Condition-Based Maintenance: Analysis of highly variant, temporal sparsely settled, unmarked data sets for performing condition-based maintenance
The enterprise “TRUMPF Machine Tools” is the global leader in the production of machine tools for sheet metal forming (laser, punching, and bending machines). At specified, but irregular intervals a “digital image” in the form of a data collection of logging and configuration information is created in a TRUMPF machine tool. Besides configuration data, such as machine series status and setting parameters, logging data, such as the notification history, is included. The data are stored in a zip-compressed container with the name MIO (Machine Information Object). Thus, the container includes a wide range of data files – each with different information that is stored either as history with timestamp or snapshot. Each data file has its own structure.
Using these data, the project being planned aims at detecting deviations (anomalies) from the so-called “normal operation”. Moreover, the intent is to reveal correlations to yet unknown factors, by analyzing the separately present notification history and other information. For example, it can be detected if the safety devices, such as guard door monitors, are not working properly because they were electrically bridged. In a first step, a correlation between customer escalation and MIO will be established. In a second step, a prediction of service cases is to be enabled. Under terms of strict confidentiality, TRUMPF can provide a significant number of MIO-files of which each show a substantial amount of data and information.
Data Innovation Community
KIT, SAP SE, Trumpf
The recently completed SDIL project used rule-based approaches to discover sequence of machine events that led to critical conditions in the machines. For this purpose, algorithms of Sequence Mining and Process Mining have been combined. Read the project’s results at: http://www.sdil.de/downloads/sdic-2016-konferenzband.pdf#page=45
Klaus Bauer, email@example.com