Improved multi-objective decision-making in manufacturing processes through uncertainty quantification and robust...

Area of Science:

• Engineering
• Manufacturing Science
• Data Science

Background:

• Manufacturing processes have complex input-output relationships, challenging optimization.
• Surrogate models approximate these relationships but struggle with uncertainty.
• Uncertainty in manufacturing can lead to suboptimal decisions and inaccurate predictions.

Purpose of the Study:

• To explore Probabilistic Graphical Models for representing manufacturing processes.
• To integrate expert knowledge with data-driven approximations for unknown relations.
• To investigate the impact of aleatoric uncertainty on multi-objective optimization.

Main Methods:

• Utilized Probabilistic Graphical Models to represent manufacturing processes.
• Integrated expert knowledge with data-driven surrogate models.
• Applied methodology to a continuous manufacturing case study.
• Employed probabilistic surrogate sampling for setpoint generation.

Main Results:

• Demonstrated enhanced Pareto front creation under non-linear process functions.
• Showcased improved decision-making by incorporating aleatoric uncertainty.
• Generated a more conservative setpoint in the continuous manufacturing case.
• Validated the effectiveness of probabilistic surrogate modeling.

Conclusions:

• Incorporating aleatoric uncertainty into surrogate modeling enhances manufacturing optimization.
• Probabilistic Graphical Models offer a robust framework for decision-making under uncertainty.
• The approach leads to more reliable and robust manufacturing optimization strategies.