Search academic texts

Information × Registration Number 0218U006219, 0112U008459 , R & D reports Title Identification of the individual functions of transformation of multi-parameter sensors popup.stage_title Head Sachenko Anatoly, Registration Date 03-04-2018 Organization Ternopol National Economic University popup.description2 The object of research is the process of measuring the energy illumination of ultraviolet radiation by semiconductor photodiodes. The purpose of the study is the development of neural network methods for the accurate identification of the individual functions of the transformation of multiparameter sensors. increase of accuracy of information-measuring systems of ultraviolet radiation. Methods of research - structural and functional analysis, error theory, electric circuits and neural networks, as well as methods of simulation modeling. Research results: the analysis of known methods and systems of measuring ultraviolet radiation and ultraviolet radiation sensors in the composition of the information-measuring system for determining the main directions of increasing their accuracy; the method of reduced complexity for the identification of the individual functions of transformation of a multiparameter sensor was developed and experimentally investigated with the help of the proposed model of the function of the transformation of a multiparameter sensor and its errors; the system of automated formation of the training sample for the study of the errors of the proposed method is developed; the architecture of the information-measuring system of ultraviolet radiation and the method of correction of errors of its components is developed, which provides interchangeability of sensors. Scientific novelty lies in the development of neural network method of identifying individual conversion functions multiparameter sensors on the results of verification, which provides: a) high prediction accuracy of testing results at the points where the actual calibration is not carried out, which allowed to significantly reduce the number of calibration points; b) resistance to random errors of experimental research - reduction of their influence by approximately an order of magnitude; further developed model transformation function multiparameter sensors and their errors, taking into account all the possible combinations of different types of errors for each sensor measurements or influence quantities, which allowed the formation automate the training sample error for the study of the proposed method. The results of the GDR have been tested at international and national scientific-practical conferences. Product Description popup.authors Кочан Володимир Володимирович Рощупкін Олексій Юрійович Турченко Ірина Василівна Турченко Володимир Олександрович popup.nrat_date 2020-04-02 Close