Tool Wear Experiment Dataset for the Evaluation of Sensor Informativeness in Longitudinal Turning (ExtraDrey)

This dataset was collected at the Institute of Production Management and Technology (TUHH IPMT) to investigate tool wear behaviour in CNC longitudinal turning and to evaluate the suitability of different in-process sensor modalities for predicting maximum flank wear using machine learning. Experiments were performed on two workpiece materials (C45+N, X5CrNi18-10) using indexable inserts. Process parameters were varied according to a Fractional Factorial Design, are fully documented in the accompanying metadata, and were held constant throughout the full life of each cutting edge until failure.

For each 30-second cutting experiment, the following in-situ online measurements are available: three-axis cutting forces and tool-side acceleration; structure-borne acoustic emission; rake-face temperature. In-situ offline measurements, recorded at 30-second intervals, include scattered light from the workpiece surface (Aq and macro profile angle) and microscope images of the cutting edge from three perspectives — flank face (FF), rake face (RF), and cutting-edge top view (CE) — at standard 150× and optionally 100× magnification, together with surface heightmaps via focus variation microscopy. Segmentation masks identifying the flank wear zone are available for selected images.

The maximum flank wear (VB_Max, µm) is provided as a quantitative wear label for each experiment, embedded in the sensor data files and listed in the accompanying documentation spreadsheet. It was creted through a machine vision programm.

The dataset is suitable for training and benchmarking machine learning models for tool condition monitoring, multi-modal sensor fusion, and wear-progression modelling. The combination of synchronised time-series sensor signals with microscope images, surface heightmaps, and segmentation masks supports investigation of the individual and combined informativeness of different sensor types for wear prediction. The segmentation masks additionally enable computer vision approaches to automated wear measurement. The documented Fractional Factorial design allows statistical analysis of parameter effects on wear behaviour, and the sensor signals can be used to validate physics-based wear models.