Alternate identifier:
(KITopen-DOI) 10.5445/IR/1000151546
Related identifier:
Schlagenhauf, Tobias [Institut für Produktionstechnik]

Wolf, Jan [Institut für Produktionstechnik]

Puchta, Alexander [Institut für Produktionstechnik]
Multivariate time series dataset of milling 16MnCr5 for anomaly detection
Additional titles:
(Abstract) The dataset was recorded during milling of 16MnCr5. Due to artificially introduced, though realistic anoma-lies in the workpiece the dataset can be applied for anomaly detection. Furthermore, milling tools with two different diameters where used which led to a dataset eligible for transfer learning.
(Technical Remarks) The dataset consists of seven folders. Each folder represents one milling run. In each milling run the depth of cut was set to 3 mm. A folder contains a maximum of three json files. The number of files depends on the time needed for each run which is a function of milling tool diameter and feed rate. Files in each folder were numerated in sequence. For example, folder “run1” contains the files “run1_1” and “run1_2” with the last number indicating the order in which the files were generated. The frequency of recording datapoints was set to 500 Hz. During each milling run the milling tool moved along the longitudinal side and then was moved back alongside the workpiece. This way machining started always on the same side of the workpiece. Table 1 provides an overview of the milling runs. Run 1 to 4 were performed with a HSS tool with a diameter of 10 mm. The tool in use was an end mill (HSS-E-SPM HPC 10 mm) developed by Hoffmann Group. During the first three runs with this end mill no tool breakage occurred. However, in run 4 the tool broke. Runs 5 and 6 were performed by milling with an end mill of the same tool series (HSS-E-SPM HPC 8 mm) that just differs in tool diameter. In contrast to this run 7 was performed by using a solid carbid tool (Solid carbide roughing end mill HPC 8 mm). Cutting with SC tools provides much higher productivity with the downside being higher tool price. In our case the SC end mill performed cuts with a feed rate of 1150 mm/min compared to 191 mm/min achieved by a HSS end mill of the same diameter. Tool breakages were recorded on all runs with end mills of diameter 8 mm. Table 1. overview of the data folders folder name | number of json files | tool diameter | tool breakage | tool type run 1 2 10 mm No HSS run 2 2 10 mm No HSS run 3 2 10 mm No HSS run 4 2 10 mm Yes HSS run 5 2 8 mm Yes HSS run 6 3 8 mm Yes HSS run 7 1 8 mm Yes SC Each json file consists of a header and a payload. The header lists all parameters that were recorded such as position, motor torque and motor current of each of a maximum of five axes of a milling machine. However, the machine used in our experiments is a 3-axis machining center which leaves the payload of 2 possible additional axes to be empty. In the payload the sequential data for each parameter can be found. A list of recorded signals can be found in Table 2. Table 2. recorded signals during milling Signal index in payload | Signal name | Signal Address |Type 13-18 VelocityFeedForward VEL_FFW|1* double 19-24 Power POWER|1* string 25-30 CountourDeviation CONT_DEV|1* double 38-43 TorqueFeedForward TORQUE_FFW|1* double 44-49 Encoder1Position ENC1_POS|1* double 56-61 Load LOAD|1* double 68-73 Torque TORQUE|1* double 68-91 Current CURRENT|1* double * 1 represents x-axis, 2 represents y-axis, 3 represents z-axis and 6 represents spindle-axis. Note that our milling center has 3 axis and therefore values for axes 4 and 5 are null.
Anomaly Detection
Transfer Learning
Machine Learning
Production Science
Related information:
Production year:
Subject areas:
Resource type:
Data source:
Software used:
Data processing:
Publication year:
Rights holders:
Schlagenhauf, Tobias

Wolf, Jan

Puchta, Alexander
Name Storage Metadata Upload Action

Number of views in the previous six months.

Dataset page views




Overall statistics

Period Landing page accessed Dataset downloaded
Dec 2023 2 0
Nov 2023 30 3
Oct 2023 15 1
Sep 2023 13 2
Aug 2023 12 0
Jul 2023 11 0
Before 10 1
Total 93 7
Uploaded by:
Created on:
Archiving date:
Archive size:
219.4 MB
Archive creator:
Archive checksum:
7fbd9e2823ea8d6a41b04edd806b40ae (MD5)
Embargo end date: