Abstract: The data of engine failures about tractors operating in field conditions is collected by running engine failure tracting tests. A mathematic model is established using the theory of random process, and non-homogenous process for the engine failure process, a hypothesis test is done with the model, and pa-rameters are estimated with the failure process model as well. The mathematical expression between the intervalic time of engine failures and the failure intensity function is studied, the concepts of failure process improvement and deterioration are defined, and the relation between the failure process improvement and deterioration and the failure intensity function is developed so that failure intensity curves are established for the initial operation of engine, and estimated mean time between engine failures is given. The result of research can be used as references for improvement of reliability and maintainability of tractor engines, and for establishment of maintenance strategy, and as theoretical basis for study on the interrelation between mechanical system failures as well.

Key words: Engine, Failure intensity, Non-homogeneous Poisson process, Reliability