Comparison of Different Methods for Numerical Approximation of Static Characteristics of McKibben Pneumatic Artificial Muscle

Transcription

1 Comparison of Different Methods for Numerical Approximation of Static Characteristics of McKibben Pneumatic Artificial Muscle JÁN PITEĽ, MÁRIA TÓTHOVÁ, STELLA HREHOVÁ, ALENA VAGASKÁ Department of Mathematics, Informatics and Cybernetics Technical University of Košice Bayerova 1, Prešov SLOVAKIA Abstract: - Pneumatic artificial muscles exhibit highly non-linear static characteristics especially due to inherent properties of their materials and compressibility of air. Therefore, it is difficult to control them without adequate modeling and simulation. The paper deals with the numerical approximation of the static characteristics of McKibben pneumatic artificial muscle. The muscle force dependence on the muscle contraction and pressure in the muscle was approximated using available software tools as Matlab Curve itting and Microsoft Excel and three obtained functions are expressed and shown for comparison in the paper. Key-Words: - numerical approximation, static characteristic, pneumatic artificial muscle 1 Introduction Pneumatic artificial muscles (PAMs) belong to the group of non-conventional actuators which have great potential in mechatronic [1] and biomedical [], [] applications due to an advantageous power to weight ratio, high stiffness, structural simplicity, natural compliance and self-dampening. When compressed air is supplied inside PAM its elastic tube has tendency to expand in the radial direction, whereas a contraction occurs in the axial direction and thus the tensile force is generated. This force is characterized by highly non-linear responses []. The principle of PAM is described in more details for example in [5], [6], [7]. McKibben artificial muscle is the most used type of PAM. An intensive attention has been and still is paid to modified version of this muscle and its applications by various companies such as Bridgestone Corporation of Japan (Rubbertuator Muscle), esto in Germany (DMSP/MAS luidic Muscles), Shadow Robot Company (Shadow Air Muscle) and Merlin Systems Corporation in United Kingdom (Humaniform Muscle). Static Characteristics and Geometric Parameters of PAM The static characteristic of PAM represents the relation (a function) between the muscle force and muscle contraction κ (under a constant muscle pressure p) regardless the time factor. PAMs manufactured by ESTO are now the most popular and commercially available and therefore static characteristics of fluidic muscles of this company were used for approximation. These characteristics of PAM type ESTO MAS-0-00N in a working range between 0 % and 5 % of the muscle contraction (which is recommended by ESTO catalogue [8]) are shown in ig. 1. ig. 1 Static characteristics of PAM type ESTO MAS-0-00N PAMs consists of the inner elastic tube inserted into the non-extensible fibers which define an expansion in the context of increasing pressure. Typical materials used for the tube are latex and silicone rubber, fibers are made typically of nylon. The tube is connected to the terminals at both ends of muscle through which the mechanical energy is ISBN:

2 transferred to the load. These terminals are most often made of metallic materials. The commercially available PAM MAS-0-00N by ESTO differs slightly from the general McKibben type muscle. The fibers of the fluidic muscle are knit into the tube, offering easy assembly compared to conventional designs of muscle [9]. Then the main geometric parameters of PAM in ig. are the initial muscle length l 0, the initial angle α 0, the initial radius in the middle of the muscle r 0, the actual muscle length l, the actual angle α, the actual radius in the middle of the muscle r, number N wrapped of single fibers and the half nylon thread length L. r0 α0 N windings l r dl α l assumed that the maximal contraction will be the same for the different pressures. In order to balance the relation between pressure p and contraction κ 0, the member (p) was added to (1). This new member was suitable for higher pressures p but there were always the differences between the experimentally obtained values and the theoretical model for smaller pressure p in relation to contraction κ. In order to obtain the approximation for small values of pressure, member (κ) was also added to equation (1). After substituting new members and using relation for the contraction κ = (l 0 -l)/l 0, it can be obtained [15]: p p r p a p, b, () 0 1 where values a = /tan α 0, b = 1/sin α 0 depend on the parameters of PAM (for our type of PAM: a = 5., b =.8, r 0 = 10 mm). The following functions are used to achieve the best approximation of the muscle curve: L = const. α0 α p r0 N p r N p a e a e p c b, () b. () ig. Correlation between geometric parameters of PAM Approximation of Static Characteristics of PAM There are several basic approximation functions for description of static characteristics of PAM which are the subject of many works [9], [10], [11], [1], [1]. ollowing there are presented three possible ways of numerical approximation of static characteristics of PAM using the given measured data. Coefficients a ε, b ε, a κ, b κ, c κ were found by the least squares method using Matlab Curve itting Toolbox. The obtained values of these coefficients are shown in Table 1. Table 1 The values of coefficients from () and () Coefficient Value a ε 0,00075 b ε 0, a κ -0,0060 b κ -0, c κ 0, Approximation Using an Analytical Modeling of PAM On the basis of physical laws (law of energy conservation, Bernoulli equation, etc.) and the geometric parameters of PAM (ig. ), relation of the static characteristics was derived [1]: l 1 p, p r0 p. (1) tan 0 sin 0 Equation (1) does not include impact of the muscle membrane on the pressure change; it was ig. orce-contraction relation approximated by () for various pressures in the muscle ISBN:

3 Table The values of coefficients from (5) and (6) ig. calculated in correlation with measured for ig. ig. presents the static characteristics of luidic Muscle type MAS-0-00N (ig. 1) obtained by approximation and they were created using () in Microsoft Excel. To describe the nature and force of the relation between the calculated results and the specified by ESTO, regression and correlation analysis were used with achieved coefficient of correlation R = R = (ig. ). Coefficient Value a 0-99,000 a 1-17,100 a -18,8000 b ,000 b 1-16,6000 b 1,7700 b -0,60 b 0,0107 The result of approximation of the static characteristics of luidic Muscle type MAS-0-00N (ig. 1) is in ig. 5. They were created using (5), (6) in Microsoft Excel. ig. 6 describes relation between the calculated result and the specified by ESTO. The coefficient of correlation R = R = was found by using regression and correlation analysis.. Approximation Deducted from the Maximum orce of PAM The muscle force as a function (p,κ) of the muscle contraction for different pressures in the muscle can be deducted from the maximum muscle force max. Note that if the muscle contraction is constant, the muscle force depends almost linearly on the pressure. However, the proportionality factor decreases with increasing contraction. Then, the result of the muscle force is defined as follows [16]: a a, (5) 0 1 p, max pmax p a ig. 5 orce-contraction relation approximated by (5) and (6) for various pressures in the muscle where p max is the maximum and p is the actual pressure in muscle. Coeficients a 0 [N], a 1 [Nm -1 ], a [Pa] were found using Matlab Curve itting Toolbox. The maximum force max as a function of the muscle contraction is introduced by a fourthorder polynomial function for the response at the maximum pressure p max = 600 kpa [16]: max b b b b b, (6) 0 1 where coeficients b 0, b 1, b, b, b were also found using Matlab Curve itting Toolbox. All coeficients from (5) and (6) are shown in Table. ig. 6 calculated in correlation with measured for ig. 5 ISBN:

4 . Approximation Using a Polynomial unction The third method applied for approximation of static characteristics of PAM (ig. 1) was a polynomial approximation. In order to approximate these static characteristics with good accuracy, a fifth-order polynomial function of two variables was used. This polynomial function contains twenty-one coefficients and its form is as follows:, p a a a a p p a p 00 a a p a 0 0 a 0 a 10 a a p a 1 1 a p p a 5 a 1 0 p a p a a 1 1 a p p 11 p a p 05 p p 5. (7) The values of all coefficients in (7) were determined also using Matlab Curve itting Toolbox and they are shown in Table. Table The values of coefficients from (7) Coefficient Value Coefficient Value a 00 0,6700 a 1-1,15E-0 a 10-97,7700 a -,77E-06 a 01,50 a 1 -,99E-08 a 0 16,100 a 0 1,56E-08 a 11-0,16 a 50-5,57E-0 a 0 0,006 a 1-1,0E-05 a 0-1,970 a 6,5E-07 a 1 0,007 a -1,1E-08 a 1,78E-05 a 1 1,50E-10 a 0-1,10E-05 a 05-8,90E-1 a 0 0,070 ig. 7 orce-contraction relation approximated by (7) for various pressures in the muscle ig. 8 calculated in correlation with measured for ig. 7 The static characteristics generated by luidic Muscle type MAS-0-00N (ig. 1) and approximated using (7) were created again in MS Excel and they are shown in ig. 7. Relation between the calculated result and the specified by ESTO describes ig. 8. Coefficient of correlation R = R = is approaching to the maximum value of correlation (strongest, R = 1). Results of Comparison and Conclusion The muscle force dependence on the muscle contraction and pressure in the muscles was approximated for modified version of McKibben PAM produced by esto Company. Three different approximation functions were tested for possible approximation of the given static characteristics of luidic Muscle type MAS-0-00N (ig. 1). Analyses were carried out in Matlab Curve itting Toolbox environment and obtained results of approximation were shown using Microsoft Excel. Different approaches using an analytical modeling (presented in section.1), results of experimental modeling (presented in section.) and only numerical approximation (presented in section.) were used for comparison. irstly, the results of approximation using an analytical modeling of PAM (ig. ) were compared and the achieved correlation coefficient was R = As it can be seen from ig. there is a significant difference especially for higher values of the muscle force. Secondly, the results of approximation deducted from the maximum force of PAM were compared and the achieved correlation coefficient was R = As it can be seen from ig. 5 when compared to ig. 1 the difference is mainly for the mean values of the muscle contraction. ISBN:

5 The best approximation results were reached using a fifth-order polynomial function of two variables with twenty-one coefficients (the achieved correlation coefficient R = ) and the function (7) with the values of coefficients in Table was used in dynamic modeling of one Do (Degree of reedom) PAM-based pneumatic antagonistic actuator using modified Hill s muscle model [17]. Acknowledgment The research work is supported by the Project of the Structural unds of the EU, title of the project: Research and development of intelligent nonconventional actuators based on artificial muscles, ITMS code: References: [1] J.M. Hopen and A. Hošovský, The servo robustification of the industrial robot, Annals of DAAAM for 005 & Proceeding 16 th International DAAAM Symposium "Intelligent Manufacturing & Automation: ocus on Young Scientists and Researchers", Vienna, 005, pp [] K. Židek and J. Šeminský, Automated rehabilitation device based on artificial muscles, Annals of DAAAM for 011 & Proceeding nd International DAAAM Symposium "Intelligent Manufacturing & Automation: Power of Knowledge and Creativity", Vienna, 011, pp [] O. Líška, M. More, D. Janáčová and H. Charvátová, Design of rehabilitation robot based on pneumatic artificial muscles, Mathematical Methods and Optimization Techniques in Engineering, Antalya, 01, pp [] M. Balara, The upgrade methods of the pneumatic actuator operation ability, Applied Mechanics and Materials, Vol. 08, 01, p [5]. Daerden and D. Lefeber, Pneumatic artificial muscles: actuators for robotics and automation, European Journal of Mechanical and Environmental Engineering, Vol. 7, No. 1, 00, pp [6] CH-P. Chou and B. Hannaford, Measurement and modeling of McKibben pneumatic artificial muscles, Transaction on Robotics and Automation, Vol. 1, No , pp [7] M. Tóthová, J. Piteľ and J. Boržíková, Operating modes of pneumatic artificial muscle actuator, Applied Mechanics and Materials, Vol. 08, 01, pp. 9-. [8] luidic muscles DMSP-.../MAS-... [online] Available at: < SupportPortal/iles/561/75768d6.pdf>. [9] J. Boržíková, M. Balara and J. Piteľ, The mathematical model of contraction characteristic k = (, p) of the pneumatic artificial muscle, Proceeding of XXXII. Seminar ASR'007 "Instruments and Control", Ostrava, April 7, 007, Ostrava: VŠB-TU, pp [10] A. Hošovský, Numerical approximation of static characteristic of PAM-based antagonic actuator, Journal of applied science in the thermodynamics and fluid mechanics, Vol. 1, No. 1, 007, pp. 1-. [11] J. Sárosi, G. Szabó and J. Gyeviki, Investigation and application of pneumatic artificial muscles, Biomechanica Hungarica, Vol., No 1, 010, pp [1] J. Sárosi, New approximation algorithm for the force of fluidic muscles, Proceeding 7 th International Symposium on Applied Computational Intelligence and Informatics (SACI 01), Timisoara, 01, pp. 9-. [1] V.T. Jouppila, et al. Sliding mode control of a pneumatic muscle actuator system with a PWM strategy, International Journal of luid Power, Vol. 15, No. 1, 01, pp [1] T. Kerscher, J. Albiez, J.M. Zollner and R. Dillmann, Evaluation of the dynamic model of fluidic muscles using quick-release, Proceeding of International Conference on Biomedical Robotics and Biomechatronics (BioRob 006), Pisa, 006, pp [15] J. Boržíková, Non-linear approximation of the static characteristic = f(p,κ) of antagonistic system, Proceeding ARTEP 008, Stará Lesná, 008, Košice: TU, pp [16] V.T. Jouppila, S.A. Gadsden and A. Ellman, Modeling and identification of a pneumatic muscle actuator system controlled by an ON/O solenoid valve, Proceeding 7 th International luid Power Conference, Aachen, 010, pp [17] A. Hošovský and M. Havran, Dynamic modeling of one degree of freedom pneumatic muscle-based actuator for industrial applications, Tehnički vjesnik, Vol. 19, No., 01, pp ISBN: