Timely replacement of oil in engines of modern agricultural machinery is economically feasible, since it allows to fully exhaust the resource of engine oil and ensure the safety of the internal combustion engine. One of the ways to determine the moment of engine oil replacement with sufficient accuracy is to use a digital twin, for the creation of which data on the influence of operational factors on oil degradation obtained experimentally are required. Features of conducting experimental studies during engine bench tests are analyzed in the paper. An algorithm for planning an experiment based on probabilistic methods has been developed. It is shown that a bench experiment to collect data for building a digital twin can be carried out within one period of engine oil replacement, it should be planned in several stages and consist of several series of experiments, since the moment of reaching the critical value of the base number is not known in advance, and when planning the main experiment covering 250 engine hours in two series of 10 experiments, the duration of each experiment is determined from the condition of equality of the error in determining the change in the base number. The duration of the first series of experiments was 62.7 engine hours, and the second — 187.3 engine hours. The developed algorithm allowed to build a plan of a bench experiment consisting of four series of experiments with sequential uniform filling of the factor space without repeating the engine operating modes. The resulting experimental plan allows to determine the data with the planned accuracy, minimizes the number of experiments and uniformly fills the factor space at any duration of the experiment.