The purpose of the work is to develop a system for diagnosing defects (damages of the armature) of air transmission line supports based on the combination of electromagnetic control principles with the interpretation of the results of diagnosis using neuro-fuzzy modeling. It is shown that in the case of uncertainty in the diagnosis of defects of air transmission line supports in real time, it is rational to build an electromagnetic system using neuro-fuzzy modeling. Practice shows that the bars of the reinforcement bars of the transmission lines support fractures oriented perpendicularly forming, and stress-corrosion cracks - parallel to it. Optimal results of the electromagnetic control are achieved in the scan direction, which intersects the defects perpendicular to their long side. In the case of electromagnetic diagnosis, the conclusion about the defect and its size is usually based on the values of the amplitude, phase and frequency of the current signal. However, the large amount of information that comes from controlling the tens and hundreds of power lines in the field under live conditions exceeds the human ability to adequately perceive it in real time and requires automation of data analysis; this will significantly increase the speed of control, reduce errors and errors associated with the actions of the staff, will allow to coordinate the results to the object of control. Under such conditions, it is rational to synthesize an electromagnetic system for diagnosing defects of air transmission lines using neuro-fuzzy modeling. That is, create a neural network and teach it to look for defects based on real data. The most informative deterministic and probable signs of defect image are considered. It is proved that the use of hybrid neural networks leads to a significant increase in the speed of eddy current control and the reliability of recognition of defects. An analysis of the causes of emergency situations of the transmission line elements showed that 30-35% are damage to reinforced concrete supports, and the existing methods of non-destructive diagnosis of the condition of reinforcement and bearing capacity of the supports are not effective enough. This requires improvement of the non-contact electromagnetic control method. The study of the complex of characteristic parameters allowed us to select the 9 most informative deterministic and probabilistic features of the defect image. The training of the neural network was carried out using samples of artificial defects such as through-through fracture of the reinforcement according to the hybrid method of training with the error level 0 and the number of cycles 400, which allowed to improve the quality of recognition of the defect image. The correctness of network training is ensured by dividing the experimental data into test and training ones. For data modeling, 2 data blocks were used: the training sample volume was 180 datasets and 60 test sets. The information blocks were filled taking into account the features of neural network synthesis - the data should change alternately (in the direction of growth) and as much as possible fill the entire area of their values. Diagnostic inputs include transmission line voltage, resistance type, lifetime, failure statistics, climatic and geological conditions, VCT signal. It is also possible to enter additional user data and more. Combining the benefits of an electromagnetic defect diagnosis system with the use of neuro-fuzzy modeling will significantly increase the speed of eddy current control and the reliability of recognition of defects of supports.

defects of supports, eddy current control, neuro-fuzzy modeling, diagnostics

Full Text:

PDF

1. Sait «Nacional`na komisiya, shho zdijsnyuye derzhavne regulyuvannya u sferax energety`ky` ta komunal`ny`x poslug» [Site of National Commission for State Regulation of Energy and Utilities]. nerc.gov.ua. Retrieved from https://www.nerc.gov.ua/?id=39678 [in Ukrainian].

2. Klyuev, V. (Eds.). (1986). Priboryi dlya nerazrushayuschego kontrolya materialov i izdeliy [Devices for non-destructive testing of materials and products: a reference]. Moscow: Mashinostroenie [in Russian].

3. Bondarenko, V.B. (2012). Problemy` elektromagnitnoyi defektoskopiyi opor povitryany`x linij elektroperedachi [Problems of electromagnet defectoscopy poles of repeated transmission lines]. Problems of energy efficiency and automation in industry and government: XІ Mezhdunarodnaia nauchnoprakticheskaia konferentsiia (21-22 zhovtnya 2015 r.) – 21-st International Scientific and Practical Conference. (pp. 112-114). Kirovograd: KNTU [in Ukrainian].

4. Bondarenko, V.B. (2019). Pidvy`shhennya nadijnosti ta produkty`vnosti vy`xorostrumovoyi diagnosty`ky` elementiv povitryany`x LEP [Improvement of reliability and performance of eddy current diagnostics of air transmission line elements]. Energy efficiency: science, technology, application: Vseukrayins`ka nauchnoprakticheskaia konferentsiia (27 ly`stopada 2019 r.) – All-Ukrainian Scientific and Practical Conference. (pp. 8-11). Ky`yiv: NPU imeni M.P. Dragomanova[in Ukrainian].

5. Leonenkov, A.V. (2005). Nechetkoe modelirovanie v srede MATLAB i fuzzyTECH [Fuzzy modeling in MATLAB and fuzzyTECH]. Sankt-Peterburg: BVH [in Russian].

6. DSTU B V.2.6-210:2016. Otsenka tehnicheskogo sostoyaniya ekspluatiruemyih stalnyih stroitelnyih konstruktsiy. [National Standard B V.2.6-210:2016. Assessment of the technical condition of operating steel building structures]. Kyiv, Minregion Ukrai'ny, 2016. 57 p. [in Ukrainian].

Copyright (c) 2019 Vladimir Bondarenko, Kateryna Petrova, Sergiy Serebrennikov