Simply Supported Beam Deflection

Getting Practical with Simply Supported Beam Deflection

A simply supported beam has a pin at one end and a roller at the other, meaning it can rotate freely at both supports and cannot take moment at either end. This is the baseline condition everyone learns in the first structural mechanics class. The formulas that come with it are straightforward, but they cover only a narrow slice of real-world situations, and treating them as gospel will cost you time and sometimes money. The two most common cases are a point load at midspan and a uniformly distributed load across the full span. For the point load case, the maximum deflection is P times L cubed divided by 48 E I. For the uniform load case, it is 5 w L to the fourth divided by 48 E I. Here E is Young's modulus, I is the second moment of area about the bending axis, L is the span, P is the point load, and w is the distributed load per unit length. The location of maximum deflection for the point load case is exactly at midspan. For the uniform load case it is also at midspan. For an eccentric point load, the peak shifts slightly away from center, but the difference is small and the midspan value is usually close enough for preliminary design. Slope at the left support for a midspan point load is P L squared divided by 16 E I, and the slope at the right support is the same magnitude with opposite sign.

I tend to work these by hand first because it keeps the numbers honest. If your first result looks too clean, you probably made a unit mistake.

Quick worked example

Take a W12x50 steel beam, span of 10 meters, with a 15 kN point load at midspan. The section property from the AISC manual gives I about 78.0 times 10 to the sixth millimeters to the fourth. Young's modulus is 200,000 megapascals. Converting the span to millimeters, L equals 10,000 mm, and L cubed is 10 to the twelfth. Plugging in, the deflection comes out to approximately 15.6 millimeters. The stress check on this same beam under that load would be the main concern, not the deflection, since 15.6 mm is roughly L over 640, which is below most code limits. But you still need to verify rotation at the supports if something attached to the beam end is sensitive to slope. For the uniform load case on the same beam with a total load of 30 kN spread evenly, w equals 3 N per millimeter, L to the fourth is 10 to the sixteenth, and the deflection works out to about 19.5 millimeters. Different load pattern, different formula, similar magnitude. That is the kind of thing you want to see immediately rather than after you have already sized the wrong section.

Where the simple formulas break down

The elastic deflection formulas assume small deformations, linear material behavior, and that the beam is long and slender relative to its depth. They also assume the supports do not move. When any of those assumptions go out the window, the standard Simply Supported Beam Deflection equations give you answers that are wrong, and sometimes quietly wrong in a way that looks plausible. Thick short beams are the first place this shows up. If the span to depth ratio is below about 10, shear deformation starts to matter. Timoshenko beam theory adds a shear correction term, and for a rectangular section the additional deflection from shear is roughly 3 P L divided by 10 G b h for a midspan point load. G is the shear modulus and b and h are the section width and depth. In practice, this adds a few percent for a steel wide flange, but for a short timber beam or a deep concrete slab, it can add ten or fifteen percent, and ignoring it makes your deflection prediction unconservative. Large deflections are the second trap. Once the deflection exceeds about L over 250, the geometry changes enough that the linear relationship between load and deflection no longer holds. The curvature assumption behind the derivation is violated. This rarely matters for steel beams in ordinary buildings, but it matters for flexible composite decks, long-span timber glulam, and some cable-supported systems. If you are working near that threshold, use a nonlinear analysis or a finite element model that updates geometry during solution.

Support settlement is a real problem

I learned this the hard way on a renovation project where an existing W18x50 beam was supporting a new masonry wall. The engineer specified a simple span of 9 meters and calculated deflection using the standard formulas. The beam looked fine on paper. What nobody had measured was that the left support, a steel column on a concrete pier, had settled about 12 millimeters relative to the right support during the previous five years. The settlement alone introduced a rigid-body rotation that changed the load path enough to increase the midspan deflection by roughly 4 millimeters compared to the idealized calculation. The beam did not fail, but the masonry above it cracked because the rotation was transferred directly into the wall. The workaround was to measure the existing support elevations before any strengthening, model the beam with the actual support positions including the differential settlement, and then decide whether to shim the low support, add a jack post, or redesign the connection. Simple formulas cannot account for this because they assume both supports are at the same level and do not move. If your structure sits on anything that can settle, you need field measurements and a model that includes those movements.

Section properties matter more than you think

Using the wrong I value is a surprisingly common mistake. The AISC manual lists Ix for the strong axis, but if your load is applied weak-axis, you need Iy, and the difference between the two for a W12x50 is enormous, about 78 million versus 50.6 million millimeters to the fourth. Using the strong-axis value when the beam is loaded weak-axis halves your predicted deflection, which is a dangerous error. Another issue is that some textbooks and online calculators list I in inches to the fourth while others use millimeters to the fourth. If you mix units without converting, the result will be off by a factor of about 41.6 million. I always write the units next to every number as I substitute, and I do a quick sanity check: a steel beam in a typical office building should deflect somewhere between L over 360 and L over 600 under service loads. If your number is outside that range, check the units first, then check the section property, then check the load magnitude.

What I actually do instead of trusting the textbook formula

For quick screening I still use the closed-form equations. They take about ten seconds and tell you whether the concept is viable. For anything that goes into construction documents, I run the model in structural analysis software with the actual boundary conditions, support stiffness, and member sizes. The software accounts for the exact load positions, any intermediate supports, varying cross-sections, and temperature effects if I include them. I also check deflection limits from the applicable code, because the allowable depends on the finish and the occupancy. For a steel beam supporting gypsum board ceiling, L over 360 is typical. For a beam supporting brittle masonry, you might need L over 600 or stricter. A deflection that is fine for a flexible finish can crack a rigid one.

Pitfalls beginners miss

The first is forgetting that the Simply Supported Beam Deflection formulas give elastic deflection only. They do not include creep, shrinkage, or long-term load effects. For timber, the long-term deflection can be two to three times the immediate elastic deflection depending on the species and moisture conditions. I multiply the elastic result by a creep factor from the relevant timber design standard instead of hoping the code will catch it. The second is assuming the beam is truly simply supported. A connection that looks like a simple shear tab can develop enough rotational restraint to change the moment distribution. If the connection is stiff, the beam behaves closer to a fixed-fixed or propped cantilever, and the midspan deflection under uniform load drops by about a third compared to the truly simply supported case. I check the connection stiffness during detailing and model the actual restraint, not the idealized pin. The third is ignoring self-weight. For long spans or heavy sections, the beam's own weight is not negligible. I always add it to the load before computing deflection, even though it is obvious, because when you are juggling ten different load combinations it is easy to forget.

When to switch methods entirely

If your beam has a varying cross-section, an intermediate support, an overhang, or a non-uniform load, the standard midspan formulas do not apply. Use the moment-area method, conjugate beam method, or virtual work for hand calculations. These take longer, maybe twenty to thirty minutes per case, but they give the correct answer without relying on a formula that does not fit the geometry. For complex structures, a finite element model is faster overall because setting up the hand calculation for each case takes more time than running the model and interpreting the results. I keep a small spreadsheet with the basic formulas pre-populated so I can run quick checks while the software model is setting up. The spreadsheet covers point load at midspan, point load at arbitrary position, uniform load, triangular load, and uniform moment. It also flags when the span-to-depth ratio is low and reminds me to check shear deformation. This cuts the preliminary check from an hour of fiddling with units and formulas down to about five minutes, and it catches most of the common mistakes before they become expensive ones.

Summary of what actually works

Use the standard elastic formulas for preliminary sizing and sanity checks. Verify the section property and the axis of bending. Check units explicitly. Account for support settlement when the structure sits on variable foundations. Include self-weight. Apply the correct deflection limit for the finish. For anything beyond the basic cases, use moment-area methods or a structural model, and always check the assumptions behind whatever equation you are using.