The paper presents the process of genset maintenance optimization in complex organizations. In the initial stages of research, military technology maintenance methods were reviewed, and a visual inspection was conducted on approximately 25 % of gensets used in the Military Forces of the Republic of Croatia. It was found that a large number of different types of gensets were in use making application and maintenance more difficult. It is therefore suggested to reduce the number of different types. To select the genset correctly, tests have been defined for the purpose of establishing the specific genset indicators. Based on the established indicators, gensets are classified, that is, grouped according to ISO 8528. In addition to the requirements defined in Product Quality Protocols and ISO 8528, there are additional requirements related to the possibility of using the devices in international military operations where different operating modes appear. One of the additional requirements is the possibility of using F-34 kerosene fuel for powering diesel gensets. This is a specific requirement which can have a negative impact on the genset’s output characteristics. To verify the impact of kerosene fuel on the output characteristics, a test was conducted on the P-B40.R1 genset. In addition to the tests, simulations were carried out in the MatlabSimuklink software package showing that an adjustment of the genset’s parameters can contribute to minimizing the resulting difference in the genset’s output characteristics where different types of fuel are used. In complex technical devices, such as gensets, a large portion of time in the maintenance process relates to the identification of the technical condition, that is, to detecting the defect. To minimize defect identification time, an expert detection system is being developed aiming to use the evident irregularities in assembled condition to define the location of a defect which is difficult to detect and thus contribute to maintenanceoptimization. The expert system workson the basis of the Bayes method. The results of the measurements, simulations and the defect detection expert system are used to define the criteria for selecting the optimum genset types according to economic and availability criteria. The selection is carried out based on the AHP method with the application of the Expert Choice software package.