NABL & ISO/IEC 17025:2017 Calibration Reference Portal
Official Lab Services: multitek.in
Solutions for: Quality Heads & Auditors Shop Floor Inspectors Tool Room Technicians Engineering Students

In-Situ Metrology for Metal Additive Manufacturing: Why Measurement Is Becoming the Qualification Bottleneck

Updated: 13 September 2026


Summary

Metal additive manufacturing (AM) has made it possible to build complex parts layer by layer, but qualifying those parts for real-world use remains a challenge. Manufacturers can't rely solely on machine settings or a quick final inspection. The National Institute of Standards and Technology (NIST) highlights the lack of reliable, validated in-situ measurements as a major roadblock. Researchers are now zeroing in on how to track thermal behavior, melt-pool activity, microstructure, stress, powder quality, and defects as the build happens—and how to tie those numbers to actual part performance.


Additive manufacturing upends a core assumption of traditional manufacturing: in machining, you chip away at something that’s already solid. In metal AM, you’re building the part from scratch, with hundreds or thousands of thermal and material events happening across many layers. A defect can form deep inside long before it’s visible.

The big question: Can we measure enough, as we build, to actually qualify a part?

In 2026, NIST made this question the centerpiece of its roadmap for in-situ metrology in metal AM.


Why In-Situ Metrology Is Critical in 2026

Earlier this year, NIST hosted a workshop on in-situ metrology for metal alloys. They see huge potential for metal AM in aerospace, defense, energy, and medical devices—but scaling up is tough without better process understanding and qualification.

One of the main obstacles: there simply aren’t enough trustworthy, validated measurements that can track how the structure and microstructure evolve during the build. The workshop focused on everything from phase changes and microstructure to AI, data, and standards. The bottom line: AM metrology is far more than just sticking another camera in the machine.


What Does In-Situ Metrology Really Mean?

In-situ metrology is about measuring key process or material behaviors as manufacturing happens. For metal AM, that might mean tracking powder spreading, laser interaction, melting, solidification, or the build-up of each layer.

Common measurements include:

  • Melt-pool size and thermal behavior
  • Temperature distribution
  • Spatter patterns
  • Powder-bed condition
  • Phase changes and stress development
  • Signals tied to defects

It’s not about collecting endless data for its own sake—the goal is to gather data reliable enough to explain, predict, and control part quality.


Monitoring Isn’t Automatically Metrology

Not every monitoring system gives you real measurements. For example, a camera might show a bright spot around a melt pool, but that pixel intensity doesn’t necessarily translate to a real, traceable temperature.

Metrology means asking tougher questions:

  • What exactly are we measuring?
  • How is the instrument calibrated?
  • What’s the model behind the measurement?
  • What affects the outcome?
  • What’s the uncertainty?
  • Is the measurement validated?

Key distinction: Process monitoring tells you something changed. Metrology tells you what changed, by how much, and with what confidence.


Why Laser Powder Bed Fusion Is So Hard to Measure

Laser powder bed fusion (PBF-LB/M) is a measurement nightmare. The laser interacts with metal powder at tiny scales and high speeds. Melting and solidifying happens in a flash, and vapor, plumes, or ejected particles (spatter) can mess with optics.

Variables like laser power, scan speed, powder properties, layer thickness, thermal history, gas flow, and machine condition all play a role. The process is dynamic—no single set of parameters tells the whole story.


Melt-Pool Monitoring: What Are We Really Seeing?

The melt pool, the tiny spot where the laser melts the metal, is a hotspot for process insight. Monitoring systems might look at its size, shape, intensity, thermal signals, or how it changes over time.

But turning those signals into reliable physical measurements is tough. Things like emissivity, camera settings, geometry, and changing material states all affect the result. A bright spot isn’t automatically a trustworthy temperature reading.


Why Thermography Matters

Infrared thermography offers a window into the build’s thermal behavior. NIST’s AM Bench program uses in-situ thermography in controlled experiments, helping validate simulation models with real data. This is crucial, because a simulation might look right but still miss key details of the actual process. Benchmark data keeps the models honest.


NIST’s AM Bench: Raising the Bar

AM Bench is NIST’s ongoing effort to create challenging, controlled datasets so researchers can test their simulations against real-world measurements. These include in-situ thermography, surface topography, and melt-pool cross-sections.

The real value? These experiments are designed so that different models can be directly compared to the same reference, making progress measurable.


Why Model Validation Still Matters

Simulations are now common in AM, helping to estimate temperatures, stress, distortion, microstructure, and defects. But a model is only as good as the measurements it’s compared against. Validation is the check that keeps digital models grounded in physical reality.

The best approach:

  1. Measure the process with controlled metrology
  2. Compare model predictions to those measurements
  3. Identify differences
  4. Improve the model or understanding
  5. Re-validate across multiple conditions

Spatter: More Than a Visual Defect

Spatter—particles ejected during the laser process—does more than look messy. Spatter can land elsewhere and cause defects. In July 2026, NIST published high-fidelity measurements of spatter size, shape, and landing spots inside a real PBF machine. Their capture method preserved each particle’s landing site for high-res analysis.

Why does this matter? Conventional imaging can miss or mis-measure spatter, and ex-situ collection loses the link between particle and its landing spot. High-fidelity measurements help us understand what monitoring systems actually see—and what they miss.


The Hard Truth: Defect Detection Isn’t Just Anomaly Detection

Spotting an “anomaly” just means something looks off. Detecting a defect means finding a real, physical flaw. Not every odd signal means a defect, and some defects won’t show up in certain sensors.

To tie in-situ data to part quality, researchers need to correlate their signals with post-process evidence: CT scans, metallography, surface measurements, mechanical tests, dimensional checks, and microstructure analysis.


What Does Qualification-Ready Metrology Look Like?

Interesting measurements aren’t always ready for industrial qualification. To be truly useful, a measurement needs to be:

  • Clearly defined
  • Repeatable for its intended use
  • Validated against reference standards
  • Linked to real part properties or defects
  • Calibrated or well-characterized
  • Accompanied by known uncertainty or limits
  • Robust across actual operating conditions

This is why NIST’s 2026 workshop focused not just on collecting data, but on which measurements actually drive process control and qualification.


Where Does Uncertainty Creep In?

Uncertainty affects every measurement stage, from optical imaging (camera focus, distortion) and thermal sensing (emissivity, wavelength) to spatter detection (frame rate, overlap) and post-process validation (CT resolution, sampling).

Even a dashboard full of precise numbers still needs careful interpretation.


Calibrating AM Monitoring Sensors

NIST’s work includes new ways to calibrate and characterize monitoring tools in real-world conditions. That’s important, since sensors are often built into proprietary machines in harsh environments. Calibration questions include:

  • Can we calibrate the sensor as installed?
  • Does the optical path distort results?
  • How often should we check performance?
  • What’s the right reference source?
  • How do we spot drift?
  • Is calibration valid across the working range?

These are classic metrology questions—now in a much tougher context.


Can AI Fix AM Monitoring?

AI can help sort through endless streams of image, thermal, and sensor data, even spotting patterns humans might miss. But it’s only as good as its training data. If you train AI on poorly labeled defects or bad sensor data, you’ll get a bad model.

That’s why NIST’s 2026 workshop included a focus on data, AI, and digital infrastructure along with measurement science.

The strongest sequence:

  1. Develop reliable measurements
  2. Build reference datasets
  3. Connect signals to real part outcomes
  4. Train and test AI with those datasets
  5. Validate AI on independent builds

In-Situ Metrology vs. Post-Process Inspection

In-situ measurement doesn’t replace post-process inspection—it complements it.

In-situ metrologyPost-process metrology
Watches the process in real timeChecks finished parts
Reveals layer-by-layer changesEvaluates final geometry/structure
Can warn about precursorsConfirms if a defect exists
Useful for process controlUseful for acceptance/validation
Must be linked to final qualityMay not reveal when/why a defect formed

The best results come when both approaches work together.


Why Qualification Is Expensive Without Good In-Process Data

If you can't trust what happened during the build, you need extensive post-process checks: destructive and non-destructive tests, microstructure analysis, dimensional checks, and more. Reliable in-situ evidence could cut some of this uncertainty—but it’s not a magic bullet.


Where Should Manufacturers Start Measuring?

Start with the failure mode that matters most. Good measurement begins with the right engineering question, not just the sensor you have.

  • Worried about lack of fusion? Focus on signatures tied directly to that defect.
  • Concerned about residual stress? Thermal imagery alone may not be enough.
  • Is powder spreading unstable? Layer-wise surface measurements might help more.
  • Spatter contamination? Make sure imaging matches physical particle measurements.

Pick measurements that answer real questions, not just the ones that are easy to take.


A Practical Workflow for Validating AM Process Monitoring

  1. Define the defect or process characteristic you want to track.
  2. Specify the physical quantity to measure.
  3. Characterize the sensor: resolution, calibration, environmental factors.
  4. Collect synchronized process data (timing, machine settings, locations).
  5. Gather independent reference evidence (post-process or destructive tests).
  6. Establish the link: Does the signal predict the property or defect?
  7. Quantify performance: false positives, missed defects, uncertainty.
  8. Validate on new builds, not just the training data.
  9. Reassess if anything (material, machine, optics, software) changes.

What Should Quality Managers Ask About AM Monitoring?

  • What exactly does this system measure?
  • How is it calibrated or characterized?
  • What’s the spatial and temporal resolution?
  • Which defects can it catch—and which can it miss?
  • How was performance validated, and with what independent method?
  • How is sensor drift tracked?
  • How are changes in software and thresholds controlled?
  • What evidence links the signal to final part quality?

What Process Monitoring Should Never Hide

  • Poor machine qualification
  • Uncalibrated or badly characterized sensors
  • Uncontrolled powder or material variation
  • Weak post-process validation
  • Unknown measurement uncertainty
  • Software changes that aren’t documented
  • The false belief that every anomaly is a defect—or that a “normal” signal means a defect-free part

Why This Matters for Aerospace, Defense, and Medical

These sectors flock to metal AM for the geometrical freedom, but they also have the toughest requirements for proof. That means process monitoring is attractive, but validation standards are higher. What’s good enough for R&D isn’t automatically good enough for production.

The intended use always sets the bar for measurement evidence.


A Metrology Opportunity for Indian Manufacturing

Indian manufacturers in aerospace, defense, tooling, energy, and precision engineering are increasingly adopting metal AM. Metrology here is more than selling a sensor—it’s about sensor characterization, thermal measurement, calibration, dimensional validation, CT/NDT correlation, uncertainty, data validation, and support for qualification.

As AM shifts from prototyping to production, measurement capability becomes part of manufacturing capability.


The Future: Measurement-Rich Additive Manufacturing

Metal AM machines will only get more measurement-rich. Expect systems that blend thermal sensing, high-speed imaging, layer topography, machine-state data, powder-bed monitoring, AI-based anomaly detection, physics-based models, and closed-loop control.

The challenge won’t be collecting the data—it’ll be proving which data are genuinely reliable for engineering decisions. That’s why in-situ metrology has become the central bottleneck for qualifying metal AM parts.


Related Reading:


Frequently Asked Questions

What is in-situ metrology in additive manufacturing?
It’s the measurement of key process or material behaviors as the build happens. That can include thermal, optical, structural, dimensional, or other measurements needed to understand the process and final part quality.

What is melt-pool monitoring?
Melt-pool monitoring uses sensors—optical, thermal, etc.—to watch the tiny region melted by the energy source. But those signals only matter if they’re validated as real, reliable measurements.

Why is spatter important in laser powder bed fusion?
Spatter particles get ejected during the laser process and can land anywhere on the build surface. Their size, shape, and landing spots can affect the process and are key variables to watch.

What is AM Bench?
AM Bench is NIST’s benchmark program for additive manufacturing. It provides controlled measurement data and challenge problems so researchers and modelers can test and validate their simulations against real experimental results.