【正文】 G laser. The textured portion of the unidirectional bearing was a= and that of the bidirectional bearing was a= . As can be seen from ?gure 2 both these a values should produce loadcarrying capacity vary close to the maximum theoretical test rig is shown schematically in ?gure 5. An electrical motor turns a spindle to which an upper holder of the rotor is attached. A second lower holder of the stator is ?xed to a housing, which rests on a journal bearing and an axial loading mechanism that can freely move in the axial direction . An arm that presses against a load cell and thereby permits friction torque measurements prevents the free rotation of this housing. Axial loading is provided by means of dead weights on a lever and is measured with a second load cell. A proximity probe that is attached to the lower holder of the stator allows online measurements of the clearance change between rotor and stator as the hydrodynamic e?ects cause axial movement of the housing to which the stator holder is ?xed. Tap water is supplied by gravity from a large tank to the center of the bearing and the leakage from the bearing is collected and recirculated. A thermocouple adjacent to the outer diameter of the bearing allows monitoring of the water temperature as the water exit the bearing. A PC is used to collect and process data online. Hence,the instantaneous clearance, friction coe?cient, bearing speed and exit water temperature can be monitored constantly. The test protocol includes identifying a reference ―zero‖ point for the clearance measurements by ?rst loading and then unloading a stationary bearing over the full load range. Then the lowest axial load is applied, the water supply valve is opened and the motor turned on. Axial loading is increased by steps of 40 N and each load step is maintained for 5 min following the stabilization of the friction coe?cient at a steadystate value. The bearing speed and water temperature are monitored throughout the test for any irregularities. The test ends when a maximum axial load of 460 N is reached or if the friction coe?cient exceeds a value of . At the end of the last load step the motor and water supply are turned o? and the reference for the clearance measurements is rechecked. Tests are performed at two speeds of 1500 and 3000 rpm corresponding to average sliding velocities of and m/s, respectively and each test is repeated at least three times. 4. Results and discussion As a ?rst step the validity of the theoretical model in Ref. [12] was examined by paring the theoretical and experimental results of bearing clearance versus bearing load for a unidirectional partialLST bearing. The results are shown in ?gure 6 for the two speeds of 1500 and 3000 rpm where the solid and dashed lines correspond to the model and experiment, respectively. As can be seen, the agreement between the model and the experiment is good, with di?erences of less than 10%, as long as the load is above 150 N. At lower loads the measured experimental clearances are much larger than the model predictions, particularly at the higher speed of 3000 rpm where at 120 N the measured clearance is 20 lm, which is about 60% higher than the predicted value. It turns out that the bination of such large clearances and relatively low viscosity of the water may result in turbulent ?uid ?lm. Hence, the assumption of laminar ?ow on which the solution of the Reynolds equation in Ref. [12] is based may be violated making the model invalid especially at the higher speed and lowest load. In order to be consistent with the model of Ref. [12] it was decided to limit further parisons to loads above 150 N. It should be noted here that the ?rst attempts to test the baseline untextured bearing with the original surface ?nish of Ra= lm on both the stator and rotor failed due to extremely high friction even at the lower loads. On the other hand the partialLST bearing ran smoothly throughout the load range. It was found that the postLST lapping to pletely remove about 2 lm height bulges, which are formed during texturing around the rims of the dimples, resulted in a slightly rougher surface with Ra= lm. Hence, the baseline untextured stator was also lapped to the same rough ness of the partialLST stator and all subsequent tests were performed with the same Ra value of lm for all the tested stators. The rotor surface roughness remained, the original one namely, lm. Figure 7 presents the experimental results for the clearance as a function of the load for a partialLST unidirectional bearing (see stator in ?gure 4(a)) and a baseline untextured bearing. The parison is made at the two speeds of 1500 and 3000 rpm. The area density of the dimples in the partialLST bearing is Sp= and the textured portion is a 188。 0:734. The load range extends from 160 to 460 N. The upper load was determined by the testrig limitation that did not permit higher loading. It is clear from ?gure 7 that the partialLST bearing operates at substantially larger clearances than the untextured bearing. At the maximum load of 460 N and speed of 1500 rpm the partialLST bearing has a clearance of 6 lm while the untextured bearing clearance is only lm. At 3000 rpm the clearances are and lm for the LST and untextured bearings, respectively. As can be seen from ?gure 7 this ratio of about 3 in favor of the partialLST bearing is maintained over the entire load range. Figure 8 presents the results for the bidirectionalbearing (see stator in ?gure 4(b)). In this case the LST parameters are Sp 188。 0:614 and a 188。 0:633. The clearances of the bidirectional partialLST bearing are lower pared to these of the unidirectional bearing at the same load. At 460 N load the clearance for the 1500 rpm is lm and for the 3000 rpm it is 6 lm.