Implementation of the turnaround project, as the most significant maintenance project, through the project task is extremely complex, and often ends in failure. Available literature in the form of books, resesarch and professional papers are based mostly on personal experiences and the empirical data without significant scientific methodological analysis. Such data are subjective, significant as a guidelines in the study, but scientifically not substantiated. One of the reasons of unavailability of the literature on turnaround projects is the complexity of the scientific field that combines topics: maintenance management, quality management and project management. The purpose and objectives of this study are defining the concept of quality assurance model of turnaround refinery plants project management. Defining the concept of the model is based on the six interrelated variables connected through the five hypotheses. All five of the model hypotheses were tested using logistic regression. The obtained results confirmed the interrelationship between variables, and all hypotheses were verified. The most significant contribution is defining the success for the turnaround project and definition and implementation of the complexity variable. The variable complexity is of great importance for the success of the turnaround project. The analysis of the model confirmed that the increase in the complexity decreases the success. Based on the obtained results it is concluded that the increase in the quality of individual elements of the model increases the success of project management, which ultimately has a positive effect on the achievement of the objectives ie. the success of the turnaround project. Furthermore, as part of the research conducted testing of presented quality assurance model of turnaround refinery plants project management.on the turnaround project realized during 2013 in Rijeka Refinery under the project name TA-13. Analysis of the presented model is performed on the basis of descriptive data. The analysis shows that descriptive data provide indication that the set model relations could be statistically confirmed, provided that the sample is representative and large enough.