SalomeCode_AsterCustom ML

Hybrid Physics+AI Methodology

Proprietary dual-model system combining improved SRSS physics with custom gradient boosting regression. Trained on 560 Salome/Code_Aster FEM simulations with bolt-type-specific calibrations.

560
FEM Simulations
96.4%
Classification Accuracy
6.95%
Resistance MAE

FEM Training Infrastructure

🔬

Salome Platform

Open-source CAD and mesh generation. Procedural geometry construction via Python scripting. Parametric bolt patterns, plate geometry, and weld definitions.

Geometry Engine
⚙️

Code_Aster

Advanced nonlinear FEM solver. Material plasticity, contact mechanics, bolt pretension modeling. EN 1993-1-8 material models.

Structural Solver
🤖

Custom ML Pipeline

Proprietary in-house development. Dual gradient boosting regressors with physics-informed feature engineering. 13-algorithm competition per bolt type.

Our Technology

Three-Stage Prediction Framework

01

Improved SRSS Physics Foundation

UC² = Σu²ᵢ + α·uVz·uMy + β·uN·uVz + γ·uVy·uMz + δ·uN·uMy + ε·uN·uMz + ζ·uVz·uMz + η·uVy·uMy

Enhanced Square Root Sum of Squares (SRSS) with 7 fitted coupling coefficients models multiaxial load interactions. Each bolt configuration (Pretension μ=0.5, Snug-Tight μ=0.0) employs calibrated coefficients optimized via least-squares regression on FEM ground truth.

Dual Calibration: Pretension (Ncap=302.9kN, α=0.42, β=-0.31...) | Snug-Tight (Ncap=303.8kN, α=0.39, β=-0.28...)
02

Custom ML Residual Learning

ΔR = fML(uN, uVy, uVz, uMy, uMz, ...) where ΔR = RFEM − RSRSS

Proprietary gradient boosting regressor developed in-house learns systematic deviations between physics baseline and FEM truth. 23 engineered features (14 load ratios + 8 physics-derived + 1 SRSS prediction) input to ensemble model. 13 competing algorithms (Ridge, ElasticNet, SVR, GBM, RF, KNN, GPR, etc.) evaluated via 5-fold cross-validation per bolt type.

Training Dataset: 280 Salome/Code_Aster cases per model • Latin Hypercube Sampling • 13-algorithm tournament selection
03

Hybrid Prediction Synthesis

Rfinal = RSRSS(7 coeffs) + ΔRML(GBM ensemble)

Final resistance combines calibrated physics baseline with learned correction term. Bolt-type-specific routing ensures Pretension and Snug-Tight connections receive appropriate coupling coefficients and ML corrections. Classification layer (Random Forest, 200 trees) maps continuous resistance to PASS/FAIL categories with 96.4% accuracy.

Output Pipeline: Physics (14.7% MAE) → ML Correction → Hybrid (6.95% MAE) → Classification (96.4% accuracy)

Salome Parametric Geometry

HEA120-HEA120 beam-column connection with 6×M20-8.8 bolts, 15mm S355 endplates. Geometry procedurally generated via Salome Python API for Code_Aster meshing.

  • Beam: HEA120 (h=114mm, b=120mm, tf=8mm, tw=5mm)
  • Bolts: 6×M20 Grade 8.8 (fub=800 MPa, As=245mm²)
  • Plates: 15mm endplate, 5mm stiffeners, S355 steel
  • Code: EN 1993-1-8 (Eurocode 3 Part 1-8)
Building 3D Geometry...
Interactive 3D model • Salome procedural generation • Code_Aster FEM-ready mesh

FEM Training Data Generation Protocol

📐

Salome Geometry

Parametric Python script generates HEA120-HEA120 connection. Bolt holes, endplate thickness, weld geometry defined procedurally.

🔲

Mesh Generation

Salome automated meshing: 8-node hexahedral elements for plates/beams, refined mesh near bolt holes and welds.

⚙️

Code_Aster Solve

Nonlinear FEM analysis: material plasticity (ε,pl ≤ 5%), contact mechanics, bolt pretension (Pretension: 155kN, Snug-Tight: 0kN).

📊

Ground Truth

Extract resistance %, bolt forces, plate von Mises stresses, plastic strains from .rmed output files.

165 Load CasesLatin Hypercube Sampling across 5D load envelope
2 Bolt ConfigsPretension (μ=0.5) + Snug-Tight (μ=0.0)
330 Total Sims165 cases × 2 bolt types = 330 Code_Aster runs
560 Training Points280 per model after augmentation and validation split

Validation & Performance Metrics

Classification Performance

96.4%
Pass/Fail Accuracy

Random Forest classifier (200 trees, max_depth=15) achieves 96.4% accuracy across 560 FEM cases. Only 6 borderline misclassifications (100-105% zone).

Regression Performance

6.95%
Mean Absolute Error

Hybrid model reduces MAE from 14.7% (physics alone) to 6.95% (physics+ML). 53% error reduction via custom gradient boosting residual correction.

Critical Zone Accuracy

8.1%
Avg Error (85-115%)

Borderline design zone (85-115% resistance) shows 8.1% average error. FEM verification mandatory for final approval in this range.

Performance by Resistance Zone

Resistance RangeAvg ErrorMax ErrorRecommendation
Fail (<85%)4.2%12.8%High confidence - no FEM needed
Critical (85-115%)8.1%15.2%FEM verification required
Pass (115-200%)3.9%9.4%High confidence - safe for preliminary design
Deep Pass (>200%)11.6%25.3%Use classification only (PASS reliable, R% less accurate)

Technical Implementation Details

ML Architecture

  • Classifier: Random Forest (n_estimators=200, max_depth=15, min_samples_split=5)
  • Regressor: Gradient Boosting (n_estimators=200, max_depth=4, learning_rate=0.1)
  • Features: 23 total (14 load ratios + 8 physics-derived + 1 SRSS baseline)
  • Validation: Leave-one-out + 5-fold cross-validation
  • Development: Custom in-house implementation (not off-the-shelf)

FEM Configuration

  • Solver: Code_Aster (open-source structural FEM)
  • Geometry: Salome Python API for parametric CAD
  • Material: S355 steel (fy=355 MPa, fu=510 MPa, E=210 GPa)
  • Plasticity: Von Mises yield criterion, εpl,max ≤ 5%
  • Contact: Friction μ=0.5 (Pretension) or μ=0.0 (Snug-Tight)

Capacity Reference Points

  • Ncap = 302.9 kN (axial tension, Pretension)
  • Vy,cap = 92.0 kN (shear Y-axis)
  • Vz,cap = 111.8 kN (shear Z-axis)
  • My,cap = 27.6 kNm (moment about Y)
  • Mz,cap = 12.5 kNm (moment about Z)

Code Compliance

  • Design Code: EN 1993-1-8 (Eurocode 3 Part 1-8)
  • Material Standard: EN 10025 (structural steels)
  • Bolt Standard: EN ISO 4014/4017 (hexagonal head bolts)
  • Safety: Partial factors γM0=1.0, γM2=1.25
  • Validation: FEM results match EN 1993-1-8 analytical checks

Engineering Usage Protocol

ConnectAI is a preliminary design screening tool, not a replacement for code-compliant FEM analysis. Follow this protocol for responsible engineering practice:

Preliminary Design (Screening)

Use AI predictions to rapidly evaluate 50-100 design alternatives. Eliminate clear failures (R < 85%) and identify promising candidates (R > 115%). No FEM needed at this stage.

⚠️

Design Optimization (Iteration)

Iterate quickly with AI (10-20 variants/day), then run Salome/Code_Aster FEM on top 2-3 finalists. Verify borderline cases (95-105%) with full nonlinear analysis.

🔒

Final Design (Approval)

Always verify critical connections with full FEM before approval. High-consequence structures, borderline resistances, or loads near capacity limits require Code_Aster validation and PE sign-off.

Disclaimer: ConnectAI provides preliminary estimates for feasibility studies and conceptual design. Final structural design must be verified by licensed professional engineers using code-compliant FEM software (Salome/Code_Aster, IDEA StatiCa, or equivalent) per EN 1993-1-8 requirements.

Experience the Hybrid AI Engine

Test real HEA120 connection with Salome-generated geometry and custom ML predictions