Iron reinforcements in Beauvais and Metz Cathedrals: from bloomery or finery? The use of logistic regression for differentiating smelting processes

A specific statistical approach was tested to determine the ironmaking processes (bloomery or indirect) used to manufacture iron reinforcements found in two French gothic medieval monuments: Metz and Beauvais Cathedrals. Slag inclusions embedded in the metallic matrix were analysed and the major element compositions were quantified, using a Si Drift EDS detector permitting the study of several hundred inclusions per artefact. First, pre-processing was applied to the raw data to discriminate inclusions from the smelting stage (or refining stage for the indirect process) from those formed during post-reduction operations (forging etc…). PCA and hierarchical clustering on the major element compositions were then performed. Secondly, a specific multivariate statistical method, logistic regression, was applied to a learning set of data from a reference set of samples. This allowed the development of a model capable of distinguishing artefacts from the two smelting processes and to link unknown samples to one or the other. This model was then used to study artefacts from the Mutte Tower of Metz Cathedral and Beauvais Cathedral. This allowed us to confirm the extensive use of iron reinforcements since the construction of Beauvais cathedral. In the Mutte Tower of Metz Cathedral clamps from Lorraine reinforced some parts of the monument from its construction between the 13th and the 15th centuries. •New approach is tested to differentiate making processes on iron reinforcements.•Several 100 slag inclusions per artefact are analysed.•Logistic regression is applied to build a model that distinguishes the processes.•The model is used to study artefacts from Metz and Beauvais cathedrals.•New data on the diffusion of the indirect process during the 15 h c. are obtained.