September 11, 2019 Dataset Open Access

Sensor data set radial forging at AFRC testbed v2

Christos Tachtatzis; Gordon Gourlay; Ivan Andonovic; Omer Panni

Sensor data set, radial forging at AFRC testbed

General information on the data set

Radial forging is widely used in industry to manufacture components for a broad range of sectors including automotive, medical, aerospace, rail and industrial. The Advanced Forming Research Centre (AFRC) at the University of Strathclyde, Glasgow, houses a GFM SKK10/R radial forge that has been used as a testbed for this project. Using two pairs of hammers operating at 1200 strokes/min, and providing a maximum forging force per hammer of 150 tons, the radial forge is capable of processing a range of metals, including steel, titanium and inconel. Both hollow and solid material can be formed with the added benefit of creating internal features on hollow parts using a mandrel. Parts can be formed at a range of temperatures from ambient temperature to 1200 °C.

For the provided data set, a total of 81 parts were forged over one day of operation. A machine failure occurred during the forging of part number 70, and this part was re-run once the malfunction had been fixed. Each forged part was then measured using a CMM to provide dimensional output relative to a target specification and tolerances. The CMM records 18 dimensional measurements.

The aim of the measurement setup is to predict the quality (in terms of dimensional properties) of the forged part from the sensor measurements during the forging process.

Structure of the data

• The sensor readings for the forging of the parts are provided in 81 csv files in the folder “Scope Traces”, named “Scope0001.csv” to “Scope0081.csv”. Each file contains the readings (columns) against time (rows). The first column displays the clock times (in milliseconds).
• A commentary on the sensors is provided in the file “ForgedPartDataStructureSummaryv3.xlsx” (NOTE: Some columns do not have sensor descriptions as this information is not available).
• The CMM data is provided in the file “CMMData.xlsx”.

Further Information

For an introduction and tutorial to this data, a set of Jupyter notebooks is available here:

https://github.com/harislulic/ZeMA-machine-learning-tutorials/releases/tag/v0.0.1

These notebooks contain Python code and a documentation of example machine learning tasks and analysis of this data set. In the near future, these will be extended to also include uncertainties in the input data.