Structural glass tends to influence  modern architecture more and more. Following are some basic design rules for the glass design. 

1. Redundant construction as basic rule 

Glass can always break, even if designed properly. Or, more scientific, the initial energy to break glass is very low, even a minor accident may destroy this material. Glass structures must be designed redundantly, so if one glass part brakes, the rest of the structure (either steel or also glass parts)  will still be safe, may be with a reduced level of safety. Therefore, glass construction are mostly secondary or tertiary structural elements (in term of structural hierarchy of a building), i.e. parts of the building envelope, but not parts of the main primary structure.

The understanding of aspect of  reduced safety after glass failure is the vital element of glass design. This case can be assessed  by means of analysis, but mostly by experiments.

2. Loads

The imposed loads for glass are the same as for any cladding detail. Main loads are wind, snow and own weight, the national codes of practice ought to be used. Care should be taken to consider the increased wind loads at corners of buildings. High-rise buildings or buildings with unusual forms may require a detailed wind assessment with wind channel research.

3. Support forces and bending moments

Since glass is typically used as plates with linear or punctual supports, bending moments and support reactions can easily be obtained by using simple FEM programs with plate elements or by literature with tabulated values for plates.  The approach with the linear-elastic theory (Kirchhoff-theory) is always a good way because its at safe side. Deflections more than plate thickness indicate the limit of that theory, i.e. nonlinear calculation could here yield lower stresses, but that effort is actually seldom required.

For the allowable stress approach the forces and bending moments need no load factors included.
Most FEM programs give you already the stresses in both direction and the  principal stresses. Please note that due to glass crack mechanism  the principal tension stress will lead to the crack, so this maximum value ought to be compared with the allowable stress.

Laminated glass (made from layers of same thickness) ought to be modeled as one layer,  imposed the reduced load =  real load divided by the number of layers. The shear composite effect ought to be neglected, since its not sure, especially with long-term-loads and under high temperature the PVB foil tend to creep.

4. Allowable stresses (as currently applied in Germany) 

Following are the values for maximum allowable stresses (they include already a global safety factor of 2.4 against 5%-quantil value for breaking) have been applied for numerous projects with the approval of building authorities:

5. Comments

With "Laminated safety glass" is meant : Glass,  laminated with PVB foil. Other glass with acrylic compound have to be especially approved by the building authority.  The thickness of the PVB foil may affect the rest safety after break of one layer. Typical PVB foil thicknesses  are 0.38mm, 0.76 mm and 1.52 mm. Glass that must withstand pedestrian load by cleaners has to have at least a 0.76 mm foil. Glass for regular use as pedestrian area is usually at least a 4-layer composite glass. Safety glass (ESG) is fully thermically prestressed glass, which breaks in small smooth pieces. It is used as car glass as well. Thermally / heat strength glass (TVG) is like the safety glass thermally prestressed, but less than safety glass. It breaks like normal float glass, but has higher allowable stresses. Being laminated, TVG has the best redundant safety properties, i.e. even if broken no parts fall down and the broken laminate acts like a carpet between supports and has still astonishing loadbearing capacities. Therefore, laminates from TVG became first choice for engineered glazing meanwhile.

All stresses are to compare with major tension stresses. Experiments have shown that in-plane stresses lead earlier to failure than plate stresses due to bending, so for the maximum allowable stress for in-plane loads (shear panel loads)  90% of the values above should be taken.

Punctual fittings consider much more detail knowledge. The common way is to test the actual fitting type with the glass type and find experimentally the maximum break load.  Common fittings allow break forces perpendicular to plane of about 5 to 15 kN (pull-out-test). If you have a typical glass of 1.35 x 2.7 m and a wind load of lets say 1 kN/m², each fitting will have to carry 0.91 kN, which gives to 5 kN ultimate load an overall safety factor of about 5. Those reserves vapor with horizontally glass, where own weight and snow are imposed, so more fittings and/or such with large  washers may have to be used.

Punctual fittings for in-plane-loads (e.g. load-bearing glass fins with bolted supports) are much more special. Some fitting manufacturer have fitting types in stock for those purposes. As a thumb rule you can obtain the average stress in the glass hole by stress =  bolt force /(glass thickness * hole diameter) . This value should not reach more than 1/3 of the allowable glass stress written above. The reason is that - in difference to steel, glass is not ductile, so stress peaks in the hole cannot plastify away, and  tend to crack the glass very early.  For that type of connection, experiments are strongly recommended. For laminated glass with drilled holes I just remember for  the known problem, that the drilled holes of the layers cannot  be placed exactly side by side, so some special knowledge is  required to tackle this problem.

The glass itself always ought to be supported in plane statically determined to avoid forces due to temperature.

For insulating glass of small (below 1,2 m) edge size, an additional stress component due to the pressure change inside the insulating gas volume must be considered.

Deflection of glass is limited to 1/100 ... 1/200 of span. Actually, in terms of breaking the deflection does not matter at all, because due to the low Youngs modulus  glass bends astonishing wide  before breaking. I suggest to choose the design deflection limit keeping in mind  the elasticity  of the joint sealing material to obtain watertight sealings. More important is the deflection of the steel substructure of the glass.  The allowable substructure movement can be checked by FEM analysis  too.

6. Reduced safety after break of redundant glass parts 

For an first analytical assessment, loads can be reduced as following:
- wind or snow loads reduced to about 66%... 75% (approximately the maximum wind value in 2 years, instead of the maximum value in 50 years which gives the standard wind and snow load of the codes of practice)
- global safety factor at material resistance side reduced to 1.6 (instead of 2.4), in other words, the allowable glass stresses may be factored by 1.5 An analysis with those parameter will give a first answer if the glass dimension is viable at all.

7. Glass and lateral stability 

In general, glass should not serve as lateral stabilization member. Actually, in most cases the load bearing capacity of glass is sufficient for lateral stability members. But, a sudden failure of the glass could lead to a chain reaction collapse of those members who lost its lateral bracing, and the whole of the building could fail. The second reason is, that lateral bracings require a tight fit to the structure, but glass ought to be supported without tight fit so temperature can work without adding additional force to the glass.