Determine the realized process-parameter values in fused filament fabrication

Determine the actual nozzle temperature, nozzle speed, layer thickness, and other process-parameter values realized during fused filament fabrication when they differ from the values specified by the designer, in order to incorporate this input uncertainty into the Bayesian neural-network-based optimization framework.

Background

The paper distinguishes uncertainty arising from imperfect knowledge of the model from variability in the manufacturing process. It notes that process parameters specified by a designer may not be realized exactly during manufacturing, creating uncertainty between the nominal input used in the prediction model and the actual process input.

The proposed Bayesian neural network and optimization framework accounts for model uncertainty and aleatory variability learned from experimental observations, but the uncertainty associated with the discrepancy between specified and realized process parameters is explicitly excluded. Determining the actual realized values would be necessary to represent this additional input-uncertainty source in the optimization of fused filament fabrication quality.