When you look at a factory roof covered in solar panels, it reads as light, thin modules lying flat, barely there. That picture is misleading. A rooftop plant is a permanent structural addition to a building that was very likely designed and built years before anyone thought about solar. And the load that matters most is not the one you can see.
There are two forces at play, and proposals that skip the second one are the ones that get roofs into trouble.
Dead load is the easy part. Wind is the one that bites.
A mounted array adds a modest, permanent dead load, indicatively in the region of 12–15 kg per square metre once you count modules, rails, clamps and mounting structure. Most industrial roofs carry that comfortably. If dead load were the whole story, structural checks would be a formality.
The force that actually governs the design is wind uplift. Wind flowing over a low-tilt array does not just push, it lifts, the same way it lifts an aircraft wing. On a shallow-pitched rooftop array that suction can be substantial, and it acts upward, trying to peel the panels, the rails and sometimes the roof sheet itself off the building. It is dynamic, it concentrates at edges and corners, and it is exactly the load a cheap proposal has no incentive to model honestly.
In India, this is not guesswork. IS 875 Part 3 is the code that governs wind loading on structures, and it is what a competent structural engineer works to. It starts from a basic wind speed assigned by region, coastal Tamil Nadu, for instance, sits in a higher-wind zone, indicatively around 50 m/s, and then adjusts for terrain category (an open coastal plot behaves very differently from a site tucked among taller buildings), height, topography and the shape of what you are mounting. Two identical arrays on two different sites can face very different design loads. Naming the standard is the easy bit; doing the calculation for your roof is the work.
The panels are light. The uplift is not. And the roof you already own is what has to hold both down.
Where the load actually lands, and where it fails
Uplift does not stop at the panel. It travels down a chain: module to clamp, clamp to rail, rail to purlin, purlin to truss, truss to column. Every link in that chain has to carry the force, and the array is only as safe as its weakest one. On a pre-engineered building (PEB), the purlins and truss spacing were sized for the original roof, sheet, some services, a design wind load, with no allowance for rows of solar catching wind. Add an array without rechecking that chain and you can quietly overload members that looked perfectly adequate on the original drawings.
Fasteners are the other quiet failure point. On a GI-sheet roof, everything ultimately depends on the pull-out strength of screws driven into thin metal, and those fasteners age, corrode and loosen. A joint that held years ago may not hold the new uplift today. On an RCC roof with a non-penetrative ballasted system, the equivalent question is whether there is genuinely enough ballast to resist lift-off across the whole array, including the edges and corners where suction peaks.
Common failure modes
- Overloaded purlins on a pre-engineered building that was designed and built without solar in mind.
- Corroded or aged GI-sheet fasteners that have lost pull-out strength, so the roof-to-array connection can no longer resist uplift.
- Inadequate ballast on an RCC roof, letting wind lift the array off, worst at the exposed edges and corners.
- Missing or ignored expansion joints, so thermal movement of a long array builds stress into the structure and the fixings.
That last one is easy to overlook. A long steel array heats and cools every day and needs room to move. Skip the expansion joints and thermal stress has nowhere to go but into the members and the roof, a slow load that shows up as fatigue and loosened fixings, not a dramatic failure.
Why the structural check comes before the price
Here is the part that separates a real EPC from a vendor selling watts. The structural assessment has to happen before the commercial proposal, not after it. If a bill of materials and a payback figure land on your desk before anyone has walked your roof, taken your building drawings, checked purlin spacing and fastener condition, and run the IS 875 numbers for your terrain, then the price is a guess, and it is a guess that has quietly assumed your roof is fine.
When the survey comes first, one of two honest things happens. Either the roof is confirmed capable and the design is built around real numbers, or the assessment flags remedial work, additional purlins, fastener replacement, more ballast, and that cost sits inside the proposal where you can see it. What you never want is to discover the reinforcement bill after signing, or worse, after the first serious monsoon squall.
Ballast vs anchored: a real trade-off, and a warranty question
On flat RCC roofs you usually have a choice. A non-penetrative ballasted system holds the array down with weight and never puts a hole in your waterproofing, which is exactly why many facility owners prefer it, because it keeps the roof warranty clean. The trade-off is that it only works if the ballast is genuinely sufficient for the design uplift; under-ballast to save weight or cost and you have engineered in a lift-off risk. An anchored (penetrative) system ties directly into the structure and resists uplift very securely, but every penetration is a potential leak and may affect the roof warranty unless it is flashed and detailed correctly. Neither is universally right. The correct answer falls out of the wind numbers, your roof type and how the warranty is worded, which is one more reason the engineering has to lead.
A rooftop plant is meant to sit on your building for twenty-five years and quietly pay for itself. It can only do that if the roof underneath it was checked as seriously as the panels on top. If you want to see how we approach this in practice, our case studies show real installations, and the engineering section explains what our structural assessment covers before we quote a single number.