Experimental dataset of composition–property relations for 46 PTFE composite formulations with specimen-level repeatability data

An experimental dataset of 46 compositions of polytetrafluoroethylene (PTFE)-based composites reinforced with eight fillers of different chemical natures: carbon fibre, basalt fibre, coke, graphite, kaolin, sodium chloride, ultradisperse PTFE and titanium dioxide. It includes 31 two-component formulations, 14 three-component hybrid formulations, and one unfilled PTFE matrix. For each formulation, specimen-level measurements (n = 5) of density, tensile strength, relative elongation and wear intensity are reported, totalling 230 observations. All materials were manufactured using a single powder-metallurgy route and tested under a single regime, thereby removing confounding from processing and test conditions. Pooled within-formulation repeatability standard deviations computed in accordance with ISO 5725-2 are provided as the physical error scale for benchmarking predictive models. The dataset supports formulation-grouped machine-learning benchmarking (e.g., leave-one-formulation-out and leave-one-filler-out protocols), descriptor-based modelling and additivity/synergy analysis of hybrid fillers.