlap.
6. Does the pulling speed change the result?
7. What do the data sheets say? Loctite 401, a cyanoacrylate, gives 7.5 MPa lap-shear strength on ABS [1]; Tec7 gives a tensile strength of 2.8 MPa after 7 days [3]. How do your values compare, and why might they differ?
8. How long does the glue take to reach its full strength?
9. Is a butt joint stronger or weaker than the lap?
10. When is each glue useful for a printed part, and when would you choose a screw instead?

Presentation

When you are done, tell us that you want to present. Give a short, simple presentation of what you found; we ask curious questions. A sincere attempt passes. Present well before the course ends, not in the exam week.