TOMRA's reverse vending machines are very reliable. Yet, the machines may on rare occasions stop. In cooperation with NR, TOMRA has investigated the occurrence of these stops using reliability analysis.

Reliability analysis is a tool for evaluating a system's ability to perform a certain task. A reverse vending machine is such a system, composed by different components. The task is to identify empty beverage containers and issue an appropriate refund. The reliability of the system, i.e. the probability that the system carries out the task, can be modelled if data are available. Typical questions that can be answered are: What is the expected time between each stop and what is the uncertainty of the estimate? What is the probability that the system does not stop within a certain time frame? What happens with the system reliability if one of the components improves or deteriorates?

To answer such questions, a model for the time between system stops has to be made. Firstly, what is a stop and what is the time unit? For TOMRA's machines a stop may be that the customer cannot use the machine. Usually the number of hours between each stop is the time unit. However, for reverse vending machines the load is proportional to the number of customers. The number of containers passed through the machine between each stop is therefore a better time unit.

In order to estimate the reliability of the system, the stop times for each component have to be modelled and the parameters (with uncertainty) in the stop time distributions have to be estimated. Analytical solutions for the system reliability are often hard to find, but by utilising stochastic simulation (Monte Carlo simulation) all questions raised above may be answered. The method is extremely flexible, and gives estimates not only for (say) the expected time between stops, but also realistic estimates of the uncertainty.