AI-Driven Process Optimization

Particle swarm optimization, reinforcement learning, and CNN-enabled calibration pipelines are transforming FFF/FDM printing by dramatically reducing cost, time, and scrap rates while improving repeatability.

Hassan et al., Composites Part C (2024) - Section 5: Process Optimization

Key Impact Metrics

87%
Cost reduction achieved by PSO Experimenter (from $24.62 to $3.22)
97%
Parameter tuning time reduction for open-source platforms
Real-time
CNN defect detection with simulated annealing calibration

Highlighted Studies

Study Highlights
PSO Experimenter
Particle Swarm Optimization
Cut stool build cost from $24.62 to $3.22 (87% reduction) while reducing parameter tuning time by 97% for open-source FDM platforms.
Autonomous Calibration System
CNN + Simulated Annealing
Combines cameras with CNN defect detection and simulated annealing to identify clogs, layer shifts, and infill drift in real time.
Reinforcement Learning Agents
Adaptive Parameter Control
RL agents iteratively learn optimal nozzle temperatures, raster angles, and speeds, enabling repeatable builds for complex geometries and recycled feedstocks.
Process optimization is tightly coupled with quality control—many systems use feedback loops to adjust parameters based on detected defects.

See also