Bayesian Based Strategies for Inspection And Maintenance to Optimize Structural Life-Cycle Performance

一个国家的民用基础设施对安全、生活质量和经济有着重要的影响。重要的是，基础设施，如桥梁，得到维护，使他们能够安全地进行预期的交通。目前，桥梁每两年进行一次目视检查。这一程序既昂贵又不确定（这在2007年I-35桥的倒塌中显而易见）。即使在检查或非破坏性测试之后，也必须决定桥梁需要修复、废弃或破坏的程度。可以开发新的数学工具来帮助这个决策过程。因此，这项工作的目标是通过一个合适的建模框架来解决老化问题，推进现有的计算工具，用于可持续管理和资源分配，以最小的成本有效地应对关键的，具有挑战性的社会需求，以更安全的基础设施。马尔可夫决策过程（MDP）具有很长的，在土木工程结构的风险管理和最低生命周期成本方面的成功实施历史。基础设施管理项目对MDP的成功依赖主要是因为MDP建议决策者根据实际检查或无损检测做出最佳顺序决策，他/她执行，大量检查数据可以纳入数学框架的容易性，以及最终在各种科学领域对这些方法的全面研究。MDP是受控随机过程，其中决策者不确定执行某个动作的确切效果。研究计划包括腐蚀劣化过程的部分优化MDP制定，基础设施管理算法的评估和决策工具的制定。来自CalTran的现场数据将用于验证该算法。该方法将依赖于随机控制方法和贝叶斯原理，以规划最佳的生命周期政策，并根据实际结构条件执行检查和维护操作。

A nation's civil infrastructure has significant impact on the safety, quality of life and economy. It is important that the infrastructure, such as bridges, be maintained so they can carry the intended traffic safely. Currently, bridges are inspected visually every two years. This procedure is expensive and uncertain (this was evident in the collapse of I-35 bridge in 2007). Even after inspection or after non-destructive testing, decision has to be made as to what extent the bridge needs to be repaired, abandoned, or destroyed. New mathematical tools can be developed that will assist in this decision making process. The objective of this work therefore is to approach the aging problem through a suitable modeling framework, advance the available computational tools for sustainable management and allocation of resources for degrading civil structures and respond effectively to critical, challenging societal demands for safer infrastructure at a minimum cost.Markov Decision Processes (MDPs) have a long, successful history of implementation in risk management and minimum life-cycle costing of civil engineering structures. The successful reliance of infrastructure management projects on MDPs is mainly due to the facts that MDPs advice the decision-maker to make optimum sequential decisions based on the actual inspection or non-destructive testing he/she performs, the ease in which extensive amount of inspection data can be incorporated in the mathematical framework, and finally the existence of comprehensive studies on these methods in a variety of scientific fields. MDPs are controlled stochastic processes in which a decision-maker is uncertain about the exact effect of executing a certain action. Research plan includes partially optimized MDP formulation of corrosion deterioration process, evaluation of the algorithm for infrastructure management, and formulation of decision making tool. Field data from CalTran will be used to validate the algorithm. The methodology will rely on stochastic control methods and Bayesian principles in order to plan an optimum life-cycle policy and perform inspection and maintenance actions based on the actual structural conditions.