M.S. Korovina - Teaching Assistant at the CAD Systems Design Department, The Artificial Intelligence Faculty, Emperor Alexander I St. Petersburg State Transport University E-mail: pgups2013@yandex.ru, big-bernar@yandex.ru

Aerial work platforms (AWP) are dangerous industrial objects of cycling operation, with working conditions determined by documentation standards. From statistics, it is known that AWP mechanisms show significant failure rate related to inappropriate maintenance, despite the fact that most of the AWPs don-t have any maintenance mode control equipment. The article demonstrates the approach to AWP operation modes control with the ANFIS fuzzy-logic output technology, combining the advantages of neural networks and fuzzy logics output systems. The research with Fuzzy Logic Toolbox in MatLab has outlined the correlation between linear acceleration vector of an AWP electric drive and driving torque values for steady and non-steady operation modes, using several membership functions with altered initial data sets and varied iterations for training and test data sets. Two neural network training methods - backpropagation and hybrid - were compared. The model was generated with grid partition algorithm (without clustering). Conclusions: 1. The technology of controlling the working modes of an AWP with a micro electro-mechanical (MEMS) accelerometer and ANFIS neural fuzzy-logic output MatLab module, allows to find out whether the work parameters of the AWP conform to documentation standards; for this purpose, the driving torque value is determined by the linear acceleration vector components - values with a minimal mean squared error of 0.17535 for the hybrid training method and linear output for a generalized bell-shaped membership function. 2. The minimal mean squared training error occurs when eveluating each of the components of a linear vector for five linguistic terms; and when evaluating driving torque values in electric engine for one hundred twenty-five linguistic terms. 3. The mean squared errors analysis for training and test data sets has revealed that the errors depend on the adaptive fuzzy neural output system settings and the number of iterations. 4. The analysis of an adaptive neural fuzzy logic output algorithm preview has revealed that the components of a linear acceleration vector represented in fuzzy logic output system, are not used for some output algorythms. 5. The filling for a membership function graphs for the linear acceleration vector components, showing the level of input membership to fuzzy terms of the algorithm, varies depending on the adaptive fuzzy neural output system settings. 6. The fuzzy output surface preview allows to find out the relation of one or two components of a linear acceleration vector with driving torque value, and to adjust low-energy mode MEMS thresholds. 7. ANFIS Fuzzy Logic Toolbox has a wide variety of user settings for input parameters; this technology requires further automatization.

 

1. Federalnye normy i pravila v oblasti promyshlennojj bezopasnosti «Pravila bezopasnosti opasnykh proizvodstvennykh obektov, na kotorykh ispolzujutsja podemnye sooruzhenija», utverzhdennye prikazom № 533 Federalnojj sluzhby po ehkologicheskomu, tekhnologicheskomu i atomnomu nadzoru ot 12 nojabrja 2013 g. [EHlektronnyjj resurs]. Rezhim dostupa: http://www.consultant.ru, svobodnyjj. (Data obrashhenija: 22.09.2016 g.).
2. Tekhnicheskijj reglament Tamozhennogo sojuza TR TS 010/2011 «O bezopasnosti mashin i oborudovanija», utverzhdennyjj Resheniem Komissii Tamozhennogo sojuza ot 18 oktjabrja 2011 g. № 823. [EHlektronnyjj resurs]. Rezhim dostupa: http://www.consultant.ru, svobodnyjj. (Data obrashhenija: 22.09.2016 g.).
3. GOST 18322-78. Sistema tekhnicheskogo obsluzhivanija i remonta tekhniki. Terminy i opredelenija. Vveden s 01.01.1980, dejjstvujushhijj. M.: Standartinform. 2007. 9 s.
4. Gerike B.L., Abramov I.L., Gerike P.B. Vibrodiagnostika gornykh mashin i oborudovanija. Kemerovo: KuzGTU. 2007. 187 s.
5. Barkov A.B., Barkova N.A. Vibracionnaja diagnostika mashin i oborudovanija. Analiz vibracii: Ucheb. posobie dlja studentov vuzov. SPb.: Izd. centr SPbGMTU. 2004. 156 s.
6. Barkov A.V., Barkova N.A., Azovcev A.JU. Monitoring i diagnostika rotornykh mashin po vibracii. SPb: Izd. centr SPbGMTU. 2000. 159 s.
7. Barkova N.A. Vvedenie v vibroakusticheskuju diagnostiku rotornykh mashin i oborudovanija: Ucheb. posobie dlja studentov vuzov. SPb.: Izd. centr SPbGMTU. 2003. 160 s.
8. GOST ISO 10816-1-97. Vibracija. Kontrol sostojanija mashin po rezultatam izmerenijj vibracii na nevrashhajushhikhsja chastjakh. CH. 1. Obshhie trebovanija. Vveden s 01.07.1999, dejjstvujushhijj. M.: Standartinform. 2009. 18 s.
9. GOST R ISO 10816-3-99. Kontrol sostojanija mashin po rezultatam izmerenijj vibracii na nevrashhajushhikhsja chastjakh. Promyshlennye mashiny nominalnojj moshhnostju svyshe 15 kVt i nominalnojj skorostju ot 120 do 15000 ob/min. Vveden s 01.07.2000, dejjstvujushhijj. M.: Standartinform. 2005. 16 s.
10. Popov G.A. i dr. Issledovanija ehlektricheskikh mashin postojannogo i peremennogo toka: metodicheskie ukazanija k laboratornym rabotam. SPb.: PGUPS. 2004. 99 s.
11. Burakov M.V. Nejjronnye seti i nejjrokontrollery: Ucheb. posobie. SPb.: GUAP. 2013. 284 s.
12. SHtovba S.D. Proektirovanie nechetkikh sistem sredstvami MATLAB. M.: Gorjachaja linija - Telekom. 2007. 288 s.
13. Djakonov V.P., Kruglov V.V. MATLAB 6.6 SP1/7/7 SP1/7 SP2 + Simulink 5/6/. Instrumenty iskusstvennogo intellekta i bioinformatiki. Ser. «Biblioteka professionala». M.: SOLON-PRESS. 2006. 456 s.
14. Demidova L.A., Kirakovskijj V.V., Pylkin A.N. Algoritmy i sistemy nechetkogo vyvoda pri reshenii zadach diagnostiki gorodskikh inzhenernykh kommunikacijj v srede MATLAB. M.: Gorjachaja linija - Telekom. 2005. 365 s.
15. Leonenkov A.V. Nechetkoe modelirovanie v srede MATLAB i fuzzyTECH. SPb.: BKHV-Peterburg. 2003. 736 s.
16. Matlab. URL= http://www.matlab.ru, svobodnyjj. (Data obrashhenija: 22.09.2016 g.).
17. Kruglov V.V., Dli M.I. Intellektualnye informacionnye sistemy: kompjuternaja podderzhka sistem nechetkojj logiki i nechetkogo vyvoda. M.: Fizmatlit. 2002. 349 s.
18. Rutkovskaja D., Pilinskijj M., Rutkovskijj L. Nejjronnye seti, geneticheskie algoritmy i nechetkie sistemy. M.: IPRZHR. 2004. 370 s.