Statistical Determination of Minimum Inspection Intervals

We were tasked to provide emergency statistical and analytical support to the passenger car dept. of a major subway system.  Working in conjunction with in-house mechanical engineers, our analyses help to classify each type of mechanical failures, determine expected time to failure for each type and the implied minimum inspection intervals, and helped to allow the intermittent return to service of the beleaguered fleet until trucks of a new design could be fitted.  The then-new subway cars utilized a lightweight bogie that was the first production build of a new design.  However, soon after the fleet entered service, cracks began to appear in the trucks at frequent intervals, which resulted in the entire fleet being put out of service pending an inspection.  When inspected, more than 200 cracks of various types were found on these nearly-new trucks.  Initially each car in the fleet needed to be inspected every 48 hours to ensure that no unsafe condition could go undetected.  However, the mechanical engineers kept track of each inspections and the defects found, which, over time, provided a dataset that allowed us to predict how long the car could be kept in service safely and with confidence before requiring another truck inspection.  The cracking issues, however, accelerated as the fleet aged, such that we were not really able to reduce the inspection interval to acceptable levels.  During the last part of this project, we focused on assisting fleet managers in reducing the use of this specific type of car until the replacement trucks arrived on the property.  Our analyses likely prevented several technical incidents involving this type of trucks by providing quantitative evidence showing that the trucks must be inspected at frequent intervals despite the subway system gaining more practical engineering with this specific issue.  Statistics is not always intuitive, and despite developing a comfort level with a common problem it is sometimes necessary to examine the data critically to ensure that the organization does not take on more operational risk than it is aware of.