A clustering-based model for rating concrete bridges using K-means technique

Bridge Maintenance, Repair and Replacement (MRR) decisions need accurate condition assessment and rating methods to ensure safety and serviceability of bridge infrastructure. In North America, the common practices to assess condition of bridges are through visual inspection. Further, the thresholds that define the severity of bridge deterioration are subjectively assigned based on the experience and judgment of the inspector or expert. The current research discusses the main deteriorations and defects identified during visual inspection and Non-Destructive Evaluation (NDE). NDE techniques are becoming popular in augmenting the visual examination during inspection to detect subsurface defects. Quality inspection data and accurate condition assessment and rating are the basis for determining appropriate MRR decisions. Thus, in this paper, a novel method for bridge condition assessment using the Quality Function Deployment (QFD) theory is proposed to develop an integrated condition rating index while identifying clear thresholds between the different ratings. The threshold identification technique is based on using K-means clustering,/c-means is one of the simplest unsupervised learning algorithms that solves the subjective determination of threshold values problem. The QFD method was applied and the developed rating index was implemented on twenty case studies in the province of Quebec. The results from the analyzed case studies show that the proposed threshold model produces robust MRR recommendations consistent with decisions and recommendations made by bridge managers on these projects. The proposed method is expected to advance the state of the art of bridges condition assessment and rating.