LPBF Process Control: Feedforward vs Feedback

Introduction

LPBF process control connects heat transfer, manufacturing and control theory to improve the consistency of metal 3D printing. This guide explains how feedforward and feedback strategies regulate laser powder bed fusion, why sensor signals need careful interpretation, and how to calculate a simple corrective power command.

The topic has a timely research connection: an ASME conference paper published online on 4 November 2026 investigates hybrid thermal-history control in Inconel 718. Its preliminary findings provide a useful context for understanding the underlying engineering rather than assuming one fixed parameter set suits every geometry.

LPBF Process Control and Melt Pool Monitoring

In laser powder bed fusion, a laser melts selected regions of successive powder layers. Heat escapes through the surrounding material and build plate, but thin walls, overhangs and changing cross-sections alter those conduction paths. Consequently, identical laser settings can produce different thermal histories across a component.

Melt pool monitoring measures process behaviour using instruments such as photodiodes, thermal cameras and optical tomography systems. Engineers may track optical emission, apparent melt pool dimensions or cooling behaviour, then compare these observations with a reference associated with acceptable processing. Monitoring becomes control only when the system uses that information to change an input.

NIST’s powder bed fusion research programme includes open-platform equipment for monitoring, controls and metrology. Such testbeds help researchers connect sensor measurements with independent evidence of part quality.

Feedforward Control vs Feedback Control

Feedforward control anticipates predictable disturbances before they affect the measured output. A thermal model can identify layers likely to accumulate heat and schedule different laser powers before printing begins. Its main weakness is model mismatch: inaccurate boundary conditions or material properties can produce the wrong correction.

Feedback control responds to a measured error during processing. If a calibrated process signal rises above its target, a suitably designed controller may reduce laser power to limit overheating. Sensor noise, measurement delay and actuator limits constrain how aggressively it can respond.

A hybrid approach combines a geometry-aware baseline with sensor-driven corrections. The November ASME study combined graph-theory thermal modelling with optical tomography and commercial closed-loop control, testing six Inconel 718 parts under different strategies. The authors reported preliminary reductions in variability of surface deviation and roughness, while identifying microstructure and mechanical-property characterisation as future work.

LPBF Process Control: A Worked Example

A simplified proportional controller can use P_cmd = clip(P_ff + Kp(y_ref – y), P_min, P_max). Here, P_ff is the feedforward laser power, y_ref is the target sensor signal, y is its measurement, and Kp converts signal error into watts. The clip operation keeps the requested power between validated lower and upper bounds.

For an illustrative teaching example, take P_ff = 220 W, y_ref = 100 arbitrary units, y = 120 arbitrary units and Kp = 0.5 W per arbitrary unit. The error is -20 units, giving P_cmd = 220 + 0.5(-20) = 210 W. With assumed bounds of 180–260 W, clipping does not change that result.

This calculation illustrates corrective action, not a recommended production setting or a stability proof. Engineers must validate the signal-to-process relationship and tune the controller experimentally. Potential applications include maintaining more consistent processing around thin features, cooling channels and heat-sensitive transitions in aerospace components.

Common Mistakes in Thermal History Control

Do not equate brighter optical emission with a directly measured temperature. Emissivity, viewing conditions and calibration influence the signal, so raw intensity needs a validated interpretation. Likewise, a quieter sensor trace does not establish low porosity or adequate fatigue strength without independent inspection and testing.

A second mistake is assuming feedback can compensate for every manufacturing problem. Poor powder spreading, shielding-gas problems or sensor contamination may require interventions beyond laser-power adjustment. In examinations, distinguish the manipulated variable, measured output, reference and disturbance before proposing a controller.

Conclusion

LPBF process control works best when predictive thermal modelling and trustworthy measurements complement each other. Treat feedforward action as preparation, feedback as correction, and independent part testing as the evidence of success. Explore more mechanical engineering topics on Mechtics, or leave a question about applying these ideas to your next manufacturing project.