The dissertation research deals with solving the problem of assessing and reducing uncertainty in the problems of automotive technical expertise of road accidents. The methodological basis of the thesis is the consistent application of a systematic approach, analytical and experimental research methods, modern scientific and technical means at all stages of the reconstruction of a road transport accident. The concept of the dissertation research is that the decision-making process in the automotive technical expertise of road accidents should be considered not only as deterministic, but also as a stochastic and uncertain process that requires the application of synthesis of deterministic, probabilistic, regression and neuro-fuzzy models for taking into account most of the factors that affect the reducing uncertainty in the formation of expert opinions.
A general concept for assessing and reducing uncertainty in the problems of automotive technical expertise of road accidents based on the application of the methods of the theory of fuzzy sets and neural networks, the morphological method and modular synthesis technology has been created. The high flexibility of the automotive technical expertise system is provided when its operating conditions change. A method has been developed and an information model has been built for assessing the quality of investigations and conducting automotive technical expertise of road traffic accidents, on the basis of which algorithms for adaptive decision-making processes with deterministic and probabilistic characteristics are synthesized and the technology of choosing the optimal alternative under uncertainty is substantiated.
The application of the methods of the theory of fuzzy sets, neural networks and experimental research methods has improved the method for estimating the coefficient of adhesion of car wheels with a road surface, which, unlike existing ones, allows to determine the value of the coefficient of adhesion for a wide range of modern cars more accurately, which, in turn, improves the quality and reduces the subjectivity of establishing the technical ability to avoid a traffic accident; a method for assessing the distance of visibility of objects when driving a car in the dark, which, unlike the existing ones, makes it possible to simplify the time-consuming procedures for studying safe modes of car movement in the dark, reduce the time spent by an automotive expert for an in-depth analysis of road accidents, increase accuracy and expert opinions objectivity. Based on the methods of differential and integral calculus, mathematical statistics, probability theory and regression analysis, methods for assessing the distance traveled by a car have been further developed, which take into account the influence of the design of the braking system and the psychology of driving a vehicle on emergency braking indicators in the investigation of road accidents; methods for analyzing car maneuvering by taking into account uneven movement (maneuvering with braking). Based on the method of fuzzy cognitive maps, it was proposed the model for assessing the reliability and safety of a vehicle in the driver-car-road-environment system (DCRE), which allows for taking into account the mutual influence of factors determining the reliability of a man-machine system, rank them and develop measures for improving traffic safety. Possible areas of application of the method are human-machine systems with a continuous nature of human activity (the DCRE system belongs to them), in which there are no clear boundaries between the operations performed, which makes it difficult to collect statistics on the probabilities of their correct execution.
To evaluate the effectiveness of introducing a new method for establishing a particular parameter in the technical expertise of road accidents is possible basing on determining the difference in the uncertainty of the system in two states A and B (A is the state of the system when the current method is used, B is the proposed one). In this case, the efficiency criteria can be: the relative difference between the normalized entropies in states A and B, the relative value of the decrease in the dispersion range of the desired parameter, the relative value of the reduction in the probability of errors of the first and second kind. The proposed approach to assessing the uncertainty of the methodological support of automotive technical expertise can increase the objectivity of the formation of expert conclusions in the process of determining the circumstances of emergencies. Compared to the deterministic approach, accounting for stochastic uncertainty narrows the range of possible solutions by 26-28%, additional accounting for fuzzy uncertainty allows improving results by another 20%.
