AN INTERPRETABLE, UNCERTAINTY-AWARE FRAMEWORK FOR BRIDGE DETERIORATION PREDICTION AND RISK-BASED MAINTENANCE PRIORITIZATION
Keywords:
Bridge management systems, Deterioration prediction, Machine learning, Gradient boosting, Conformal prediction, Uncertainty quantification, Probability calibration, SHAP, Maintenance prioritization, Risk-based decision makingAbstract
Aging bridge stocks and constrained maintenance budgets make it essential to forecast structural deterioration and to prioritize interventions where they avert the most harm. Most machine-learning pipelines for bridge condition, however, emit point predictions and are coupled to prioritization rules that rank bridges solely by current condition or age, ignoring both predictive uncertainty and the consequences of failure. This paper proposes an integrated, interpretable and uncertainty-aware framework that (i) forecasts future component condition and the probability of entering a Poor state with gradient-boosted trees, (ii) equips the forecasts with distribution-free prediction intervals via split conformal prediction and with calibrated failure probabilities via isotonic regression, (iii) explains the learned deterioration drivers with SHAP, and (iv) ranks bridges with a Bridge Maintenance Priority Index (BMPI) that multiplies a risk-averse, uncertainty-inflated deterioration risk by a consequence score combining traffic exposure, network criticality, detour penalty and asset value. On a 9,800-bridge longitudinal benchmark that reproduces the schema and deterioration behaviour of the U.S. National Bridge Inventory, the classifier attains 0.84 ROC-AUC, conformal intervals achieve near-nominal 90% coverage, and isotonic calibration more than halves the expected calibration error. At a 15% maintenance budget the BMPI averts 51.6% of the consequence-weighted deterioration risk, versus 34.5% for condition-first and 21.6% for age-first ranking. The pipeline is reproducible and transfers directly to operational inventory data.References
[1] American Society of Civil Engineers. 2021 Report Card for America's Infrastructure: Bridges. ASCE, Reston, VA, USA, 2021.
[2] Federal Highway Administration. Recording and Coding Guide for the Structure Inventory and Appraisal of the Nation’s Bridges. Rep. FHWA-PD-96-001, U.S. Dept. Transportation, Washington, DC, USA, 1995.
[3] Madanat S, Mishalani R, Ibrahim W H W. Estimation of infrastructure transition probabilities from condition rating data. J. Infrastruct. Syst., 1995, 1(2): 120-125.
[4] Mauch M, Madanat S. Semiparametric hazard rate models of reinforced concrete bridge deck deterioration. J. Infrastruct. Syst., 2001, 7(2): 49-57.
[5] Thompson P D, Johnson M B. Markovian bridge deterioration: developing models from historical data. Struct. Infrastruct. Eng., 2005, 1(1): 85-91.
[6] Morcous G. Performance Prediction of Bridge Deck Systems Using Markov Chains. Journal of Performance of Constructed Facilities, 2006, 20(2): 146-155.
[7] Ranjith S, Setunge S, Gravina R, et al. Deterioration Prediction of Timber Bridge Elements Using the Markov Chain. Journal of Performance of Constructed Facilities, 2013, 27(3): 319-325.
[8] O'Connor AJ, Sheils E, Breysse D, et al. Markovian Bridge Maintenance Planning Incorporating Corrosion Initiation and Nonlinear Deterioration. Journal of Bridge Engineering, 2013, 18(3): 189-199.
[9] Wellalage NK, Zhang T, Dwight R. Calibrating Markov Chain-Based Deterioration Models for Predicting Future Conditions of Railway Bridge Elements. Journal of Bridge Engineering, 2015, 20(3): 04014060.
[10] Lee J, Sanmugarasa K, Blumenstein M, et al. Improving the Reliability of a Bridge Management System (BMS) Using an ANN-Based Backward Prediction Model (BPM). Automation in Construction, 2008, 17(6): 758-772.
[11] Kim YJ, Yoon DK. Identifying Critical Sources of Bridge Deterioration in Cold Regions Through the Constructed Bridges in North Dakota. Journal of Bridge Engineering, 2010, 15(5): 542-552.
[12] Bektas BA, Carriquiry A, Smadi O. Using Classification Trees for Predicting National Bridge Inventory Condition Ratings. Journal of Infrastructure Systems, 2013, 19(4): 425-433.
[13] Lim S, Chi S. XGBoost Application on Bridge Management Systems for Proactive Damage Estimation. Advanced Engineering Informatics, 2019, 41: 100922.
[14] Assaad R, El-Adaway IH. Bridge Infrastructure Asset Management System: Comparative Computational Machine Learning Approach for Evaluating and Predicting Deck Deterioration Conditions. Journal of Infrastructure Systems, 2020, 26(3): 04020032.
[15] Papadopoulos H, Proedrou K, Vovk V, et al. Inductive Confidence Machines for Regression. In Proceedings of the 13th European Conference on Machine Learning (ECML). 2002: 345-356.
[16] Vovk V, Gammerman A, Shafer G. Algorithmic Learning in a Random World. New York, NY, USA: Springer, 2005.
[17] Lei J, G’Sell M, Rinaldo A, et al. Distribution-Free Predictive Inference for Regression. Journal of the American Statistical Association, 2018, 113(523): 1094-1111.
[18] Lundberg SM, Lee SI. A Unified Approach to Interpreting Model Predictions. Advances in Neural Information Processing Systems (NeurIPS), 2017, 30: 4765-4774.
[19] Romano Y, Patterson E, Candès E. Conformalized Quantile Regression. Advances in Neural Information Processing Systems (NeurIPS), 2019, 32: 3543-3553.
[20] Platt J. Probabilistic Outputs for Support Vector Machines and Comparisons to Regularized Likelihood Methods. In Advances in Large Margin Classifiers. Cambridge, MA, USA: MIT Press, 1999: 61-74.
[21] Niculescu-Mizil A, Caruana R. Predicting Good Probabilities with Supervised Learning. Proceedings of the 22nd International Conference on Machine Learning (ICML), 2005: 625-632.
[22] Guo C, Pleiss G, Sun Y, et al. On Calibration of Modern Neural Networks. Proceedings of the 34th International Conference on Machine Learning (ICML), 2017: 1321-1330.
[23] Breiman L. Random Forests. Machine Learning, 2001, 45(1): 5-32.
[24] Friedman JH. Greedy Function Approximation: A Gradient Boosting Machine. Annals of Statistics, 2001, 29(5): 1189-1232.
[25] Chen T, Guestrin C. XGBoost: A Scalable Tree Boosting System. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD), 2016: 785-794.
[26] Ke G, Meng Q, Finley T, et al. LightGBM: A Highly Efficient Gradient Boosting Decision Tree. In Advances in Neural Information Processing Systems (NeurIPS), 2017, 30: 3146-3154.
[27] Chawla NV, Bowyer KW, Hall LO, et al. SMOTE: Synthetic Minority Over-Sampling Technique. Journal of Artificial Intelligence Research, 2002, 16: 321-357.
[28] Pedregosa F, Varoquaux Gaël, Gramfort Alexandre, et al. Scikit-Learn: Machine Learning in Python. Journal of Machine Learning Research, 2011, 12: 2825-2830.