Validation and limits¶
Point 10.2 of the Italian NTC 2018 asks the designer to assess the reliability of the code. This page summarises the material; it is a validation, not a certification.
Field of use¶
Active MASW surveys: source in line with a geophone spread at constant spacing. 1D model with horizontal layers, up to 5 modes.
Limits to assess¶
- The model is 1D: lateral changes, topography and dipping layers are not represented.
- The inversion is stochastic and non-unique: different profiles reproduce the same curve.
- Investigation depth is of the order of λ/2 of the longest wavelength reliably read. Below that the profile is extrapolated: the half-space and a 30 m Vs,eq need low-frequency energy.
- Picking is the operator's choice, and the result depends on it.
- ρ and ν are assumed; the curve is weakly sensitive to both.
- G, Ed, K, E are dynamic moduli; NSPT and Qc are indicative empirical correlations.
- Not covered: the desktop's ReMi and f-k view, passive sources, Love waves.
Test cases¶
Closed forms. Among others: synthetic wave at 200 m/s, Vs,eq of 300 m/s with 10 m at 200 and 20 m at 400 m/s, bedrock at 20 m (266.67 m/s, category E), resonance at Vs/4H.
Comparison with Easy-MASW desktop on ten projects. The f-v map matches within 10⁻⁹, misfit and error % of the desktop model within 10⁻¹², theoretical curves within 10⁻⁹ relative, and a fixed-seed run is identical. This shows the port did not change the calculation, not that the method is correct: for that, the closed forms and the literature apply.
Differences from the desktop¶
| Desktop | MASW NX |
|---|---|
| Depths summed in single precision: in one example H = 29.999998 m and category E | double-precision sums: H = 30 m and category C |
| IBM SEG-Y samples read as IEEE | correct conversion |
| Depths of the six default layers inconsistent with thicknesses | cumulated depths |
| Vs,eq computed on the saved model | computed on the solution chosen on screen |
Unchanged from the desktop: the frequency column label, offset by less than one df, and the undivided v/f in the initial estimate, which only affects the starting point.
The full document is available on request.
Found an error on this page? Let us know or open a Pull Request.