18 Oct 2024

Automated Change Detection in Ground-Penetrating Radar using Machine Learning in R

Abstract: Ground-penetrating radar (GPR) is a useful technique for subsurface change detection but is limited by the need for a subject matter expert to process and interpret coincident profiles. Use of a machine learning model can automate this process to reduce the need for subject matter expert processing and interpretation. Several machine learning models were investigated for the purpose of comparing coincident GPR profiles. Based on our literature review, a Siamese Twin model using a twinned convolutional network was identified as the optimum choice. Two neural networks were tested for the internal twinned model, ResNet50 and MobileNetV2, with the former historically having higher accuracy and the latter historically having faster processing time. When trained and tested on experimentally obtained GPR profiles with synthetically added changes, ResNet50 had a higher accuracy. Thanks to this higher accuracy, less computational processing was needed, leading to ResNet50 needing only 107 s to make a prediction compared to MobileNetV2 needing 223 s. Results imply that twinned models with higher historical accuracies should be investigated further. It is also recommended to test Siamese Twin models further with experimentally produced changes to verify the changed detection model’s accuracy is not merely specific to synthetically produced changes.

2 Jun 2022

Probabilistic Neural Networks that Predict Compressive Strength of High Strength Concrete in Mass Placements using Thermal History

Abstract: This study explored the use of artificial neural networks to predict UHPC compressive strengths given thermal history and key mix components. The model developed herein employs Bayesian variational inference using Monte Carlo dropout to convey prediction uncertainty using 735 datapoints on seven UHPC mixtures collected using a variety of techniques. Datapoints contained a measured compressive strength along with three curing inputs (specimen maturity, maximum temperature experienced during curing, time of maximum temperature) and five mixture inputs to distinguish each UHPC mixture (ce-ment type, silicon dioxide content, mix type, water to cementitious material ratio, and admixture dosage rate). Input analysis concluded that predictions were more sensitive to curing inputs than mixture inputs. On average, 8.2% of experimental results in the final model fell outside of the predicted range with 67.9%of these cases conservatively underpredicting. The results support that this model methodology is able to make sufficient probabilistic predictions within the scope of the provided dataset but is not for extrapo-lating beyond the training data. In addition, the model was vetted using various datasets obtained from literature to assess its versatility. Overall this model is a promising advancement towards predicting mechanical properties of high strength concrete with known uncertainties.