TrendMiner Use Case
Situation
Infant cereals in at a food manufacturing factory go through a pasteurization process. The old pasteurization unit was thought to cause a bottleneck, so the company installed a new one. Two different lines (Line 1 and Line 2) had been set up. The lines are long pipes where a mix of flour, water, and additives is added to steam. The feed undergoes a thermic treatment where high temperatures (about 120º C) remove microorganisms. After that, a flash tank is connected to the outlet to remove the remaining steam and cool the product.
Goals
- Understand what is causing the pasteurization line shutdowns
- Propose a solution to prevent them
Challenges
- Many variables involved in the pasteurization unit could be causing the trip
- Engineers lack information about the large number of automatic conditions that can cause the line to trip
- Programmers and more experienced process engineers do not have time to search for a cause
Approach
- Load and visualize the trends of the main variables associated with the pasteurization unit
- Analyze the behavior of the process during a shutdown
- Conduct value based and similarity searches to retrieve shutdown periods
- Add layers to compare different shutdowns
- Export meaningful charts to create reports and share them with the control team
Results
Overall, engineers realized 1-2% increased production, 5.4% wastes decrease, and €38,400/year in savings
- Because both lines were connected to the same main pipe before they split, opening the valve of the Line 2 branch caused pressure in Line 1 to fall below 1.5 bars
- After the pressure had dropped for 30 seconds, Line 1 shut down
- The event only occurred when the general pressure of the main pipe fell below 5 bars
- After comparing different shutdowns, engineers determined that they could adjust the automatic shutdown to happen when the pressure dropped to 1.8 bars for at least 45 seconds
- Furthermore, whenever Line 2 is started while Line 1 already is running, the pressure will increase automatically to 7 bars for one minute
- Engineers used this information to create a dashboard where they could see other areas for improvements (maximum level reached in a tank, for example)
