Multivariate time series data of milling processes with varying tool wear and machine tools

Machining is an essential part of modern manufacturing. During machining, the wear of cutting tools increases, eventually impairing product quality and process stability. Determining when to change a tool to avoid these consequences, while still utilizing most of a tool's lifetime is challenging, as the tool lifetime can vary by more than 100% despite constant process parameters [1]. To account for these variations, all tools are usually changed after a predefined period of time. However, this strategy wastes a significant proportion of the remaining lifetime of most tools. By monitoring the wear of tools, all tools can potentially be used until their individual end of life. Research, development, and assessment of such monitoring methods require large amounts of data. Nevertheless, only very few datasets are publicly available. The presented dataset provides labeled, multivariate time series data of milling processes with varying tool wear and for varying machine tools. The width of the flank wear land VB is used as a degradation metric. A total of nine end milling cutters were worn from an unused state to end of life (VB ≈ 150 µm) in 3-axis shoulder milling of cast iron 600–3/S. The tools were of the same model (solid carbide end milling cutter, 4 edges, coated with TiN-TiAlN) but from different batches. Experiments were conducted on three different 5-axis milling centers of a similar size. Workpieces, experimental setups, and process parameters were identical on all of the machine tools. The process forces were recorded with a dynamometer with a sample rate of 25 kHz. The force or torque of the spindle and feed drives, as well as the position control deviation of feed drives, were recorded from the machine tool controls with a sample rate of 500 Hz. The dataset holds a total of 6,418 files labeled with the wear (VB), machine tool (M), tool (T), run (R), and cumulated tool contact time (C). This data could be used to identify signal features that are sensitive to wear, to investigate methods for tool wear estimation and tool life prediction, or to examine transfer learning strategies. The data thereby facilitates research in tool condition monitoring and predictive maintenance in the domain of production technology.