Key Takeaways

  • TIMA, MLA and QEMSCAN are all SEM-based automated mineralogy platforms. They scan a polished sample, fire an electron beam, and classify every mineral grain from its X-ray spectrum, at rates up to about 100,000 grains per hour.
  • All three now lean on machine-learning classifiers under the hood for particle and phase identification, not just fixed spectral lookup tables.
  • QEMSCAN grew out of CSIRO research and, like MLA, is now a Thermo Fisher product. Zeiss Mineralogic is a close alternative that explicitly markets AI-based deep-learning classification.
  • TESCAN TIMA is the newer entrant and increasingly the default choice for high-throughput exploration and geometallurgy programs.
  • Output feeds directly into liberation analysis, grain-size distribution, and the clay, hardness and recovery ML models used elsewhere in your geometallurgy workflow.

TL;DR

To do automated mineralogy with AI, send polished sample mounts through an SEM platform such as TIMA, MLA, Mineralogic or QEMSCAN instead of manual point counting. The instrument’s built-in classifier turns raw X-ray spectra into mineral, liberation and grain-size data in hours rather than weeks. You then use that table as training data for your downstream geometallurgy models.

What Is Automated Mineralogy?

Traditional mineralogy relies on a person at a microscope. A petrographer counts a few hundred points on a thin section and estimates what minerals are present. It is skilled work, but it is slow and hard to repeat consistently between people.

Automated mineralogy replaces that person with a scanning electron microscope (SEM). The beam rasters across a polished thin section or grain mount. At each point, an energy-dispersive X-ray (EDS) detector records a spectrum, and software assigns a mineral phase. As a result, you get a full map of every grain, not a sample of points. From that map the software calculates modal mineralogy, grain size, mineral associations and liberation.

How Do I Automate Mineralogy With AI?

The workflow has three parts: pick a platform, let its classifier do the phase work, and then reuse the output.

Choosing an automated mineralogy platform

You are choosing between closely related instrument families, not between an “AI tool” and a “manual tool”. Thermo Fisher offers both MLA, the mineral liberation analyzer, and QEMSCAN. Zeiss offers Mineralogic. TIMA is TESCAN’s competing platform. It has become a common default for new exploration and geometallurgy labs because of its multi-detector throughput.

In practice, the choice mostly depends on what your lab or contract lab already has installed. The outputs are functionally equivalent across all of them: modal mineralogy, liberation curves, grain-size distributions and elemental deportment.

How the classifier reads each spectrum

The “AI” part is not a bolt-on. It is the core of how modern instruments work. TIMA and Mineralogic both use trained classification models to resolve ambiguous or mixed-pixel spectra. Older lookup-table approaches often misclassified those pixels, especially at grain boundaries. Mineralogic explicitly markets “AI-based deep learning algorithms” for this step. However, the classifier is still only as good as its mineral list, so a mineralogist should review unknown phases on each new deposit.

Feeding automated mineralogy into other models

Once you have quantitative mineralogy for a set of samples, that table becomes a feature set. For example, clay model workflows and comminution hardness prediction both consume automated mineralogy output as a primary input. Recovery prediction does too. Therefore, the value of each scan grows with every model that reuses it.

What to Expect From Automated Mineralogy in Practice

Automated mineralogy is precise, but it is not cheap per sample. Sample preparation (crushing, mounting, polishing and carbon coating) takes time. Moreover, each measurement covers only a small polished area. That makes it a calibration tool rather than a way to characterize every metre of core.

Most programs therefore combine it with faster, cheaper methods. A typical pattern uses hyperspectral core scanning for continuous coverage along the hole. Automated mineralogy then provides a smaller, high-quality set of samples to calibrate those proxies. It also complements traditional petrography, which still explains textures that an SEM map alone can miss.

If you do not already have in-house SEM capacity, commercial labs (ALS, SGS, Bureau Veritas) run TIMA, MLA and QEMSCAN as a standard service line. So this is accessible without a capital instrument purchase.

FAQs

What is the difference between TIMA, MLA and QEMSCAN?

All three are SEM platforms that classify minerals from EDS spectra and report modal mineralogy and liberation. QEMSCAN and MLA are Thermo Fisher products, while TIMA comes from TESCAN. The differences are mostly in detector setup, throughput and software, not in the type of data you receive.

How long does automated mineralogy take?

The scan itself usually takes hours per sample batch, compared with weeks for equivalent manual point counting. Sample preparation adds time before the scan. Commercial labs quote turnaround in days to weeks, depending on their queue.

Does automated mineralogy use machine learning?

Yes. Modern platforms use trained classifiers to assign a mineral to each spectrum, especially for mixed or noisy pixels. The output then often becomes training data for further machine-learning models in geometallurgy.

Can automated mineralogy replace petrography?

Not entirely. It is faster and more quantitative for mineral abundance and liberation. However, a petrographer still adds value on textures, paragenesis and optical properties that an SEM map does not capture well.

Try It With Geocluster

Automated mineralogy output is dense and multi-variable. It is only useful once it is joined against your geological, geotechnical and metallurgical databases. That research and integration work is what Geocluster is designed to support. If you are building geometallurgy models on top of TIMA, MLA or QEMSCAN data, it helps you connect that mineralogy to everything else in your dataset. You can also read how the Geocluster Research Harness works.