INTRODUCTION: Hydrodynamic bearings, dry gas seals etc. have surface texture such as grooves or dimples on the sliding surfaces. Due to these surface textures, dynamic pressure is generated, however it is known that the dynamic pressure generated by shape and depth greatly changes1, therefore a lot of research2,3 has been done previously. As an example of optimum design, Hashimoto et al.4 optimizes the groove shape by expressing the shape of the groove as a spline function and introducing a correction value into the function. In addition, Imura et al.5 optimizes the dimple arrangement by manipulating the dimples as a particle with virtually repulsive force and changing the arrangement at equilibrium. However, in these optimizations, optimization is performed only with the type of the texture as the basis, and no design method of surface texturing capable of encompassing both textures has been proposed so far. Under these circumstances, the authors focused on Cellular Automation (CA) method. The CA method follows the rule with only the state of the self and the neighboring cell as a variable, updates its own state and forms the whole organization. By applying the CA method with the state of a single cell as the presence or absence of texture, the whole texture distribution which is compounded discretely and continuously according to the rule is determined. By optimizing the rule according to the purpose by the evolutionary cellular automaton (ECA)6 method which combines the CA method with the genetic algorithm (GA) method, the surface containing both textures. In this research, we applied the ECA method to the design of surface texturing and aimed at developing a new design method. The object of texturing is a dry gas seal which is a typical non-contact type shaft seal, and the study on the application of optimization using ECA method is reported.
METHODS: In the Cellular Automation (hereinafter: CA) method, it is necessary to determine a rule of neighborhood for the cell pattern. Figure 1 shows an example of state updating by Neumann neighborhood which is a representative one. A black cell represents a cell having a texture, a white cell represents a cell having no texture, and a center cell is a cell of interest. In the Neumann neighborhood the focused cell sets by the state of five cells including itself, the left, right, top and bottom 4 adjacent ones. The cells vary step by step according to rule of Neumann neighborhood. The number of state is 32(=25) and if you set the state appropriately high performance tribo-element is designed. In addition, in this study, Moor neighborhood which is one of typical rule was used in this study. Combining the CA method and Genetic algorism, the surface texturing on dry gas seals face are optimized.
RESULTS and DISCUSSION: Figure 2 shows the Pareto solutions of each rule obtained by optimization. S-N and S-M represent the Neumann and the Moor neighborhood when the seal face is set as one area. Q-N and Q-N represent the ones when the seal face is set as four areas.
As shown in Fig. 2, the S-N and Q-N rules based on Neumann neighborhood have better Pareto solutions than the S - M and Q - M rules based on the Moore. Therefore, in this optimum design, rules based on Neumann are considered to be appropriate. Figure 3 shows the distribution of the texture obtained by optimization. According to the textures in (A), (B), and (D), in the QN rule, the texture concentrates in one divided region and a rule is formed to form a distribution like pocket grooves. However, a ring-shaped texture is formed on the inner diameter side of FIG. 3 (A) where