ArQave · Archaeological materials science
From excavation report to phase diagram.
ArQave is a web application for the extraction, analysis and visualisation of compositional data from archaeological materials. It brings together seven tools commonly used in archaeometallurgy, and a shared reference library, in a single workspace.
Tools
Seven tools for compositional analysis.
Each tool accepts a file and returns its results as downloadable files. Select a tool for a schematic of its output. Use of the tools requires registration.
PDF table extraction
Extracts data tables from excavation reports and specialist appendices, including scanned documents, into one Excel workbook.
Data preparation
Converts any table of analyses into one row per sample with numeric oxide and element columns. Detects header rows, orientation, units and values below the detection limit.
Group building
Assigns samples to analytical groups by burial, feature, material or phase, and summarises the composition of each group.
Cluster analysis
Groups samples by chemical composition using k-means or agglomerative clustering, and returns a labelled table and plot.
Principal component analysis
Computes up to five principal components and reports explained variance, scores and loadings.
Ternary diagrams
Plots normalised compositions on five published phase diagrams, selected automatically or by the user.
Micrograph analysis
Segments pores and cementite in micrographs of iron and steel, and estimates phase fractions and bulk carbon content.
Workflow
A typical sequence of analysis.
Select a step to see its inputs, outputs and the tool used.
Methods
Procedures used by each tool.
PDF table extraction
Scanned pages are processed by optical character recognition (OCRmyPDF with Tesseract). Tables are detected with Tabula and Camelot. Tables that fail quality checks are excluded and the reason is recorded.
Data preparation
Header rows are detected and merged, and the table is transposed when oxides or elements are listed in the first column. Values given as mean ± uncertainty keep the mean. Values in ppm are converted to wt% and Fe₂O₃ to FeO (factor 0.8998) when selected. Oxide totals outside 95 to 105 wt% are reported.
Group building
Samples are assigned to groups by the values of one or more selected columns. For each group the number of samples and the mean and standard deviation of every numeric variable are reported.
Cluster analysis
Numeric columns are normalised so that each sample sums to 100 per cent, and standardised. Samples are grouped by k-means or by agglomerative clustering with Ward linkage (scikit-learn).
Principal component analysis
Numeric columns are standardised to zero mean and unit variance before decomposition. Between one and five components are computed.
Ternary diagrams
Three components are normalised to 100 per cent. The diagram is chosen from the mean composition, or by the user, among CaO-SiO₂-Al₂O₃, K₂O-SiO₂-Al₂O₃, FeO-SiO₂-Al₂O₃, MnO-SiO₂-Al₂O₃ and FeO-CaO-SiO₂.
Micrograph analysis
Pixels are classified by grey-level thresholds into pores, cementite and pearlite. Bulk carbon is estimated by the lever rule, using 6.67 wt% C for cementite and 0.77 wt% C for pearlite.
Reference library
Publications are stored as PDF files with title, authors, year, site and material. Entries are searched and filtered on these fields and are visible to every registered user.
Resources
Reference library.
A shared collection of publications on archaeological materials, searchable by title, author, site, material and year. Registered users may add publications with bibliographic details. All entries are visible to every registered user.
Open the library →- 2021Enkomi (Cyprus): Using pXRF spectroscopy to identify LBA copper alloys
- 2021A Late Bronze Age hoard of bronzes rediscovered, probably from Palaepaphos in Cyprus
- 2014A compositional study of Cypriot bronzes dating to the Early Iron Age using pXRF
- 2014Chemical analyses of copper alloy artefacts from Pyla-Kokkinokremos
Contact
Questions and access requests.
ArQave is developed and maintained by CityAI. For questions, access requests or to report an error, please write to info@cityai.space.