Proceedings of the ASME 2010 International Mechanical Engineering Congress & Exposition IMECE2010 November 12-18, 2010, Vancouver, British Columbia, Canada IMECE2010-40074 MODELING AND OPTIMIZATION OF AN ELLIPTICAL SHAPE ULTRASONIC MOTOR USING COMBINATION OF FINITE ELEMENT METHOD AND DESIGN OF EXPERIMENTS Hamed Sanikhani School of Mechanical Engineering, Sharif University of Technology, Tehran, Iran h.sanikhani@gmail.com Javad Akbari Center of Excellence in Design, Robotics, and Automation, School of Mechanical Engineering, Sharif University of Technology, Tehran, Iran akbari@sharif.edu Ali Reza Shahidi Research Center for Science and Technology In Medicine (RCSTIM), Tehran, Iran a.shahidi@gmail.com Ali Akbar Darki School of Mechanical Engineering, Sharif University of Technology, Tehran, Iran a.a.darki@gmail.com ABSTRACT Standing-wave ultrasonic motors are a modern class of positioning systems, which are used to deliver a high precision linear or rotary motion with an unlimited stroke. The design process should be performed through an effective optimization algorithm in order to guaranty proper and efficient function of these motors. An optimization method of ultrasonic motors is proposed based on the combination of finite element method and factorial design as a design of experiments in this study. The results show the ability of this method in optimal design of ultrasonic motors especially those which have a complex structure and multi modes operation principle. 1. INTRODUCTION Standing-wave ultrasonic motors are a modern class of positioning systems, which are used to deliver a high precision linear or rotary motion with an unlimited stroke. Significant advantages such as large thrust force per unit volume, fast response and good controllability, non magnetic and noiseless operation, nano-scale positioning accuracy and high braking force without power consumption makes these devices a suitable alternative for conventional electromagnetic actuators [1, 2]. The working principle of ultrasonic motors is production of an ultrasonic elliptical micro-scale motion at the contact surface of stator (vibrator) and rotor (slider). This motion is converted to a macro-scale motion of rotor by friction effect. In the most of the standing-wave ultrasonic motors, this elliptical motion is generated by superposition of two orthogonal motions, resulting from mode shapes of two adjacent natural frequencies. These mode shapes of the motor are excited by means of piezoelectric actuators. Simultaneous excitation of two modes would be feasible and the elliptical motion would be amplified subjected to resonance effect, if the natural frequencies locate close together. This effectively increases the ultrasonic motor performance. Therefore, the coincidence of these mode shapes’ frequencies is the most important objective in the proper design of these multi modes ultrasonic motors. Furthermore, in order to achieve noiseless operation, natural frequencies should be located in the ultrasonic range (over 20 kHz). Based on the working principle of the ultrasonic motors, various researches have been performed about the motor structure design and its simulation, in the recent years. In some of these works, vibrational behaviors of the motor have been modeled using exact analytical equations of motion [3, 4]. However, application of the exact methods is limited to models with simple structure such as bar type or rectangular shaped motors. In the more practical problems because of complexity of structure, nonlinear effects of contact and electromechanical coupling of piezoelectric elements, finite element method (FEM) is considered as the most general and applicable method in modeling and simulation of the ultrasonic motors. Using the trial and error method is not practical and efficient to optimum design of the ultrasonic motors, especially for those with complex structures and multi modes working principle. Thus, design process should be performed through a proper optimization algorithm. Lu et al. introduced a 1 Copyright © 2010 by ASME rectangular ultrasonic motor and tried to locate operating mode shapes at the same frequency by changing the stator length [2]. Ho studied an elliptical ultrasonic motor and minimized difference of frequencies of two orthogonal mode shapes by adjusting the ratio of the major axis diameter to the minor axis diameter of the ellipse [5]. Shiyang and Ming used two parameters particle swarm optimization (PSO) to design an ultrasonic motor [6]. Bouchilloux and Uchino implemented a genetic algorithm (GA) optimization on a two-dimensional finite element model of an elliptical motor with four design parameters [7]. Fernandez and Perriard designed a motor with two symmetric mode shapes by factorial design method [8]. Because of this symmetry, natural frequencies are identical. So, the optimization process only focused on maximizing amplitude of motion. In this work, modeling, simulation and optimization of an elliptical ultrasonic motor using combination of finite element and factorial design are proposed. First parameterization of the model of the motor is performed. A mathematical model is fitted on the obtained data from finite element modal analysis based on three-level factorial design, which is a conventional method of design of experiments (DoE). Next, the values of the design parameters are optimized using a genetic algorithm in order to coincidence of frequencies of the two orthogonal modes in the ultrasonic range (over 20 kHz). Finally a finite element optimization is applied in the vicinity of the optimum point. This optimization compensates approximation between finite element model and the mathematical model. The final values of design parameters are obtained from these two optimization processes. actuators, a vertical sinusoidal motion would be generated on the tip surface of stator, and a horizontal sinusoidal motion would be generated, if out-phase voltage applies. These resultant motions are given by: sin 2 (1) 0 sin 2 (2) 0 where and are natural frequency and phase difference between excitation voltage and resultant motion, respectively. Subscripts n and t refer to the normal and the tangential modes. 2. STRUCTURE AND OPERATING MODES OF THE MOTOR The schematic of the proposed ultrasonic motor is illustrated in Fig. 1. Stator is composed of an elliptical shell which is connected to a fixed base using a flexible mechanism. This mechanism is designed in a manner that supports the stator with the minimum restriction of the motor mode shapes. Moreover, it is used to preloading the stator against the rotor. The vibrational motion of the stator is excited by two multilayer piezoceramic rings as motor actuators. These actuators are fixed and prestressed between the stator shell and a central mass. This prestressing mechanism increases the life of actuators. The central mass is installed in order to separation of the motion of the two actuators. A cylindrical ceramic part is inserted in the shell as tip of the stator to improvement of frictional condition of the contact area. This ceramic tip has a high wear resistance and coefficient of friction and thus improves the transmission of motion between the stator and the rotor. The two orthogonal mode shapes of the motor (normal mode and tangential mode) are shown in Fig. 2. The elliptical motion of the motor tip is generated by simultaneous excitation of these mode shapes. If in-phase sinusoidal voltage applies to the Fig. 1. Schematic of the ultrasonic motor Fig. 2. Orthogonal mode shapes of elliptical shell of the stator: (a) normal mode, (b) tangential mode 2 Copyright © 2010 by ASME If the dimensions of the motor are determined so that the two natural frequencies are close together, simultaneous excitation of two modes would be possible near their resonances. To achieve this goal, the driving voltages of two actuators should have a π⁄2 phase difference at equal frequency. This frequency could be varied between two natural frequencies of the motor. Using this method of stimulation, the desired elliptical motion is expected to be achieved by superposition of two orthogonal motions described as Equations (1, 2). 3. OPTIMIZATION 3. 1. PARAMETERIZATION OF THE MODEL The parameterization of the model is the first step in the optimal design process. The parametric model increases the design flexibility and also provides possibility of the implementation of appropriate optimization methods. In this work, parameterization of the model is carried out using four influence design parameters. These parameters, that are shown in Tab. 1, include the ratio of ellipse axis (ER), thickness of the central mass (Tcm), thickness of the shell (Ts) and height of the stator (Hs). model. However, study of the effects of these parameters on the natural frequencies for finding their optimum values is not possible in practice, because of the geometrical complexities and the wide range of parameter variations. Thus, deriving a mathematical model with the ability of simulation of the motor modal behavior has significant advantages in such cases. In addition various optimization methods could also be implemented on shuch models easily. A factorial design, which is a conventional method of design of experiments (DoE), is applied in order to obtain the mathematical model. In this method data is collected from a primary model (in this case finite element model) at the certain points of the variation ranges of the parameters. Then, a polynomial is fitted on these data. Depending on the problem, the number of parameters and required accuracy, this polynomial could be linear, quadratic or cubic (using higher degree polynomials is not very common). In a three-level factorial design which is used in this work, each normalized parameter could have three values: 0, 0.5 and 1 and thus 3P points are defined in the design space for a P-parameters problem. The two natural frequencies of the motor are mathematically modeled with two quadratic polynomials, as shown in Equations (4) and (5). Tab. 1. Design parameters Design parameter Parameter coefficient Initial range of parameter ER k1 1-1.4 Tcm k2 11-Hs(mm) Ts k3 3-4.5(mm) Hs k4 12-14(mm) The second column of Tab. 1 (parameter coefficient) shows normalized parameters which are related to design parameters by Equation (3). 1 2 3 4 (4) 1 2 3 4 (5) To determine all 81 coefficients of each equation ( and , 81 sets of data obtained from the finite element analysis are used. Substituting these data into above equations yields a set of linear equations for each frequency (Equations (6) and (7)). 1 1 2 3 4 (3) 1 1 2 3 4 where DP indicates the design parameter, DP and DP are minimum and maximum values of it and k is related parameter coefficient. Using these normalized coefficients helps to facilitate modeling and optimization process and also comparison of the results. 1 1 2 3 4 1 1 2 3 4 (6) DP DP k DP 1 k 3. 2. DERIVING A MATHEMATICAL MODEL USING FACTORIAL DESIGN It is possible to obtain natural frequencies at each point of the design parameters domain by solving the finite element (7) The coefficient matrices and could be determined using standard methods of linear algebra. In this study, least squares method is used to find the coefficients. This method provides less average error in the whole of the approximation domain. 3 Copyright © 2010 by ASME 3. 3. OPTIMIZATION USING GENETIC ALGORITHM As mentioned before, the main objective of design of the multi modes ultrasonic motors is the coincidence of the natural frequencies. In addition, this adjustment should be occurred in the ultrasonic range to the noiseless operation of the motor. The objective function which is defined in order to satisfy these conditions is defined as Equation (8). In this equation, the frequencies are calculated using the mathematical models. | | 20000 , between the two frequencies that this error has been eliminated after the final finite element optimization. Results of this optimization are demonstrated in Tab. 3. (8) The minimization of this function could be performed using standard optimization methods. In the present work, a genetic algorithm is implemented on this 4-parameters objective function to find the optimum point. However, this point may differ slightly from the optimum point driven from the finite element model because of the approximation error. To eliminate this error, a final finite element optimization is carried out in the vicinity of the mathematical model optimum point. Due to the small variation ranges, it is expected that the convergence would be achieved after a few number of iterations. Fig. 3. Comparison of objective function value at the DoE points and the optimum point 4. RESULTS AND DISCUSSION Tab. 3. Results of final finite element optimization 4. 1. OPTIMIZATION RESULTS The characteristics of the optimum point which have been obtained by the implementation of GA on the mathematical model are shown in Tab. 2. The coincidence of the natural frequencies has been provided at 20743 Hz in the ultrasonic range. To show the ability of this optimization method, the objective function value at the factorial design points and the optimum point is plotted in Fig. 3. Parameter coefficient Optimum values k1 0.86413 k2 1 k3 0.089759 k4 1 Tab. 2. Results of GA optimization Parameter coefficient Optimum values k1 0.86407 k2 1 k3 0.089745 k4 1 fn 20743 (Hz) ft 20743 (Hz) f 0 (Hz) Objective function value -743 Finite element modal analysis has been carried out with the optimized parameters. The FE values of the frequencies of the normal and the tangential modes are equals to 20760 Hz and 20730 Hz, respectively. So, there is about 30 Hz difference fn ft 20763 (Hz) 20766 (Hz) f 3 (Hz) Objective function value -760 4. 2. VERIFICATION OF THE MATHEMATICAL MODEL In order to verify the accuracy of the mathematical model, the values of the objective function and the natural frequencies have been compared to the finite element results at some verification points which are selected based on a reasonable procedure. In this procedure, three of the four parameters are fixed at their optimum values and the other one is varied from 0 to 1 with step size of 0.2. Considering the existence of four parameters, 24 verification points are determined. This verification shows that average relative errors in the approximation of the frequencies are less than 0.1% and average error of the objective function value approximation is less than 40 Hz. The comparison of these results is shown in Fig. 4. 4 Copyright © 2010 by ASME 4. 3. SENSITIVITY ANALYSIS Using the mathematical model, variation of the objective function with each design parameter is determined as a sensitivity analysis (Fig. 5). The results show that k1 and k3 have greater effect on the value of the objective function. This analysis would be useful to determine required manufacturing tolerances. a) k1 is variable Fig. 5. Sensitivity analysis using mathematical model b) k2 is variable COLCLUSION c) k3 is variable In this work, an optimization method of ultrasonic motors is proposed based on the combination of finite element method and factorial design as a design of experiments. In this method, finite element model is used to determine natural frequencies of the motor and a mathematical model was fitted on data obtained from FEM using three-level factorial design. The accuracy of the mathematical model to estimate the natural frequencies is verified. Genetic algorithm minimization and sensitivity analysis is implemented on the mathematical model to find optimum values of the design parameters and effects of them on the motor natural frequencies. The results show the ability of this method in the optimal design of the ultrasonic motors especially those which have a complex structure and multi modes operation principle. REFERENCES [1] Hemsel, T., Mracek, M., Twiefel, J., and Vasiljev, P., 2006, “Piezoelectric linear motor concepts based on coupling of longitudinal vibrations,” Ultrasonics, 44, pp. 591-596. d) k4 is variable Fig. 4. Comparison of the mathematical model results with the finite element results at the verification points : a) k1 is variable, b) k2 is variable, c)k3 is variable, 4) k4 is variable [2] Lu, C., Xie, T., Zhou, T., and Chen, Y., 2006, “Study of a new type linear ultrasonic motor with double-driving feet,” Ultrasonics, 44: pp. 585–589. 5 Copyright © 2010 by ASME [3] He, S., Chen, W., Tao, X., and Chen, Z., 1998, “Standing wave Bi-directional linearly moving ultrasonic motor,” IEEE transactions on ultrasonics, ferroelectrics, and frequency control, 45(5), pp. 1133-1139. [4] Roh, Y., and Kwon, J., 2004, “Development of a new standing wave type ultrasonic linear motor,” Sensors and Actuators, 112(2), pp. 196-202. [5] Ho, S. T., 2006, “Characteristics of the Linear Ultrasonic Motor using an Elliptical Shape Stator,” Japanese Journal of Applied Physics, 45(7), pp. 6011. design of ultrasonic motors,” Sensors and Actuators, 148, pp. 285-289. [7] Bouchilloux, P., and Uchino, K., 2003, “Combined Finite Element Analysis - Genetic Algorithm Method for the Design of Ultrasonic Motors,” Journal of Intelligent Material Systems and Structures, 14(10), pp. 657-667. [8] Fernandez, J.M., and Perriard, Y., 2006, “Sensitivity Analysis and Optimization of a Standing Wave Ultrasonic Linear Motor,” IEEE transactions on ultrasonics, ferroelectrics, and frequency control, 53(7), pp. 1352-1361. [6] Shiyang, L., and Ming, Y., 2008, “Particle swarm optimization combined with finite element method for 6 Copyright © 2010 by ASME