Critical Review

Pagni and Joshi’s paper provided a theoretical solution to the problem of glass breaking in fires at a time when the scholarship was limited. With the benefit of almost 20 years hindsight it is arguably easy for enthusiastic postgraduate students to be critical of this pioneering work. Yet the model has subsequently been validated by experimental work and the fundamental results have not been substantially improved upon. That stated, an interested reader might examine the following aspects of the treatise.

No experimental evidence was undertaken or cited to support the assumption that glass breakage initiates at the glass edge. (Subsequent work, and in particular that of Skelly and Roby, validate this assumption.)

The paper identifies the importance of quantifying the tensile breaking stress of the glass, the compartment fire history and determining radiative and convective heat transfer parameters.

Determining the tensile breaking stress of a particular pane of glass a-priori (that is prior to the fire) is perhaps the most problematic issue in application of the model.  The model is linearly dependent on glass breaking stress. However the paper states that values for soda-lime glass are wide-ranging (40 MPa to 128 MPa) and can be significantly influenced by defects and edge imperfections caused by cutting and fitting. 40 MPa is offered as a conservative value for the purpose of calculation of time to breakage.

The paper assumes that the edge of the glazing is simply supported.  While this is true in the majority of practical glazing systems, it may not be valid for rebated glazing, subject to the skill and experience of the glazier.

At the time of this research compartment fire histories could be reasonably well predicted for simple design fires using early computer models or correlations.  Subsequent improvements in theory and computational power have increased the accuracy with which simple compartment fires can be modeled.

The paper assumes that heat transfer parameters are constant.  Later work by Pagni and Joshi utilized linearly variable heat transfer parameters which allowed good prediction of experimental glass surface temperatures.

Pagni and Joshi’s analysis of absorption of flame radiation was modeled through the thickness of the glazing based on an estimate of the decay length for the glass. Their 2D temperature profile results indicate that for relatively fast compartment fires this was an unnecessary and computationally expensive complication because most of the heat flux was absorbed close to the exposed glazing surface.

The paper supports and refines the conventional wisdom of the time that glass breakage could be correlated with a critical hot layer temperature of approximately 250 to 370°C in a relatively fast compartment fire.  However, it extends this by providing a method for predicting glass breakage for slow heating where glass conduction allows the shaded edge to be heated appreciably by the exposed field.

The paper’s abstract identifies that it allows ‘accurate’ prediction of the breaking of glass. However the paper does not actually quantify the inaccuracies in the predictive calculation.

The paper assumes even heating of the exposed field of the glass.  ‘Real’ compartment fires can create conditions where this assumption is not valid.  The depth of the upper hot gas layer can be expected to soak the upper portion of a window while the lower sections remain relatively cool.  Similarly a fire initiated below and at the edge of a window will heat one side substantially more than the other.   These complicating cases are not addressed by the paper.

Given the difficulty of determining specific material properties and the stochastic nature of ‘real’ compartment fires one is lead to question the need for ‘accurate’ determination of glazing failure.  Although glazing surface temperatures can be predicted to within 10% of experimental measurements difficulty in determining the tensile breaking stress preludes accurate quantitative practical application of the calculation method.   Arguably the determination of conservative lower bounds for time to breakage would find more use in practical application.

The paper recognises, but does not address the need for prediction of fall out of the glass. This aspect of the venting problem has been the subject of a significant body of later research.  It is noteworthy that these later publications invariably support the theory of crack initiation at the interface between the shaded and unshaded portions of the glass.