Real-Time Feedback-Enabled Simulation Modeling of Dynamic Construction Processes

根据美国人口普查局的数据，2015年，美国建筑业的支出将超过1万亿美元。建筑和基础设施项目由人员、设备和材料相互连接的网络组成。大多数情况下，找到最佳的工作策略，并做出及时的操作决策，从而在最小化项目完成成本和时间的同时实现最大的生产力，这不是微不足道的。与制造和工业系统不同，建筑项目涉及动态（不断发展的）布局、复杂的资源交互、工作流程和过程中的不确定性，以及可能导致偏离计划和不必要的延迟的不可预见的条件。数据显示，只有30%的建设项目按时并在预算范围内完成。因此，建设项目运营层面决策的准确性和及时性至关重要。该奖项支持基础研究，通过将过程级数据无缝集成到决策中来减少不确定性，从而提高施工决策的准确性。这将通过建立理论基础和显著推进建筑仿真建模的现状来实现，通过在真实项目发展时实现与仿真模型的实时交互，并通过反馈回路传达仿真输出以引导真实项目中的事件。因此，这项研究的结果将有利于美国经济和社会，因为它导致更好的决策，从而减少浪费，返工，成本，时间，并确保安全。此计划的多学科性质，有助扩大代表性不足及多元化的学生群体参与综合研究及教学活动，并对工程教育产生积极影响。本项目中基于知识的仿真建模框架使过程级模型能够自主学习并适应不断变化和发展的建筑系统。作为这种仿真模型输入的过程级知识是从无处不在的感官数据中获得的，这些数据描述了现场过程的关系、相互作用和不确定性属性，并能够生成和维护更准确的仿真模型。在这样做的过程中，要实现这一框架的充分认可和应用，还需要克服一些科学障碍。研究团队将设计和测试从数据挖掘、机器学习、预测和控制中提取的方法，以填补从设备和人员交互中捕获和挖掘复杂数据和元数据的现有知识空白。由此产生的过程级知识将足够丰富，可以在任何时间点描述、建模、分析和规划施工系统的不确定性，从而帮助调整工作现场的资源分配和操作场景。

According to the U.S. Census Bureau, in 2015, the U.S. construction industry will surpass $1 Trillion Dollars in spending. Construction and infrastructure projects consist of interconnected networks of people, equipment, and materials. Most often, finding optimal work strategies, and making timely operational decisions that lead to maximum productivity while minimizing project completion cost and time is not trivial. Unlike manufacturing and industrial systems, construction projects involve dynamic (constantly evolving) layouts, complex resource interactions, uncertainties in workflows and processes, and unforeseen conditions that can result in deviations from plans and unwanted delays. Figures show that only 30 percent of construction projects finish on time and within budget. Therefore, the accuracy and timeliness of operational-level decision-making in construction projects is of utmost importance. This award supports fundamental research to enhance construction decision-making accuracy by reducing uncertainties through the seamless integration of process-level data into decision-making. This will be achieved by building the theoretical foundation and significantly advancing the current state of construction simulation modeling through enabling real-time interaction with a simulation model as the real project evolves, and communicating the simulation output through a feedback loop to steer the events in the real project. Therefore, results from this research will benefit the U.S. economy and the society since it leads to better decision-making which results in reducing waste, rework, cost, time, and ensures safety. The multi-disciplinary nature of this project will help broaden participation of underrepresented and diverse student groups in integrated research and pedagogical activities, and positively impact engineering education.The knowledge-based simulation modeling framework in this project enables process-level models to autonomously learn from and adapt to ever-changing and evolving construction systems. Process-level knowledge that serves as the input of such simulation models is obtained from ubiquitous sensory data that describe relationships, interactions, and uncertainty attributes of field processes, and enable the generation and maintenance of more accurate simulation models. In doing so, some scientific barriers are yet to be overcome to realize the full accreditation and application of this framework. The research team will design and test methods that draw from data mining, machine learning, forecasting, and control to fill the existing knowledge gaps in capturing and mining complex data and meta-data from equipment and human crew interactions. The resulting process-level knowledge will be rich enough to describe, model, analyze, and project the uncertainties of construction systems at any point in time and consequently help adjust resource allocations and operational scenarios on the job site.