UC 4.1 Flotation

Boliden and ABB are actively engaged in refining this specific use case. Their joint efforts aim to optimize its functionality and explore further possibilities. This collaboration underscores their commitment to advancing the use case and driving innovation.

The flotation process is the step where the overburden from the mine is separated from the valuable and sellable minerals. Since not all valuable minerals are sold on the market, an increase in yield of the flotation process has a very direct economic impact.

Impact: The AI should have two major impacts. First, it should provide an assessment of the state of the flotation process, especially about the build up of valuable metals in the different process steps. Secondly, the AI should help to avoid different operating strategies between different shifts. If strategies are changed, recovery could be lost. The operator-AI partnership also helps share and unify strategies across operators

Desired Outcome: There are various different outputs of the AI. In a first step, the AI should give the operator an indication of the process state. Another relevant outcome are specific recommendation how to operate in the current situation increasing AI transparency and reasoning justification

User Interface

Dashboard: Grafana-based and interactive visualization to explain each forecast.

ML Model Summary

Use case: multi-step prediction (60min) to forecast sensor readings that indicate production quality, i.e., Lead, Copper, or Zink yield.

Explainer Component Overview

  • Uncertainty of prediction using quartile envelopes.
  • Precision of historic predictions regarding quartile envelopes.
  • Self attention-based input signal and time step importance matrix that indicate which input data are essential for the prediction made by the model.
  • Time step importance for each individual input signal.

Feedback Component Overview (planned)

1.Operators labeling phases of the process which operated under sub-optimal conditions



2.Using proxy model to simulate the effect of operators’ reprioritizations of feature importance.

Scroll to Top