1 Advances in Hydrodynamic Journal Bearing Research
Hydrodynamic journal bearings (HJBs) are excellent devices able to support rotating shafts under the action of loads with negligible friction and good damping properties. Furthermore they are: passive, maintenance free, cheap and easy to realize. HJB are widely used in ICE’s facilities, such as crankshaft and turbocharger, and when, more in general, rotating shafts have to sustain loads (hydropower, power generation, naval facilities, process industry, chemical industry). In some applications, for their characteristics, HJB constitutes the only available choice for designers but in the majority of the cases the mix with all the properties is the key of the success of these mechanical components. Aspects like: manufacturability, reliability, cost and working precision are increasing their importance respect to performances nevertheless they rarely appear in the design procedures. Modern studies about sliding bearing market [1-2] show that in the near future the sliding speed will progressively increase while the dimension of the machine will decrease and consequently the load is going to decrease. Furthermore major companies (SKF, NTN SNR, Alstom and Siemens) presented several papers about their clear will to use water as a lubricant (NASA [3] and de Kraker et al. [4]) and polymeric materials for the sleeve coating. Nevertheless, the design, of this mechanical components, is regulated by the Raimondi-Boyd charts [5] and by some national regulations (ISO [6] and DIN [7]) which do not allow a big margin of customization and adaption of the product to the real design needing. The global aim of this work is to solve the multi-objective and multivariable problem of HJB design in a fast and industrially feasible way. The solution flow has also to be flexible to accommodate future modification, take into account different materials, lubricants and geometrical designs. The local aims are: solve the RE PDE using numerical methods and Artificial Neural Networks, devise a series of objective and implement code for their minimization with Genetic Algorithms and Artificial Bee Colony Algorithm, design and build a test rig machine to collect data in order to validate 3D CFD model developed to predict thermal behaviour of journal bearings and finally demonstrate with two different study case the effectiveness of the method.
........
2 Theoretical Models for HJB Performance Calculation
2.1 Introduction
Within this chapter the theoretical model used for performance calculations is presented. The theory adopted is an efficient revision of previous works combined together to simplify the performance calculations without loosing accuracy. The core of the calculation is the use of a smoothing agent in the finite difference scheme that provides a natural diagonalization of the system matrix with consequent simplification of the iterative method needed to extract the solution. Furthermore this chapter presents results obtained with the proposed method and series of comparisons are made with the major works presented in literature.
2.2 Fundamentals of Fluid Lubrication Theory
A good and complete explanation of how a viscous liquid film can provide lubrication between two moving and loaded surfaces has been given by Reynolds in [65] who also derived an equation to describe the behaviour of this lubricating film. This equation substantially derives from the Navier-Stokes equations of momentum and continuity or, in a more practical way, it can be found considering the equilibrium of an element of fluid subjected to shear, due to viscous force, and applying to this element the continuity principle. It is clear that the two surfaces involved in the problem have to move with a sufficient relative speed in order to generate shear forces and their geometrical shape is made to allow film height variation which is necessary for the generation of pressure gradients. In addition, it can be assumed that between the two surfaces there is a continuous film of lubricant and in any case there would be a system providing lubricant supply in any load and speed condition. There are other 5 important assumptions for the calculations: Pressure is constant trough the film thickness which is valid since the film is in the range of several micrometres. Body forces and fluid inertia is neglected. Viscosity and density are constant in the fluid domain. These are just a first step assumption to derive the equation. Lubricant behaves like a Newtonian fluid that is always valid except for some polymeric oils.
3 CALCULATION TOOLS: ARTIFICIAL NEURAL NETWORKS, GENETIC ALGORITHMS AND ARTIFICIAL BEE COLONY ............ 41
3.1 Introduction ................................ 41
3.2 Artificial Neural Network ................................. 41
4 FLUID DYNAMIC SIMULATIONS ............ 55
4.1 Introduction ................. 55
4.2 Model ...................... 55
4.3 Geometry .................... 57
4.4 Calculation Grid .................... 59
5 OBJECTIVES OF THE OPTIMIZATION .................. 71
5.1 Introduction ........71
5.2 Objective Formulation ....................... 71
5.3 Optimization Parameters and Procedure ...... 75
7 Results and Discussions
7.1 Introduction
In this chapter the results of the study are collected. In chapter 7.4 the results of the experimental measurements are compared with the relative simulations and calculations. In chapter 7.5 some indications from CFD simulations are given. Then, in chapter 7.6 and chapter 7.7 two examples of solution of the algorithm for two different kind of HJB, one with round inlet and one with square inlet, are presented. In Table 7-1 and Table 7-2 a short resume of the techniques used and their properties is presented. The results appear in the following order: Experimental results and CFD comparison. CFD results for HJBs optimization. Optimization results of two study cases. Optimization results of a different objective function.
7.2 Results of the Experiments and CFD Validation
The results of the experiments come out in form of graphs in which the measured friction torque and temperatures are compared with the calculated and simulated values. In Fig. 7-1 temperature e profile around the bearing is presented. The red dots are 360 one for each degree around the line generated from the intersection of the bearing surface and the symmetry plane. The angular position is counter clockwise direction on y’ axis. The experimental data are four for each angular position (45°, 135°, 225° and 315°) because four tests have been carried to ensure measurement repeatability. In Fig. 7-2 the inlet temperature is 25°C (298 K), the maximum temperature is 36.1 °C, the distribution is a bell shaped function. The experimental data are in good agreement with the simulation which is carried out for speed 1000 rpm and 1.5 kN. The maximum measure temperature is 37.15 °C and in general the measured data re higher respect to the simulated. When the speed is increase to 2000 rpm the temperature is much higher than the previous case. Is important to point out that according to current work measurements the speed has a big effect on the temperature rise while the load has minor ability in increase the heat produced by friction. This is crucial because demonstrates again the need of improve the design methods used to shape journal bearings because the current trends show that in the future the speeds of machines will progressively increase and their dimensions will decrease. Also in this case the curve has a bell shape with a “nose” more flat the previous. Lubricant inlets is still at 25 °C and the maximum temperature is 46.3 °C simulated and 51.4 °C measured. The increased speed also tend to translate the curve on higher angular positions, this means that the fluid is pumped with an increased force but the hydrodynamic pressure of the inlet do not allow the temperature to rise after 230°. The third graph, Fig. 7-3, shows the path of the temperature when the speed is 4000 rpm, the maximum achievable with our sensor. The maximum measure temperature is 65.0 °C and the maximum simulated is 65.25 °C. The flat part of the curve is now inclined and barely visible.
..........
Conclusions
This research has been granted by the 973 - 2013CB632305 (The working Behaviour and the Lubricant Mechanism of High-Quality Synthetic Lubricants in Aero-power Transmission Systems), the work has the aim to devise and verify a design flow to optimize hydrodynamic journal bearings. To accomplish this aim several steps have been made:
1. An extensive literature review regarding the journal bearing design has been carried out. The main outcomes of this phase of the work (Phase 1, Fig. 1-5) show that there are still open issues in the field of HJB optimization and their analysis with CFD tools, while there is a clear claim of this solution from numerous producers.
2. The second phase of the work had the clear aim to reproduce the previous work done in the field of HJB performance calculations improving the diagonalization of the matrix generated with the finite difference scheme that solves the Reynolds equation. In this part of the work a software has been implemented to calculate the main performance of HJB and compare the results with previous literature.
3. A series of Artificial Intelligence tools have been developed and modified to be used for the solution of tribological problems. In chapter 3.2 the third phase of the work has been carried out and an ANN has been build and trained in order to calculate HJB performance. Furthermore, two minimization techniques, based on Genetic Algorithms and Artificial Bee Colony Algorithm have been analysed and used to minimize the objective functions which are the core of the optimization process.
4. As presented in chapter 4 and in the appendix a CFD simulation tool has been developed with the use of Matlab, Ansys Fluent and Ansys Gambit in order to calculate the HJB performance specially taking into account the influence of the inlet geometry.
5. The 3D CFD model has been validated with experimental results. To collect real data a test rig machine has been designed and built. The test device has the capability to measure temperatures, friction torque and monitor the lubricant temperature rise in real working conditions. The data are used to confirm the prediction of the ANN and the CFD tool.
6. Two study cases have been solved to prove the feasibility of the algorithm with two different bearings, with different inlet geometry, lubricant and diameter. The results clearly show the ability of the algorithm to simultaneously reduce the friction torque and the mass flow which are trade off objectives.
............
参考文献（略）