Development and evaluation of motor carrier safety procedures

This research is an attempt to improve roadside inspection procedure performance in selecting potential out-of-service (OOS) vehicles for inspection. Several analytical models based on historical data on safety violations by motor carriers were developed and may be employed in the development of an “expert” system to actually be used in the field by roadside inspectors. The potential for using the Inspection Selection System (ISS) software for pre-inspection screening was also evaluated.; Initially, the distribution of OOS vehicles at the carrier level was analyzed using cross-classification tables. Tables using carrier OOS Rate, carrier Violation Rate and Carrier Size showed a reasonable degree of correlation between OOS Rates and Violation Rates and the potential for using large carriers as an explanatory variable.; Two analytical methodologies were employed in the development of the models to be used in the “expert” system: Binary Logistic Regression and Direct Cross-Classification. Both the regression and the cross-classification models were developed using historical inspection data merged with prior carrier data on OOS Rate, Violation Rate and Carrier Size. Reasonable regression models were calibrated with OOS Rate and Violation Rate as the primary independent variables and Carrier Size included as a dummy variable for large carriers. Two simple cross-classification models were developed using OOS Rate and Violation Rate for separate tables.; Carrier Size produced only small improvements in the performance of the models. The cross-classification models performed about the same as the comparable regression models but not as well overall as the regression model with both OOS Rate and Violation Rate. The cross-classification models are recommended for implementation since no calibration is needed.; The ISS software was effective in identifying additional OOS vehicles. The visual pre-inspection screening by inspectors also appeared to be effective.