Search

For heavy vehicle systems, within the power transmission chain from the driving machine to the working components (wheels), gears have a large share of application in the form of gearboxes, differentials and speed reducers. Their purpose is to achieve the highest possible output torque at the working components. In one heavy-duty truck, the gearbox lost its working capacity due to the failure of its vital parts: a tapered roller bearing and a helical gear. The analysis presented in this paper, conducted on the basis of experimental and analytical methods, showed that, due to inadequate installation of the bearings’ outer ring in the gearbox housing in one intermediate shaft support, the roller bearing performed its elementary function in very unfavorable working conditions of mechanical and thermal nature. Due to the damage accumulation, the roller bearing gradually lost its working capacity. As a result, the intensity of the working conditions of the entire gearbox increased. At the moment when they reached the critical value from the aspect of the gear teeth bending strength, volumetric destruction occurred, i.e fracture of the helical gear tooth, and thus the cessation of gearbox power transmission.

Predmet ove doktorske disertacije je analiza dinamičkog ponašanja kotrljajnih ležaja primenom veštačkih neuronskih mreža. Na bazi rezultata eksperimentalnog ispitivanja obučene su veštačke neuronske mreže koje su sposobne da predvide amplitude brzine vibracija ležaja. Vibracije koje ležaj generiše zavise od niza uticajnih parametara koji se mogu podeliti na konstrukcione, tehnološke i eksploatacione. Modeli dobijeni primenom veštačkih neuronskih mreža određuju zavisnosti između uticajnih parametara i amplituda brzine vibracija koje ležaj generiše. Validacija neuronskih modela izvršena je na osnovu eksperimentalnih rezultata. Analiziran je uticaj svakog parametra ležaja na amplitude brzine vibracija u karakterističnim područjima frekvencija. U radu su prikazani i rezultati međusobnog uticaja više parametara. Modelima su dobijene preporučene vrednosti uticajnih parametara ležaja. Pri analizi tehnoloških parametara uvedeni su: parametar ekvivalentne površinske hrapavosti, parametar ekvivalentne valovitosti i parametar ekvivalentnog odstupanja od kružnosti staza kotrljanja. Novouvedeni parametri omogućavaju bolje razumevanje uticaja na dinamičko ponašanje. U radu je pokazano da su neuronski modeli sposobni da na osnovu parametara ležaja predvide klasu kvaliteta ležaja., The subject of this doctoral dissertation is the analysis of the dynamic behavior of ball bearings using artificial neural networks. Based on the results of the experimental test, artificial neural networks were trained to be able to predict the amplitudes of the bearing vibration velocity. The vibrations generated by the bearing depend on a number of influential parameters that can be divided into construction, technological and exploitation. Models obtained by applying artificial neural networks determined the dependences between the influencing parameters and the amplitudes of the vibration velocity generated by the bearing. Validation of neural models was performed based on experimental results. The influence of each parameter on the vibration velocity amplitudes in the characteristic frequency ranges was analyzed. The paper also presents the results of the mutual influence of several parameters. The models obtained the recommended values of the influential bearing parameters. In the analysis of technological parameters, the following parameters were introduced: the parameter of equivalent surface roughness, the parameter of equivalent waviness and the parameter of equivalent roundness error of raceways. The newly introduced parameters provide a better understanding of the impact on dynamic behavior. The paper shows that neural models are able to predict the bearing quality class based on bearing parameters.

U ovom radu izvršeno je definisanje pojma optimizacije, njen istorijski razvoj i matematička formulacija, sa akcentom na heurističku optimizaciju. Predstavljene su grupe metoda matematičke optimizacije, a heurističke metode su detaljnije opisane. Predstavljene su prednosti metoda GA, PSO, TLBO, koje su odabrane kao reprezentativne heurističke metode. Razvijene su modifikacije, hibridna metoda i nova heuristička optimizaciona metoda. Na osnovu metode GA, njegovim modifikovanjem, dobijena je metoda iGA. Modifikacija rTLBO razvijena je na osnovu originalnog TLBO algoritma. Razvijeni su novi pristupi hibridizacije heurističkih metoda, a „redna“ hibridizacija primenjena je za razvoj hGPT algoritma. Ovaj hibridni algoritam koristi iGA, PSO i TLBO algoritme. Na osnovu heurističke pojave, razvijena je originalna optimizaciona metoda nazvana DINDI. Metoda je izvedena na osnovu eksperimentalnih ispitivanja i praćenja deteta u toku igre (dečak Dindi). Razvijen je originalni optimizacioni softver, a u softver su implementirane razvijene heurističke metode. Softver je samostalan i omogućava opštost u pogledu optimizacije. Moguće je optimizovati bilo koji problem matematički formulisan ovim softverom. On omogućuje korisniku sve potrebne aktivnosti u fazama predprocesiranja, procesiranja i postprocesiranja. Razvijene metode testirane su prema preporukama iz literature i na načine kojima je moguće potvrditi njihove kvalitete. Za testiranja su korišćene test funkcije bez ograničenja, sa ograničenjima i inženjerske test funkcije. Prema preporukama iz literature, izvršeno je poređenje razvijenih metoda, sa atraktivnim metodama iz literature. Razvijena optimizaciona metoda (DINDI) primenjena je za rešavanje praktičnih problema mašinskog konstruisanja. Za problem optimizacije reduktora, razvijen je odgovarajući oblik univerzalnog matematičkog modela i sprovedena je optimizacija za konkretne primere reduktora. Za analizirane primere postižu se značajno bolje karakteristike reduktora prim, This dissertation defines the term optimization, gives its historical development and mathematical formulation, with an accent on heuristic optimization. The mathematical optimization method group is presented, and heuristic methods are described in detail. As representative heuristic methods the advantages of GA, PSO, TLBO methods are shown. Modifications, a hybrid method and a new heuristic optimization method were developed. Based on GA, through its modification, the iGA method was created. The rTLBO modification was developed based on the original TLBO method. New approaches to hybridization of heuristic methods were developed, and “inline” hybridization was used to develop the hGPT algorithm. This hybrid uses iGA, PSO, and TLBO algorithms. Based on heuristic occurrences, an original optimization method was developed named DINDI. The method was developed based on experimental research and monitoring a child playing (a boy named Dindi). An original optimization software was developed which implements the developed heuristic methods. The software is a standalone program and allows for general use in terms of optimization. It is possible to optimize any mathematically formulated problem using this software. The software allows the user to access and use all necessary activities in the phases of preprocessing, processing and post processing. The developed methods of testing are compliant to guidelines from literature and to the methods of confirming their qualities. Unconstrained, constrained, and engineering test functions were used for testing. According to suggestions from literature, a comparison of developed methods with current methods from literature was conducted. The developed optimization method (DINDI) was used for solving practical problems in mechanical constructions. For the problem of gear train optimization a form of universal mathematical model was developed and the optimization for specific examples of gear trains was conducted. For the analyzed ex