From Lab to Field: What Geotechnical Tests Can Be Automated Today?

If you’ve ever babysat a triaxial test at 2 a.m. just to change a load step, you already know why automation is a big deal in geotechnics.

The good news? A lot of the boring, repetitive parts of our work can be automated now.
The bad news? Some of the truly risky decisions still need a human brain.

Let’s walk from lab, to field, to data, and then talk honestly about where judgment still rules.

---

Automation in the Geotechnical Lab: Triaxial, Oedometer, and Beyond

In the lab, automation has moved way past “just a data logger.”

Triaxial testing

Modern triaxial systems can now:

• Control cell pressure¹, back pressure, and deviator load automatically
• Run stress paths, CU/CD sequences, and cyclic loading without you standing there
• Perform scheduled B-value checks, consolidation steps, and shear at constant rate
• Log load, displacement, pressures, volume change at high resolution

What this changes in practice:

• No more manual valve tweaking every 5 minutes
• Fewer “oops, I forgot to record that point” moments
• Better reproducibility² between operators and between days

You still have to:

• Prepare the specimen properly
• Check for leaks, odd behaviors, and unrealistic curves
• Decide what test to run and why

The machine can follow a script. You still have to write the script.

Oedometer / consolidation

Automated oedometers can:

• Apply load increments³ or maintain constant stress automatically
• Track vertical displacement⁴ over time with high precision
• Follow standard loading/unloading/reloading sequences while you sleep

Key benefits:

• Much cleaner t–s curves
• Easier to pick t₅₀, t₉₀, and estimate coefficients
• Multiple specimens can run in parallel with minimal supervision

Again, automation removes the click–wait–measure–write cycle. It doesn’t replace understanding creep, sampling disturbance, or drainage conditions.

Other lab tests

You’ll also see automation creeping into:

• Direct shear – automated normal load and shear displacement
• Permeability – constant/falling head with controlled gradients and logging
• Resonant column / cyclic tests – automatic frequency sweeps and amplitude control
• Particle size analysis – laser-based systems instead of pure sieve marathons

In most cases, the pattern is the same:

The hardware gets smarter, the software logs everything…
but someone still has to decide whether the data makes sense.

---

Field Investigations Going Digital: CPT, DMT, and In-Situ Sensing

Out in the field, some techniques are born for automation. Others, not so much.

CPT & CPTu (and friends)

Cone Penetration Testing is probably the poster child for automated geotech in the field:

• Penetration is machine-controlled at near-constant rate
• Tip resistance, sleeve friction, pore pressure⁵ are all measured continuously
• Data streams straight into software for layer interpretation, qc–fs plots, correlations
• Some systems add seismic (SCPT) for shear wave velocity, or other sensors

What’s automated:

• The pushing process
• The high-frequency data logging
• Initial classification and correlation plots

Where humans still matter:

• Setting the right test locations and spacing⁶
• Interpreting weird zones (e.g. gravel pockets, organic layers, man-made fill)
• Tying CPT data back to boreholes, samples, and lab tests

DMT and other in-situ tests

Dilatometers (DMT), pressuremeters, and similar tools are also seeing:

• Automated pressure control
• Real-time plotting of p–V curves
• On-site calculation of indices and suggested parameters

Again, the grunt work is automated.
But the decision “do I trust this test in this soil?” is still human.

In-situ sensing and monitoring

Automation is also changing long-term monitoring:

• Automated piezometers, inclinometers, settlement plates linked to data loggers
• Wireless sensor networks on slopes, excavations, embankments, and foundations
• Cloud dashboards with alerts when thresholds are exceeded

This is where automation shines:

• You get continuous records instead of one reading per site visit
• Early warning for movements, water level changes, or load changes
• Much easier to compare predicted vs observed performance

But even with beautiful dashboards, someone still has to:

• Decide if the alarm is noise or real
• Plan what to do when the data says “this is moving faster than expected”

---

Data Acquisition, Logging, and Reporting: Let Software Do the Heavy Lifting

If there’s one part of geotechnics that automation can happily own, it’s data handling.

What software is already good at

• Collecting measurements at high frequency and synchronising channels
• Storing data in standard formats, with time stamps and metadata
• Plotting standard graphs (q–ε, e–log p, t–s, p’–q, monitored trends, etc.)
• Running routines: curve fitting⁷, parameter estimation, correlations
• Building templated reports⁸ with tables and charts in a consistent format

Once you’ve set up a reasonable template, the software can:

• Turn raw data into clean plots and summary tables
• Cut down hours of copying, pasting, and axis formatting
• Reduce human typo risk when transcribing values

Where you still need to pay attention

• Garbage in → beautiful garbage out
• Software can’t tell if your specimen was disturbed or if a saturation step failed
• Automated fitting might choose a “mathematically perfect” curve that ignores the physics

I like to think of it this way:

Let the software do the heavy lifting,
but always do one sanity check by hand.

Pick a few key points, run quick hand estimates, and see if the auto-calculated parameters feel realistic for that soil and test type.

---

Limits of Automation: Where Human Judgment Still Matters

It’s tempting to ask: “Can we automate everything?”
Short answer: no, and we shouldn’t.

Here’s where I’d still bet on a human:

1. Ground models and conceptual understanding

No amount of automation replaces:

• Building a ground profile from boreholes, CPTs, lab tests, and geology
• Deciding which layers are structurally important
• Recognising when the data is lying (or missing something)

Software can classify; it can’t “feel” that a sand layer is probably marine, or that a stiff clay is actually desiccated crust over a softer core. That’s you.

2. Choosing the right tests (and the right level of conservatism)

Automation can run a triaxial test.
It cannot decide:

• When a simple oedometer + shear box is enough
• When you really need advanced stress-path triaxial
• How conservative a design should be for a critical structure with uncertain ground

Those trade-offs involve risk, cost, and consequences—very human decisions.

3. Interpreting “ugly” data

Real data is messy:

• Pore pressure doesn’t always behave like the textbook
• CPT logs show odd spikes and drops
• Lab curves don’t line up perfectly across specimens

An experienced engineer can:

• Spot patterns
• Reject bad data
• Adjust parameters in a way that respects both soil behaviour and project needs

Automation cannot argue with you when the soil simply “feels wrong.”
That argument is still between your brain, your experience, and your code of ethics.

4. Communication and responsibility

Finally, no software:

• Signs a geotechnical design certificate
• Stands up in a meeting and explains risks to a client
• Goes to site when something fails and helps figure out what to do next

That’s still the job of a human geotechnical engineer.

---

Conclusion

So, what can we automate today?

• A huge part of the repetitive lab work: triaxial, oedometer, shear, permeability.
• Large parts of field testing and monitoring: CPT, DMT, in-situ sensors.
• Most of the data collection, logging, plotting, and basic reporting.

What we can’t (and shouldn’t) automate:

• Building honest ground models
• Choosing the right tests and interpreting messy data
• Balancing safety, cost, and uncertainty in real designs
• Taking responsibility for what happens if the soil surprises us

My approach is simple:
Automate anything that doesn’t need judgment, so you can spend more of your time on the parts that absolutely do.

That’s where geotechnical engineers still earn their keep.

---