After a cyclone crosses the coast the calls are rarely about modules that broke, they are about modules that left. Zone loading, clamp position as a warranty condition, torque records, ballast against penetration, and the inspection routine for after the storm.
After a cyclone crosses the coast, the calls we get are rarely about modules that broke. They are about modules that left. A panel that lifts out of its clamps in a gust does not stay on the roof and it does not land harmlessly. Two seasons of pre-monsoon squalls and one named storm will find every clamp that was fitted by hand and never torqued, and every corner of the roof where the structure was designed as though the whole array saw the same load.
Wind flowing over a building separates at the edges and rolls into vortices at the corners. The suction generated there is a multiple of the suction in the middle of the roof. Design codes handle this by dividing the roof into zones, field, edge and corner, with rising pressure coefficients as you move outward. The Bangladesh National Building Code gives the basic wind speed by location, and the components and cladding approach familiar from ASCE 7 is what the mounting supplier's calculation will be built on.
The practical consequence is simple and it is ignored constantly. Corner modules need more fixings, heavier ballast or a setback that keeps them out of the worst zone. An array laid out on a uniform grid with uniform fixings has a designed strength that matches the field of the roof and a weakness at every corner.
Often the cheapest engineering decision is to leave the corner and edge strips empty. You give up a few kWp and you avoid designing for the highest suction on the building. A parapet helps, but only if it is tall enough to matter and only if the parapet itself is sound, which on older Bangladeshi factory buildings is worth checking rather than assuming.
There is a second pattern within the array. The first row and the last row of a tilted arrangement carry a different load from the middle rows, because the front row takes the full oncoming flow underneath the panel. Wind deflectors or a closed skirt on the leading row change the load path and earn their cost on exposed sites near the coast or in open ground.
Every module manufacturer publishes a mounting instruction defining where on the frame the module may be clamped, given as a permitted range measured along the long side, sometimes with a separate set of permitted positions for short-side mounting. Clamping outside that range voids the mechanical warranty, and more to the point it changes where the glass and the cells are stressed under load.
We have surveyed arrays where the rail spacing was set to suit the existing purlins and the clamps then landed wherever the rails happened to fall. The order of work is what prevents that. Confirm the purlin or rafter spacing on site, select the module and read its permitted clamping range, then set the rail spacing to suit both. Done the other way round, the roof ends up with clamps in the wrong place and no way to correct it without taking modules off.
The clamp manufacturer publishes a torque figure and it is not advice. Under-torqued, the module works loose under cyclic wind loading and the joint degrades every time the wind blows. Over-torqued, you deform the frame or strip the T-slot, which is the more dangerous failure because it looks perfectly tight from a metre away.
On an RCC roof the choice is between fixing through the waterproofing and holding the array down with weight. Both are legitimate and they fail in different ways.
Penetrating fixings are stronger and more predictable. The uplift path runs into the slab and the design is an anchor calculation with a known answer. The risk is water. Every penetration is a leak waiting for a poor detail, and a leak over a production floor is expensive in ways that have nothing to do with solar. Done properly, with a chemical anchor, a raised plinth so the fixing sits above any ponding, and a flashing detail that was drawn rather than improvised, penetrations are what we would use in an exposed coastal location.
Ballast avoids the penetration and protects a membrane warranty. It brings three problems. The dead load is significant and has to be checked against the structure by someone qualified to check it. Ballast resists uplift by weight alone, so the corner zones need the most weight, which is exactly where you can least afford to add it. And ballast resists sliding by friction, so the friction coefficient against the actual roof surface matters, and it drops when that surface is wet.
The defensible answer in a high wind zone is usually a mixture. Ballast in the field of the roof if the structure will carry it, mechanical fixing at edges and corners, and a perimeter that is either properly restrained or left empty. A fully ballasted array sitting at the corner of an exposed coastal roof is a bet against the next storm.
Bangladesh carries its cyclone risk in two windows, pre-monsoon around April and May and post-monsoon around October and November, and there is normally two to three days of warning. That is enough time to do something useful with.