Stochastic Modelling Of Bridge Condition For Concrete Highway Bridges

Abstract

Highway Bridges Are An Integral Part Of The Built Environment. In The Recent Years, Transport And Government Agencies Have Recognized The Increasing State Of Bridge Deterioration. Routine Inspection And Evaluation Of Existing Bridges Are Necessary To Ensure That They Remain Safe For Public Use. Traditional Inspection Techniques Have Several Drawbacks, Hence, There Is A Need For Accurate Future Prediction And Classification Of Bridge Condition To Enable Timely Intervention And Restriction Of Deterioration In Its Early Phases. This Research Study Proposes To Analyze The Significance Of Various Factors Affecting Concrete Highway Bridge Deterioration And Condition Rating, As Well As Classify The Bridge Condition On The Basis Of The Significant Parameters. Random Forest, A Machine Learning Technique Is Used In This Study To Model The Bridge Data Extracted From The Archives Of The NBI Database For The State Of Texas In 2019. The Results Obtained Suggested That Out Of The 40 Variables Inspected, 12 Were Insignificant For The Prediction Of Bridge Deterioration. RF Classification Was Applied On These 28 Significant Variables, Yielding 99.8% Accuracy. Moreover, The AUC Obtained From The ROC Space Yielded 1, Indicating Excellent Performance Potential For RF In Evaluating The Condition Of Concrete Highway Bridges.