September 25th 2020

Data Driven Production Improvement

Research unit digital transformation: Project for Tiroler Rohre GmbH

Data Driven Production Improvement: Research project for Tiroler Rohre GmbH from November 2019 until September 2020
Data Driven Production Improvement: Research project for Tiroler Rohre GmbH. Photocredit: Tiroler Rohre GmbH


Project Duration: November 2019 – September 2020

Involved Parties:

  • Tiroler Rohre GmbH
  • Management Center Innsbruck
  • Kempten University of Applied Sciences

Team MCI: FH-Prof. Dr. Christian Ploder; MCiT Master’s student Jakob Kübelböck, FH-Prof. Dr. Reinhard Bernsteiner


The project aimed to optimize the wall thickness of pipes based on given production data. To be more precise, Tiroler Rohre GmbH produces pipes with a variety of different dimensions, and based on the production technology, the thickness varies. Additionally, the permanent measurement of the wall thickness during the production process is complex. If the pipe is too thin, the specification levels would not be reached, and the tube has to be classified as waste, as the tube may not withstand the pressure. If the wall is too thick, this would be a very inefficient way of material usage, and the pipe gets unnecessary heavy. With the automated data acquisition of the centrifugal pipe casting machine, patterns between individual parameters are to be recognized, and in further steps, they will be improved by machine learning. However, in the final stage, this process optimization is intended to make it easier for the machine operator, above all, by enabling the operator to read off on a cockpit, which parameters need to be changed to return to optimum casting.

The MCI part of this project is focused on the data collection, transforming/storing/analyzing the data, and take a look at the data quality aspects paired with the process of understanding. First of all, the production process was modeled with all the necessary data. Subsequently, the database architecture was rebuilt for the centrifugal pipe casting machine. For the data quality aspects, logics for data filtering have been set up at the database. During the data understanding, lots of different analysis tools have been implemented to ensure the fit of the real-life with the data stored in the system, and some data verification has to be done.

What we achieved

First of all, we gained an understanding of the existing process and captured it - this was achieved mainly through visits to the production site. By creating a BPMN diagram, not only we but also Tiroler Rohre GmbH was able to record the process in a standardized way. By capturing the process, we were then able to identify over a hundred essential process parameters, which is a fundamental insight for the goal of the project, and it was possible to classify them based on their importance and divide them into measurements and fixed values.

We also rebuilt the database, which was a challenge because this company has been collecting data for decades, and its architecture has grown historically. We filtered the data critical to the process and reassembled all relevant data using an Extract, Transform, Load (ETL) process. On the one hand, the data can be accessed quickly, but on the other hand, they are also clearly summarized, which makes data analysis easier. For this purpose, R and partially Python was used for statistical methods, and correlations were applied. Significant findings were then packed into a data report.

In the end, we helped Tiroler Rohre GmbH to find out how well their wall thickness measurement works. We deduced the pipe weight from the respective wall thickness values by using the concept of cylinder calculation. We compared the actual weight of the pipe with the pipe weights calculated by us, which were formed by taking (possibly) decisive factors into account. In this way, we tried to show the possible disadvantages of their wall thickness measurement. We also applied this whole procedure separately to the middle third, as there was the suspicion that this was very robust and low in fluctuation.

The next part of the data simulation will be done together with colleagues from the Kempten University of Applied Sciences based on their experience and expertise in this particular production process.

Now, this project comes to an end with only some little steps in front of us, and we will use the possibility to give a huge thanks to our project partners – especially Dipl.-Ing. Harald Tschenett and Dipl.-Ing. Roland Satlow –  for the exciting project and their willingness to open discussions and new ideas.

Prof. (FH) Dr. Christian Ploder | Professor Bachelor's program Management, Communication & IT
Prof. (FH) Dr. Christian Ploder Professor +43 512 2070 - 3536
Prof. Dr. Reinhard Bernsteiner | Professor Master's Program Management, Communication & IT
Prof. Dr. Reinhard Bernsteiner Professor +43 512 2070 - 3532