Version 3.0Thermal response test interpretation software
A thermal response test runs for days, often unattended, on a construction site, and the standard interpretation assumes a constant heat injection throughout. TRT Analysis can also interpret a test whose injection was interrupted or varied.
Windows 10, Windows 11 and macOS. Nothing runs in the cloud.
What an interrupted test costs
When a generator trips, the grid fails or the injection stops for a few hours, the record no longer meets the assumptions of the straight-line methods. The test is then either repeated at full cost, or interpreted as if nothing had happened, with a bias that nobody has quantified.
Both options are costly, because a thermal response test is the only in-situ measurement of the ground a borehole field will be drilled into. The conductivity and the borehole resistance it returns set the required length, the undisturbed temperature sets the reference that length is sized against, and together they account for a large share of the installation's capital cost.
The software was written by a former consultant who designed, ran and interpreted dozens of these tests, and lost a few of them to an interruption. It combines two complementary methods so that an interrupted or pulsed record can still be interpreted instead of repeated, and it also shows when a test does not support a reliable result.
Calibrating a test in real time
Most TRT software fits the short-term model through an offline optimizer and returns the result once it converges. In TRT Analysis, you calibrate it yourself: six sliders drive the short-term g-function against the measured record, and the residual statistics update with each move. You can see which parameters the test constrains well and which ones it barely affects.
What it computes
Undisturbed ground temperature
- Three ways to determine it: manual entry, the circulation phase, or a logged vertical temperature profile
- A vertical profile also yields the mean annual ground temperature, read from the depth range the seasonal signal no longer reaches
- Dual-range sliders select the averaging window on the circulation phase and the profile, adjustable a step at a time with the keyboard
- All three methods reported side by side in the console, next to the value the short-term g-function currently carries on its own slider
- Sets the reference both interpretation models are read against, and the boundary condition the field is later sized to
First-order approximation
- Infinite line-source regression on the heating phase
- Independent regression on the recovery phase, as a cross-check rather than a second measurement
- Ground conductivity and experimental effective borehole resistance
- Critical time drawn on the chart, showing which data must be excluded
- p-linear average of the mean fluid temperature, alongside the arithmetic mean
- Works on all five test-borehole configurations: single and double U-loop, coaxial, standing column
Short-term g-function
- Borehole response reconstructed by a network pre-trained on 3D transient finite-element results, to within 1 × 10⁻⁴ °C
- Convolved with the heat-injection history as it was measured, so the model follows every ramp, pause and shut-off
- Standard, pulsed and interrupted tests
- Heating and recovery interpreted together, not one after the other
- Six-slider calibration with live residual statistics
- Single U-loop only
Diagnostics, new in 3.0
- Reynolds number and flow-regime flag, per phase
- Histograms of heating power, flow rate and model residuals
- Residual mean, standard deviation and RMSE
- Values read off any curve or histogram bar on hover
- Scaled cross-section of the test borehole
Reading the record
- CSV import of the test series, or paste straight from a spreadsheet, and copy the loaded table back out
- Low-pass filter for measurement noise
- Interactive selection of the circulation, heating and recovery phases
What makes it different
Two complementary methods.
The first-order approximation is the reading practitioners know and standards recognise: fit a line to the late-time temperature against the logarithm of time, take the conductivity from the slope and the borehole resistance from the intercept. It asks for a constant injected power over the window being fitted, and it applies to any borehole configuration, because the line is fitted to the mean fluid temperature and to the power per metre, neither of which depends on the number of pipes in the borehole.
The short-term g-function makes the opposite trade-off. It reconstructs the borehole's own response from the first minutes rather than only at late times, then convolves it with the power that was actually measured. Because the real signal drives the convolution, the model follows the test through its ramps and its stoppages, which is why it can interpret the early hours, the recovery, and the pulsed or interrupted records the line-source model cannot. Its limitation is the configuration: the network was trained on single U-loop boreholes only.
Each method covers what the other cannot. When a test includes both a recovery phase and a steady injection, running both methods and comparing them gives the most reliable interpretation: they use different data and different regressions, so their agreement confirms the result, and a discrepancy usually points to an unstable injection, a test cut short, or groundwater flow.
A surrogate of a physical model
The neural network is a fast, deterministic substitute for a detailed finite-element computation; it is not fitted to your data. The same inputs always return the same response. Its only role is to replace a finite-element run that would be too slow to follow a slider.
Two ways to compute the excitation
The excitation driving the convolution can be built from the two measured temperatures, or from an energy balance on the measured power and flow rate. Both describe the same physical quantity, so a marked difference between the two fits is informative in itself: it indicates that the power, the flow rate and the temperature difference are not consistent with each other, usually because of a mis-scaled flow meter or a fluid other than the one declared.
One project file, three modules
The conductivity, the resistance and the undisturbed temperature identified here are stored in the project file that the sizing module opens, and the same file is then used by the network module. Nothing has to be retyped.
Published methods
The models in TRT Analysis are published, peer-reviewed and citable. TRT Analysis is developed by a spin-off of Polytechnique Montréal, and a share of its revenue goes to the graduate students whose research it implements.
- Marcotte, D., Pasquier, P., 2008. On the estimation of thermal resistance in borehole thermal conductivity test. Renewable Energy 33(11), 2407–2415.
- Pasquier, P., Marcotte, D., Bernier, M., Kummert, M., 2013. Simulation of ground-coupled heat pump systems using a spectral approach. Proceedings of the 13th IBPSA Conference, 2691–2698.
- Dusseault, B., Pasquier, P., Marcotte, D., 2018. A block matrix formulation for efficient g-function construction. Renewable Energy 121, 249–260.
- Pasquier, P., 2018. Interpretation of the first hours of a thermal response test using the time derivative of the temperature. Applied Energy 213, 56–75.
- Pasquier, P., Marcotte, D., 2020. Robust identification of volumetric heat capacity and analysis of thermal response tests by Bayesian inference with correlated residuals. Applied Energy 261, 114394.
- Pasquier, P., Lamarche, L., 2022. Analytic expressions for the moving infinite line source model. Geothermics 103, 102413.
One application, one download
TRT Analysis
GHE Analysis
GLN AnalysisTRT Analysis is one module of Analysis Suite. TRT Analysis interprets the thermal response test, GHE Analysis designs the ground loop, and GLN Analysis balances the network that connects it. They share the heat-carrier fluid, the project record and the console.
The conductivity and the borehole resistance fitted here can be carried into GHE Analysis, and you choose which technique's value travels. A ground property measured on site does not have to be retyped to size a field with it.
Pricing
TRT Analysis is included in every Analysis Suite licence, from the smallest tier up. What changes between tiers is the size of the projects you may design, never the interpretation methods.
Free for academic research and classroom use — see the academic licence.
Questions
What does TRT Analysis interpret?▾
TRT Analysis reads a thermal response test record and returns the ground conductivity, the borehole resistance and the undisturbed ground temperature — including tests interrupted by a power outage or a stopped pump, which the standard line-source method cannot use.
Does it handle a test interrupted by a power outage or equipment failure?▾
Yes. The short-term g-function model is convolved with the heat-injection history exactly as it was measured, so it follows every ramp, pause and shut-off in the record instead of assuming a constant injection throughout.
Does the software cross-check its own results?▾
Yes, when the test allows it. The first-order approximation and the short-term g-function use different data through different regressions, so running both on the same test and comparing is a built-in consistency check, not an extra step.
Can I see which parameters the test actually constrains?▾
Yes. Six sliders calibrate the short-term g-function directly against your measured record, and the residual statistics update with every move — so you see in real time which parameter the data actually pins down and which one is only guessed.
Is TRT Analysis hard to learn?▾
No. The analysis tree on the left takes you through a project in order, one panel at a time. Every panel opens its own help page, and each manual starts with a quick-start chapter that walks through a complete analysis. The manuals can be read in the application, online or as a PDF.
