CAREER: Time-Critical Applications of Nonparametric Regression: Harnessing the Power of Information Technology for Infrastructure Systems Management

弗吉尼亚大学土木工程系提案号9984149非参数回归的时间关键应用：利用基础设施管理系统的信息技术的力量随着对基础设施系统的更大需求，土木工程师被要求开发管理系统，以使基础设施的能力得到有效利用。信息技术的进步使基础设施管理系统能够收集描述系统状态的大量数据。目前，这些数据用于制定应对当前情况的管理策略。然而，根据预测的服务需求制定前瞻性策略是可取的。非参数回归模型在支持需求预测方面具有很高的潜力。非参数回归的前提是，通过搜索过去经验的数据库，可以找到与系统当前状态相似的案例。然后，这些过去的情况可以用来预测未来的系统状态。不幸的是，搜索大型数据库的复杂性是如此之大，以至于非参数回归可能无法满足基础设施管理系统的时间要求。这项工作的目的是解决非参数回归的搜索复杂性，使其成为基础设施管理需求预测的可行工具。研究将集中在数据管理、数据结构和先进的计算方法。数据管理将被认为是试图智能地修剪数据库的记录，而不会不可接受地降低模型的准确性。我们将研究数据结构，通过创新的方式对数据进行排序和存储，从而降低搜索复杂度。先进的计算技术，特别是分布式计算，也将被探索。教育工作将侧重于培养土木工程师有效应用信息技术，这是一项关键的基础设施管理要求。将编写材料以加强土木工程课程中的信息技术教育，学生将组成跨学科小组，将采取步骤吸引和留住代表性不足的群体，并将编写基于网络的教育材料。

Brian L. SmithUniversity of VirginiaDepartment of Civil EngineeringProposal Number 9984149Time-Critical Applications of Nonparametric Regression: Harnessing the Power of Information Technology for Infrastructure Management SystemsAs greater demands are placed on infrastructure systems, civil engineers are being called upon to develop management systems to enable the efficient use of infrastructure capacity. Advances in information technology allow infrastructure management systems to collect large quantities of data describing the state of the system. Currently, this data is used to develop management strategies that react to current conditions. However, it is desirable to develop proactive strategies based on forecasted demand for service. Nonparametric regression models have high potential to support demand forecasting. The premise of nonparametric regression is that by searching a database of past experiences, one can find cases that are similar to the current state of the system. Then, these past cases can be used to predict the future system state. Unfortunately, the complexity of searching a large database is so great that nonparametric regression may be unable to meet the time requirements of infrastructure management systems. The purpose of this effort is to address the search complexity of nonparametric regression, making it a viable tool for infrastructure management demand forecasting.The research will focus on data management, data structure, and advanced computing approaches. Data management will be considered to attempt to intelligently prune records of the database without unacceptably degrading the accuracy of the model. Data structures will be investigated that will reduce search complexity through sorting/storing the data in innovative ways. Advanced computing technologies, particularly distributed computing, will also be explored. The education effort will focus on preparing civil engineers to effectively apply information technology, a critical infrastructure management requirement. Materials will be developed to strengthen information technology education in the civil engineering curriculum, students will work on interdisciplinary teams, steps will be taken to attract and retain underrepresented groups, and web-based educational materials will be developed.