Research · Test planning

Predict, then test

My thesis tested oak and poplar LVL on 400 mm specimens. The next step is a beam. This page takes the measured board properties, predicts how a beam of that layup will behave, and lays out the full-scale bending test that would confirm the prediction or show where it is wrong.

Measured means measured on pressed boards at the University of Sopron. Predicted means calculated here and not yet tested at beam scale. Nothing at beam scale on this page has been tested.

1. Choose a beam

Span
3.60 m

Four-point bending with the load points at one third of the span, laminations stacked across the depth (the way an LVL beam is used). A span of 18 times the depth is the usual standard geometry for timber bending tests (EN 408); check the current edition before using it.

2. The test setup, to scale

actuatordisplacement sensor (mid-span)strain gauges: top, mid-depth, bottomspan 3600.0 mmL/3L/3
section, neutral axis dashed

oak veneer poplar veneer

Layer thickness keeps the 1.4 mm oak to 2.8 mm poplar ratio of the pressed boards, scaled up to the depth you chose. A deep beam would use more plies; the proportions are what matter here.

3. What the model predicts

Bending stiffness EI794.7 kN·m² Predicted
Effective modulus of the section11921 MPa Predicted
Failure load (both load points together)87 to 120 kN Predicted
Weak pointOuter tension or compression fibre of the weaker species reaching its strength

The failure range runs from no size effect (upper) to a size exponent of 0.12 on depth, the value Eurocode 5 uses for LVL (lower). Which one holds for oak and poplar hybrids is exactly what the beam test would show. Shear deformation uses an assumed E/G ratio and a shear correction factor of 5/6; neither has been measured for these boards.

Change the assumptions

4. What to read, and what would mean the model is wrong

Hold the first stage at 40% of the lower failure estimate, 34.8 kN in total, where the beam should still be linear.

ReadingPredicted at that loadModel is in trouble if
Mid-span deflection, relative to the supports37.97 mm Predicted (96% bending, 4% shear)it is off by more than 20%. The thesis model was within 9.2%, and board stiffness varied by up to about 6%.
Strain, bottom face (tension)2630 µε Predictedit differs from the top-face strain by more than 15%: the neutral axis is not where the model puts it, so the layers are not sharing load as assumed.
Strain, top face (compression)2630 µε Predictedas above.
Strain, mid-depthclose to zero (shifts slightly toward the stiffer face if the layup is not symmetric)it reads a large value: the load is off-centre or the beam is twisting. Stop and re-seat.
Support settlementsmall and equal at both endsone end moves much more: bearing crushing or a tilted support. Correct the deflection for it.
Failure load (both points)87 to 120 kN Predictedit fails below the lower value: look at shear, bearing or lateral buckling before blaming the bending model. Above the upper value: the model is too cautious.
  • Test at least five beams per layup so a failure range means something. The thesis used eight specimens per board.
  • Restrain the beam sideways near the load points, and use bearing plates at the supports.
  • Record load and every sensor continuously, so the stiffness comes from a fitted slope and not from one point.

5. The prediction rule, checked on data I already have

The beam model above uses only the oak and poplar control boards. Here is the same rule applied to the thin flatwise coupons that were also tested in the thesis, with the layers stacked across the depth as in a beam. This is the closest check available before a real beam test.

BoardMeasuredPredictedDifferencen
P169.9 MPa Measured68.1 MPa Predicted-3%3
P258.9 MPa Measured68.1 MPa Predicted+16%3
TO1123.0 MPa Measured114.7 MPa Predicted-7%3
L192.1 MPa Measured92.8 MPa Predicted+1%3
L2123.8 MPa Measured109.2 MPa Predicted-12%2
L3120.2 MPa Measured107.8 MPa Predicted-10%3

Four of the six land within about 10% and the oak-rich boards come out a little low, so the rule is cautious there. The poplar board P2 comes out 16% too high because the two poplar boards differ from each other by about 30% in edgewise strength, and the rule uses their average. With only two or three coupons per board, this is a hint and not a proof, which is the reason to test a beam. The order of oak and poplar plies in L1, L2 and L3 is the one shown on the layup studio page.

Source: M.Sc. thesis, Timber Industry Engineering, Wood Engineering and Creative Design department, University of Sopron, 2024. Edgewise specimens 400 × 19 × 14 mm on a 360 mm span, n = 8 per board. Flatwise coupons n = 2 or 3. This page is an engineering example, not a design tool for real structures.