The Kansas Department of Transportation (KDOT) has had a Maintenance Quality Assurance (MQA) program since 1999 to evaluate the effectiveness of maintenance activities. KDOT samples 3,360 0.1-mile inspection sites annually and rates the effectiveness of specific elements in the following categories: pavement surface, shoulders, roadside, drainage, and traffic guidance. In order to allocate this workload to each of KDOT's 112 subarea maintenance offices, the inspection site selection scheme selects 30 inspection sites in each subarea. Because of the manpower-intensive nature of thiseffort, there was a desire by KDOT to examine ways to modify their process without increasing the number of sampled inspection sites. Three questions were addressed in this research. First, is the MQA selection model truly random, or is there an inherent bias in the system due to selecting aneven number of inspection sites in each subarea? Second, would the current selection process allow a buffer to be included between selected inspection sites? Third, what would be the trade-offs to changing the number ofinspection sites per subarea? By populating a database of all possible inspection sites in the state network with actual MQA data, filling in the unsampled inspection sites with estimated data, and using Monte Carlo simulation methods, it was possible to run many thousands of selection trials and to analyze the distribution of results to understand the range of the possible maintenance ratings that might result under various conditions. This Monte Carlo methodology could be applied by other agencies for the evaluation of MQA programs.
Abstract