During the last years, RIO TINTO has been involved in an UE research and development program in the field of industry 4.0 called Monsoon.
This R&D programs aims at developing digital twin model with predictive functions that could be used in a wide range of industries.
For the aluminium smelting industry, a digital twin model predicting the green anode quality was developed and implemented in an operating paste plant. Presentations at different steps of this project have been made at TMS and Forum Aluminium in 2019.
The objective of the digital twin model is to provide a support to paste plant technical staff to improve green anode density. In other words, the objective is to always achieve the best possible green anode density for a given set of raw materials.
The digital twin model operates in 3 steps:
1 – CLASSIFICATION
Real time detection of the 30 minutes’ periods of lower quality with two classes : low quality that appears in red and good quality in red
2 – IDENTIFICATION
Identification of the main 5 variables that caused the lower quality and real time information of the process engineer to help him to understand what is happening in the paste plant.
3 – OPTIMIZATION
In case of low quality, recommendation on 3 actuators to improve the situation
Actuator 1 : Pitch temperature
159.6 => 176.9°C
Actuator 2 : Heater temperature
195.0 => 201.8°C
Actuator 3 : Mixer / cooler temperature
159.0 => 153.0°C
The recommendation is based on a list of 3 actuators with set point; the possible gain in density is also given by the function.
To feed the algorithm, data is collected from an historian. The digital twin model is training itself every 8 hours using the data from the last 6 months. This learning process allows the system to self-adapt to changes in the physical world and to quickly take into account process step changes.
The project is well on tracks:
- Successful implementation of a digital twin model in an aluminium smelter
- The tool is visual, simple and auto adaptive
- The paste plan digital twin opens optimization possibilities in paste plant
This project unlocks many possibilities for the use of digital twin models and their associated predictive functions in every area of aluminium smelter.