Français FrançaisDeutsch DeutschSitemapSearch


Statistical Calculation and Development of Glass Properties

Add this Page to Favorites

Analysis and development of glasses with desired physical properties based on chemical composition - property statistics

This website promotes the statistical analysis and calculation (modeling) of physical properties of silicate, borosilicate and other specialty glasses based on global glass databases for facilitating process modeling, process control, and development. In addition, property measurement techniques are described because only accurate experimental data allow reliable predictions.




Otto Schott

Common Glasses


Glass Databases


Statistical Modeling Principles


Measurement Error Development


Glass Properties:


Viscosity - Liquidus Temperature


Density of Glass at Room Temperature


Density and Thermal Expansion of Glass Melts


Thermal Expansion below the Glass Transition


Optical Properties at Room Temperature


Electrical Conductivity / Resistivity of Melts


Chemical Durability (Hydrolytic Class)


Thermal (Phonon) Conductivity


Water Solubility in Glass Melts


Elastic Properties


Surface Tension

Ernst Abbe

Specific glass properties of interest can be determined from the chemical glass composition, for example given a commercial soda lime or borosilicate glass, and the property-composition models on this website.

Innovation compared to previous glass property calculation models:



For statistical glass modeling up-to-date global glass databases are used, i.e., property values from a high number of sources (laboratories). Outlying datapoints are excluded from the calculation.


The analysis of global glass databases does allow a significant reduction of systematic errors through comparative glass modeling, i.e., through investigation of categories as described in the linear regression tutorial (PDF, 0.4 MB) on page 11. Systematic differences between laboratories are tested and if necessary corrected.


The statistical glass modeling does not only result in a prediction of the property value itself. It also includes an error estimation for glass in mass production, given the desired error confidence level, e.g. 95%, and the uncertainty of the glass composition of interest (see viscosity calculator).


In the statistical property calculation approach glass property standards are included, e.g., NIST 710A, 710, 717A, 711, 731, DGG I, which ensures high accuracy (see viscosity model comparison).


The model application limits, considering component combination limits are evaluated in detail. Therefore, it is unlikely to obtain inaccurate results through unintentionally entering a glass composition for which the model is not valid for.


The procedure involves a leverage analysis, i.e., glasses with high influence on the model and/or unusual compositions are analyzed in particular and excluded from the calculation if in doubt.


The statistical analysis of global glass databases gives new insights that are difficult to obtain otherwise, such as an improved understanding of the mixed-alkali effect for the viscosity of glasses.

Advantages of the statistical modeling approach for the calculation of glass properties:



Superior accuracy: Because the models are based on many datasets including various measurement techniques, systematic errors of specific investigators can be analyzed and resolved. In often investigated composition areas the large number of data does allow a higher prediction accuracy than measurements within one single laboratory permit.


Time and financial savings: Calculations can be completed within a few minutes. In contrast, one single experimental investigation may take several hours or days, including high personal and equipment expenses, and the measurement accuracy still needs to be confirmed statistically afterwards by comparing results with those of other investigators.


Optimization: Several models can be combined and reversed for determining the best glass composition from a desired set of physical properties. Besides property-composition models other functions can also be considered in the optimization procedure such as financial estimators or processing calculations.


Compatibility: Statistical glass modeling makes data comparable over wide composition ranges, different measurement techniques, and from various investigators. Previous models can be integrated, e.g., Lakatos et al. in "High temperature glass melt property database for process modeling".


Broad application range: The accuracy of new measurements mostly can not be derived directly from similar results in the scientific literature, or from NIST or DGG glass property standards. The statistical modeling approach may include several glass types with different chemical compositions; it makes them comparable. Consequently, the model allows predictions and glass development in composition areas that are not covered by common industrial models that are valid for only one specific glass type. The results of experiments inside and outside conventional ranges can be predicted and compared economically.


Transparency: The statistical calculation of physical glass properties is based on mathematical methods, established over several decades and well comprehensible. Further details are summarized in statistics textbooks and at Principles of Statistical Glass Modeling. For comparison: Services of other organizations concerning the calculation of glass properties are offered (see Glass Links, Calculation of Glass Properties), however, the basics of the calculations are not given sufficiently and therefore the predictions might be questionable.

The benefits of the calculation of physical glass properties are demonstrated in articles by Bingham and Marshall ("Reformulation of container glasses for environmental benefit through lower melting temperatures"; Glass Technol., vol. 46, 2005, p 11-19), and by Wallenberger et al. ("The effect of boron on the properties of fibreglass melts"; Glass Technol., vol. 47, 2006, p 148-152). The models presented on this website do allow the development of similar and additional process and property improvements.

Principles of Statistical Glass Modeling >

Overview of existing glass property models >

Alexander Fluegel,, December 7, 2007