PyelogP: Automated Energy-Based Determination of Preconsolidation Pressure in Clay Deposits

The 30-second take

• What: The authors release an open-source Python library that estimates preconsolidation pressure σ′p via Becker et al. (1987) strain energy, combining cubic splines, Kneedle knee detection, and split-point regression with OCR and curvature constraints.
• Why it matters: Abundance angle: expert graphical reading of e–log(P) curves is still scarce and shaky on disturbed soft clays. A millisecond, documented estimator is a step toward default geotech numbers — mid-horizon, not a dated building-code takeover.
• Who should care: Geotechnical engineers, lab managers processing oedometer suites, and anyone who currently treats Casagrande construction as the only σ′p story.

What the paper actually did

Estimating preconsolidation pressure (σ′p) from one-dimensional consolidation (oedometer) tests is central to settlement analysis. Traditional graphical methods such as Casagrande can be uncertain, especially on rounded e–log(P) curves from disturbed soft clays and silts.

PyelogP is an open-source Python library that calculates σ′p using the strain-energy method of Becker et al. (1987) as an automated, reproducible alternative. The algorithm combines natural cubic spline interpolation, knee-point detection via the Kneedle algorithm, and split-point linear regression in work–pressure space. Physically informed thresholds — including overconsolidation-ratio limits and second-derivative maxima d²e/d(log σ′)² — set pre-yield and post-yield fitting boundaries.

The method is evaluated on 22 experimental consolidation datasets, including Saint-Alban clay and San Francisco Old Bay Clay. Agreement with published values is reported as R² = 0.912, RMSE = 0.374, MBE = −0.080. The O(N²) algorithm takes a few milliseconds per typical oedometer curve and less than 150 milliseconds on the largest datasets in the study.

What makes this disruptive

The disruption is replacing a hand-drawn Casagrande construction — especially on rounded, disturbance-prone curves — with a documented energy-method pipeline that runs in milliseconds and hits R² = 0.912 against published σ′p on 22 clays. Reproducibility is the point: the same spline, Kneedle, and split-regression choices can be versioned.

The scarcity it touches is accurate earth-systems and foundation numbers that still depend on scarce expert eyes. Automated σ′p is a step toward more default geotechnical intelligence. It is not a new constitutive theory of clay.

Stay with the reported errors (RMSE 0.374, slight negative bias) and the clay set named (Saint-Alban, Old Bay, and others in the suite of 22).

Why it matters (outside the lab)

Abundance lens (today’s luxuries → tomorrow’s defaults): Disruptive Concepts reads climate/earth-systems tooling as a move on a scarcity map — not as a finished product.

Scarcity today: accurate monitoring and site characterization that still need scarce expert interpretation of lab curves.

If this line of work scales: cheaper modeling and measurement that makes geotechnical intelligence more widely usable. Horizon: mid-horizon; measurement practice and codes both matter.

Near-term: parallel PyelogP with Casagrande on your oedometer stack. Medium-term: code acceptance and more soil types decide whether this becomes a default. No invented year for fully automatic settlement design.

Limitations & open questions

This is a preprint. R² = 0.912 is versus published σ′p values on 22 datasets, not versus an independent mechanical ground truth. Disturbed soft-clay curves are exactly where graphical methods struggle; automation can still fit a confident-looking knee that is geologically wrong.

OCR limits and second-derivative maxima are physically motivated constraints that can also bias unusual profiles. O(N²) is fine for typical N but is not a complexity victory for huge traces. Casagrande is critiqued, not exhaustively benchmarked in the abstract.

Not yet a default: this does not demonetize geotechnical judgment on a fixed date. Codes, liability, and more soils still sit between a library and tomorrow’s default report.