Istraživanje mogućnosti primene veštačke inteligencije u predviđanju performansi kočnog sistema motornih vozila

Abstract

Osnovni zahtevi koji se postavljaju pred današnje kočne sisteme motornih i priključnih vozila, u pogledu bezbednosti vozila i saobraćaja, se odnose na njihovo dalje unapređenje kroz razvoj novih, inteligentnih, rešenja. Suština ovih zahteva jeste da se omogući pomoć vozaču kroz inteligentno upravljanje sistemima na vozilu, odnosno njihovim performansama u različitim, dinamički promenljivim, radnim uslovima. Pošto kočne performanse vozila zavise od performansi kočnica, koje funkcionišu na principima trenja i samim tim imaju vrlo nepredvidiv karakter, i od usklađenosti tih performansi sa trenutnim uslovima prijanjanja u kontaktu pneumatika sa tlom, koji se mogu intenzivno menjati tokom samo jednog ciklusa kočenja, realizacija ovih zahteva je izuzetno kompleksna. To je osnovni razlog za sprovođenje istraživanja u pogledu razvoja i implementacije inteligentnijih načina upravljanja performansama kočnog sistema na osnovu uslova prijanjanja u kontaktu pneumatik–tlo. U ovoj doktorskoj disertaciji su istraživane mogućnosti primene tehnika iz oblasti veštačke inteligencije u cilju modeliranja složenih dinamičkih uticaja radnih režima kočnica motornih vozila i uslova u kontaktu pneumatik–tlo, kao i predviđanja ovih uticaja u cilju upravljanja performansama kočnica, a time i performansama kočnog sistema, u toku ciklusa kočenja. Zbog nemogućnosti modeliranja složenih dinamičkih uticaja radnih režima kočnica motornih vozila na njihove izlazne performanse, odnosno na vrednosti klizanja u kontaktu pneumatika i puta pomoću klasičnih matematičkih metoda, uvedena je nova inteligentna metoda bazirana na dinamičkim veštačkim neuronskim mrežama i fazi logici. U skladu sa time, u ovoj disertaciji su istraživane mogućnosti primene dinamičkih veštačkih neuronskih mreža i fazi logike u cilju modeliranja, predviđanja i inteligentnog upravljanja performansama kočnica, odnosno performansama kočnog sistema. Predmetno istraživanje je usmereno ka razvoju sposobnosti kočnog sistema ka inteligentnom prilagođavanju sile kočenja dinamičkim promenama podužnog klizanja točka (pneumatika) u kontaktu sa putem u toku ciklusa kočenja. Ovakav koncept upravljanja performansama kočnog sistema, na osnovu prethodnih i trenutnih vrednosti posmatranih uticajnih veličina i identifikovanih uslova prijanjanja tokom kočenja, podrazumeva predviđanje potrebne vrednosti pritisaka aktiviranja kočnica, na prednjoj i zadnjoj osovini, za date uslove kočenja (vrednosti pritiska aktiviranja kočnice, vrednosti brzine točka na prednjoj/zadnjoj osovini, temperature u kontaktu frikcionog para kočnice na prednjoj/zadnjoj osovini i vrednosti klizanja u kontaktu pneumatik–tlo) kako bi se u kontaktu pneumatika i tla postiglo željeno (optimalno) klizanje u podužnom pravcu.

In terms of vehicle and traffic safety, the main demands imposed to the braking systems of motor vehicles and trailers are related to their further improvement through development of new, intelligent, solutions. It could enable the driver assistance function through an intelligent control of the vehicle systems performance in different and dynamically changing operating conditions. Since the braking performance of vehicles depend on the performance of the brakes, which based their function on the friction, it is a difficult to control stochastically changed the brakes performance. Furthermore, harmonization of that performance with the actual conditions in the tire-road contact, which is also intensively changed during a braking cycle, the realization of demands towards an intelligent control the braking system performance is very complex. This is the main reason for conducting research regarding development and implementation of more intelligent ways for control of the braking system performance. In this doctoral thesis, possibilities for employing of an artificial intelligence have been investigated in order to model and predict the impact of the brakes operating regimes and the complex conditions in the tire-road contact in order to provide intelligent controlling of the braking system performance during a braking cycle. Due to the impossibility for modeling of complex dynamic influences of brakes’ operating conditions on their performance and consequently on the value of the longitudinal wheel slip using conventional mathematical methods, a new method has been introduced based on an integration of dynamic neural networks and fuzzy logic. Accordingly, this thesis investigated possibilities for the proper integration of dynamic artificial neural networks and fuzzy logic in modeling, prediction, and intelligent control of the brakes’ performance, i.e. performance of the braking system. It should provide inherent capabilities of the braking system towards an intelligent adaptation of the braking forces to the dynamic changes of the longitudinal slip ratio in the tire–road contact during a braking cycle. This concept for control of the braking system performance, based on previous and current values of observed influential factors, means predicting of the brake applied pressure values, on the front and rear axle, for the given braking conditions (brake applied pressure, wheel speed on the front/rear axle, brake interface temperature on the front/rear axle, and wheel slip) in order to achieve the desired and/or optimal slip level in the longitudinal direction. Furthermore, the braking system should continuously learn about the complex and stochastic influences between these factors during a braking cycle. Since this is especially important for commercial vehicles, the focus of research has been directed on possibilities for improving the performance of electronically controlled braking system. It is done not only to achieve the optimal value of the longitudinal wheel slip in the tire-road contact, but also enables later optimization of the lateral wheel slip.