Fig. 3: Decomposition of PCE model outputs for predicting repair material performance. | Communications Engineering

Fig. 3: Decomposition of PCE model outputs for predicting repair material performance.

From: Full-cycle prediction of crack healing in self-healing concrete using generalized polynomial chaos expansion

Fig. 3: Decomposition of PCE model outputs for predicting repair material performance.

First-order PCE model results showing a the distribution of polynomial coefficients, b the main Sobol index (Si) and the total Sobol index (STi) for each input variable (where a, k, r, v, m represent crack surface repair rate, crack water seepage repair rate, crack resistivity-based repair rate, crack ultrasonic velocity repair rate, and crack anti-chloride repair rate, respectively), and c a parity plot comparing model predicted values (Pv) against true measured repair rates (Tv). d Comparison of prediction errors between the first-order model and a benchmark method at curing times of 7, 14, and 28 days. e Result of dimensionality reduction for the second-order PCE model based on minimizing the Leave-One-Out Cross-Validation (LOOCV) error. Second-order model results showing f Sobol sensitivity indices (where a, k, m zrepresent crack surface repair rate, crack water seepage repair rate, and crack anti-chloride repair rate, respectively) and g the distribution of polynomial coefficients. h Enhanced comparison of prediction performance between the second-order and first-order models.

Back to article page