The increased demands imposed upon maintenance personnel by ever increasing vehicle complexity has markedly stressed the ability of these individuals to perform vehicle diagnostics in a timely fashion. As a response to this growing problem the vehicle industry and its supporting communities have developed sophisticated automated test equipment to support the maintenance function. The availability of test equipment capable of conveniently performing complete vehicle tests has led to a natural extension of vehicle diagnosis, in the form of vehicle prognosis. Vehicle prognosis directly addresses the issue of timely vehicle maintenance by identifying potential failures in advance of or during their occurance.The prognostic approach selected for a system is necessarily dependent on the types of failures occuring within the system. Of the three general types of failures, random, stress related and detectable, only the latter two represent cases for which vehicle prognosis is currently feasible. Prognosis for stress related failures is accomplished by monitoring system stress until internal components have completed their life cycles. Prognosis for detectable failures is accomplished by monitoring the progress of a measurable parameter as it degrades with system use. The rate at which this parameter changes with use provides information vital to the development of a prognosis. It is the latter approach which is discussed in this document.This paper describes the process of developing algorithms for diesel engine power prognosis using the specific example of a Cummins VT903 engine. It also includes a discussion for the extension of the techniques presented to other engines. Finally, the economic and technological issues affecting algorithm scope and utility are analyzed.