Simulation Study, Neural Network-based Chain Lubrication

In this thesis, a Chain Dropping Lubrication System designed by a foreign company is analyzed and researched by neural networks. This system is designed ingeniously and reliably. In comparison to traditional methods, the system can lubricate effectively and save lubrication greatly. So it is very important to do the research to this system by simulation for system optimization and similar system design. It is proved that the oil drop rate of the system is very important for lubricating effect. So, firstly, the system is analyzed by mechanics and hydrodynamics to investigate that the main factors, which influence the lubrication drop rate, especially focus on ambient temperature, lubrication type, taper angle and control valve position of the system.Secondly, after studying the mapping capacity of feed-forward neural network for complicated system and utilizing the BP and RBF neural networks to be trained by some selective arithmetic and experimental results, the characteristics of lubrication drop rate vs. main factors model is established. Based on the BP networks model, which is optimal between two networks, the system and the structure design that the taper angle of the system's oil valve is adjusted will be simulated and applied to find which angle will be brought out best performance.Microsoft Visual C++ 6.0 and MATLAB engine calculation is adopted for the implementation of the simulation program. The comparison of network simulated results and experiment output shows the results are satisfactory.