液靜壓軸承因其高剛性、高承載力、幾乎無摩擦力等特性，已經被廣泛使用在精密機械加工機產業，而使用於其中之節流器更是影響液靜壓軸承表現的關鍵元件，因此本文致力於研究實驗室過往所設計之溝槽式、自補償式與結合前兩者的複合式節流器，進行物理理論之模擬與實驗驗證，完善此新構形節流器的設計。
然而，由於對於整個液靜壓軸承系統，其數學模型過於冗長，又為非線性，在做理論模擬時須做非常多的假設條件；再加上實驗的過程中，來自於環境干擾因素的誤差原因，有些理論模擬並無法完全考慮，因此最後導致模擬結果的不準確。
為解決此問題，本研究利用多層感知器模型可以接受多輸入多輸出與能處理非線性問題等特性，以實驗所取得之數據做為訓練資料，建立一多層感知器網路模型，預測液靜壓軸承系統搭配不同節流器時，隨著供油壓力、負載、節流器結構參數等數值變化時其他相應參數可能的改變，並依此結果輔助新節流器設計時理論模擬的不足，以達到最佳化設計的目標，最後也期待本論文之研究結果能實際應用於精密機械的設計過程之中。

Hydrostatic bearings are widely used in precision machine tools because of their superior characteristics of high stiffness, high load-carrying capacity, high damping, nearly frictionless and long life. The stiffness of hydrostatic bearings is mainly affected by the use of flow restrictors. With different kinds of flow restrictors, hydrostatic bearings will perform with distinct stiffness. Thus, the goal of this paper is to design a hybrid type restrictor. However, accurate estimation of the bearing stiffness is often unattainable, which may attribute to variation of environment conditions, resistance from oil tube and incompatible assumptions in the physical model of hydrostatic bearings.
This paper proposes use of multilayer perceptron (MLP) to learn and predict the stiffness of hydrostatic bearings. Different from classical metamodeling method, the MLP model has stronger filtering capacity. It can accept different kinds of variables as input and build a Multi-input Multi-output (MIMO) system. According to this malleable nonlinear model and functions, the MLP model can find the hidden patterns in the data and predict the result.
The simulation was implemented based on the derived equations and an experimental setup was constructed to investigate the stiffness of hydrostatic bearings. Various kinds of flow restrictors such as groove and self-compensation type restrictor were used in the experiment and pressure, flow rate, load, temperature and oil-film displacement were measured by the corresponding sensors directly. Based on the collected data, the MLP model for the stiffness of hydrostatic bearings can be effectively trained. Compared to the simulation, the predicted results obtained from the MLP model constructed in this paper are more applicable. Therefore, with the method of multilayer perceptron, the stiffness of hydrostatic bearings can be precisely modeled and improved.

[1] Girard, L. D., 1862, Application des Surfaces Glissantes, Bachelier, Paris.
[2] Malanoski Telingater, V. S., 1972, “Hydrostatic Slideways Using Standard Bearings,” Machines and Tooling. Vol.43, No.2, pp.15-20.
[3] Ling, T. S., 1962, “On the Optimization of the Stiffness of Externally Pressurized Bearings,” Trans. ASME. J. Basic Eng. Vol. 84, pp.119-122.
[4] O’Donoghue, J. P., 1972, “Parallel Orifice and Capillary Control for Hydrostatic Journal bearings,” Tribology International Vol. 5, pp.81-82.
[5] Sharma, S. C., Sinhasan, R., Jain, S. C., Singh, N., and Singh, S. K., 1998, “Performance of Hydrostatic/Hybrid Journal Bearings with Unconventional Recess Geometries,” Tribology Transactions Vol.41, pp.375-381.
[6] Kang, Y., Shen, P. C., Chen, C. H., Chang, Y. P. and Lee, H. H., “Modified determination of fluid resistance for membrane-type restrictors,” Industrial Lubrication and Tribology, 2007, pp.123-131.
[7] Mohsin, M. E., 1963, “The Use of Controlled Restrictors for Compensating Hydrostatic Bearing,” Third International Conference on Machine Tool Design Research pp.129-424.
[8] Rowe, W. B. and O’Donoghue, J.P., 1970, “Diaphragm Valves for Controlling Opposed Pad Hydrostatic Bearing,” Proc. I. E. Tribology. Vol.184, pp.1-9.
[9] Moris, S. A., 1972, “Passively and Actively Controlled Externally Pressurized Oil-film Bearing, ” Trans. ASME. Ser. F, Vol. 94.
[10] Rowe, W. B., 1976, Hydrostatic Bearing Patent Application no.51919/67. Publish 15th November170 (1):602. Assigned to Wickman Ltd.
[11] Rowe, W. B., 1969, Hydrostatic bearing patent application no.22072/66. Publish 15th November 170 (1):602. Assigned to Wickman Ltd.
[12] Tully, N., 1977, “Static and Dynamic Performance of an Infinite Stiffness Hydrostatic Thrust Bearing,” Trans. of ASME. Vol. 99, No.1, pp.106-112.
[13] Osumi, T., Mori, H., Ikeuchi, K., 1985, “Effects of Stabilizer on Initial Response of Self Controlled Externally Pressurized Bearings,” Trans. JSLE, Vol. 20, pp.651-657.
[14] Yoshimoto, S., Anno, Y., Amari K., 1990, “Static Characteristics of Hydrostatic Journal Bearing with a Self Controlled Restrictor Employing Floating Disk,” Trans. JSLE. Vol. 56, pp.3360-3367.
[15] Yoshimoto, S., Kikuchi, K., 1999, “Step Response Characteristics of Hydrostatic Journal Bearings with Self-controlled Restrictor Employing Floating Disk,” Trans. Int. Vol.121, No.4, pp. 315-320.
[16] Schoenfold, R., “Regulator for adjusting the fluid flow in a hydrostatic or aerostatic device.” US Patent number 6276491B1, 2001.
[17] Mayer, J. E. and Shaw, M. C., 1963, “Characteristics of Externally Pressurized Bearing Having Variable External Flow Restrictors,” ASEM Journal of Basic Engineering, Vol. 85, pp.291.
[18] Schoenfold, R., Flow-rate regulator for liquid and/or gaseous materials. D.E. Patent No. 3150117 A1 (1983).
[19] Schoenfold, R., 2001, “Regulator for Adjusting the Fluid Flow in a Hydrostatic or Aerostatic Device.” US Patent number 6276491B1.
[20] 王翊丞，2015，“液靜壓軸承複合型節流裝置設計與實驗”國立清華大學動力機械系碩士論文。
[21] 宋震國、王翊丞，2015，“主動式補償液靜壓軸承及其節流器”，中華民國發明專利第I597435號。
[22] McCulloch, W.S. & Pitts, 1943, W. Bulletin of Mathematical Biophysics 5: 115.
[23] Tropp, H. S., Chichester John, V. N., 2003, Encyclopedia of Computer Science, 4th edition, John Wiley and Sons Ltd.
[24] Hebb, D.O., 1949, The Organization of Behavior. New York: Wiley & Sons.
[25] Rosenblatt, Frank, 1957, The Perceptron--a perceiving and recognizing automaton. Report 85-460-1, Cornell Aeronautical Laboratory.
[26] Minsky, M., Papert, S., 1969, Perceptrons: An Introduction to Computational Geometry. Cambridge, MA, USA: MIT Press.
[27] Hopfield, J. J., 1982, "Neural networks and physical systems with emergent collective computational abilities", Proceedings of the National Academy of Sciences of the USA, vol. 79 no. 8 pp. 2554–2558.
[28] Rumelhart, D. E., Hinton, G. E., Williams R. J., 1986, “Learning representations by back-propagating errors”, Nature, Vol. 323, pp. 533-536.
[29] Hinton, G. E. and Salakhutdinov, R. R., 2006, “Reducing the Dimensionality of Data with Neural Networks”, Sience, Vol.313, pp. 504-507.
[30] Canbulut, F., Sinanoglu, C., Yildirim, S., 2004, Analysis of Effects of Sizes of Orifice and Pockets on the Rigidity of Hydrostatic Bearing Using Neural Network Predictor System, KSME International Journal, Vol. 18 No. 3, pp. 432 ~ 442.
[31] Canbulut, F., Sinanoglu, C., 2004,An investigation on the performance of hydrostatic pumps using artificial neural network, JSME International Journal, Vol. 47 No. 3, pp 864-872.
[32] Canbulut, F., The experimental analyses of effects of the geometric and working parameters on the circular hydrostatic thrust bearing, JSME International Journal, Vol. 48 No. 4, pp 715-722.
[33] Canbulut, F., Yildirim, S., and Sinanoglu. C., 2004, Design of an artificial neural network for analysis of frictional power loss of hydrostatic slipper bearings, Tribology Letters, Vol. 17, No. 4, pp 887-898.
[34] Rafieea, J., Arvania, F., Harifib, A., Sadeghic, M.H., 2007, Intelligent condition monitoring of a gearbox using artificial neural network, Mechanical Systems and Signal Processing 21, pp 1746–1754.
[35] Saxena, A., Saad, A., 2007, Evolving an artificial neural network classifier for condition monitoring of rotating mechanical systems, Applied Soft Computing 7, pp 441–454.
[36] Bruus, 2008, H. "Theoretical microfluidics. Oxford master series in condensed matter physics.".
[37] 鍾洪，張冠坤，2007，液體靜壓動靜壓軸承設計使用手冊，電子工業出版社：p.30。
[38] Slocum, A. H., 1992, Precision Machine Design, Prentice Hall Inc. Dear born, Michigan.
[39] 李威志，2012，“新型主動式補償液靜壓軸承之靜動態特性分析”國立清華大學動力機械系碩士論文。
[40] 鮑思丞，2016，“智慧型線性液靜壓滑塊模組設計”國立清華大學動力機械系碩士論文。
[41] 程瑾文，2017，“泛用型虛實整合系統架構建立─以液靜壓節流器之調整為驗證載具”國立清華大學動力機械系碩士論文。
[42] 張軼峯，2018，“液靜壓複合式節流器設計─輔以類神經網路之決策” 國立清華大學動力機械系碩士論文。
[43] Fox & McDonald, 2012, Fluid Mechanics, 8ED, John Wiley & Sons Inc.
[44] Neuron, Wikipedia, 網路資源, https://en.wikipedia.org/wiki/Neuron.
[45] 李宏毅，機器學習課程講議，http://speech.ee.ntu.edu.tw/~tlkagk/courses_ML17_2.html.
[46] Stanford CS class, CS231n: Convolutional Neural Networks for Visual Recognition., http://cs231n.github.io/neural-networks-3/.
[47] Rowe, W. B., 1983, Hydrostatic and hybrid bearing design, Butterworths & Co. Ltd.
[48] Spirol盤形彈簧使用手冊，https://www.spirol.com/library/main_catalogs/SPIROL-Disc-Springs-us.pdf