Work optimization with association rule mining of negative effective deterioration in building components

facility;condition;assessment;association rule;data mining;inspection optimization

Bartels, Louis Byron

关键词: facility;    condition;    assessment;    association rule;    data mining;    inspection optimization;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/107883/BARTELS-DISSERTATION-2020.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

The objective of enterprise building infrastructure management is to provide optimal allocation of maintenance, rehabilitation and replacement (MR&R) resources and to preserve the condition of building components over a planning horizon. While most approaches have studied it as a finite resource allocation problem, the presence of an underlying building network configuration has been largely ignored. The development of a network model of building components introduces several challenges, as well as opportunities, for MR&R decision-making and optimized building preservation, which cannot adequately be handled by the existing decision-making frameworks. One such challenge to enterprise building portfolio management is the lack of understanding for the correlation of one component’s condition state to another. Building component network-level optimization is not available as in other infrastructure domains, which makes calculating the benefit of component work activities on other building components very difficult to determine.This research focuses on using structured query language (SQL) based association rule mining to find frequent patterns of observed condition deterioration among different component types. A new metric, negative effective deterioration, is introduced which is based on actual deterioration observed from inspection data, relative to expected condition states. Frequent patterns of antecedent and consequent component pairs having negative effective deterioration states are discovered, and support and confidence factors indicate the strength of these correlations. The building component network model can improve enterprise work planning by considering the effects of negative effective deterioration on other correlated components. This new concept of network-level work optimization is used to identify antecedent component investment opportunities which have a large consequent component payoff by lowering the risk of future preventable deterioration. This process can decrease the long term deferred deficiency backlog by focusing limited MR&R resources not just on the components in the worst condition, but on those that have the most adverse effect on the condition of associated components. The proposed method can discover a large number of component correlations with negative effective deterioration, which are usually neglected by facility managers, and be able to manage the facility components in the way that predicts unnecessary added deterioration and enable rapid correlation analysis. Furthermore, the correlation of building components is used to present a new approach in determining which components in a building are most critical to require a re-inspection.