An Algorithm for the Welding Torch Weaving Control of Arc Welding Robot

Transcription

1 ELEKTRONIKA IR ELEKTROTECHNIKA, ISSN , VOL. 2, NO. 2, 205 An Algorithm for the Welding Torch Weaving Control of Arc Welding Robot Wen-liang Zhu, Fu-sheng Ni 2, Cao-gen Hong 3 College of Harbor Coastal and Offshore Engineering, Hohai University, Nanjing 20098, China 2 Engineering Research Center of Dredging Technology, Hohai University, Changzhou 23022, China 3 Jiangsu Automation Research Institution, Lianyungang, , China zliang200@qq.com Abstract The teaching learning method is generally applied in six DOF (degrees of freedom) robot programming. Hoever, for complicated eaving elding, there is a draback using this method. That is that numerous teaching control points are needed in the programming. Based on a method of the robot trajectory planning, this paper focus on addressing this draback by utilizing an algorithm for the elding torch eaving control. This algorithm includes the novel formulation to calculate the tracking position, and can produce an accurate and continuous eaving path along various eld lines. The teaching programming can easily determine these eld lines hich contain fe teaching points. Therefore, this algorithm can reduce the programming time for the robotic elding application. A lot of experiments have been done to validate the algorithm. The experimental results sho that the eaving control algorithm is correct and can produce various eaving paths. Index Terms Robot control, elding, motion planning, softare algorithms. I. INTRODUCTION The development of robotic elding has been truly remarkable and is one of the major application areas for industrial robots noadays, T. S. Hong et al. [] discusses many advantages and disadvantages of different robotic elding technologies. Arc elding is considered to be one of the most promising applications of intelligent robots. This situation first stems from a lo manual productivity due to the severe environmental conditions resulting from the intense heat and fumes generated by the elding process. Second, arc elding is the third largest job category behind assembly and machining in the metal fabrication industry [2]. Robot eaving elding technology is a typical application of arc elding industry, so it encourages intensive research and development of sophisticated control algorithm and application-specific softare package. J. Wang et al. [3] developed a novel sing arc system to realize high quality narro gap elding at lo cost. This system uses a motor of hollo axis to turn directly the micro-bent conductive rod and then to eave circularly the arc around itself axis of torch. The method of the robot trajectory planning can also be used Manuscript received March 5, 204; accepted October 0, 204. for robot eaving elding control, but little ork of this method as reported currently. This paper presents an algorithm for the elding torch eaving control of arc elding robot by employing this method. In an arc elding application, a task planning is the specification of the geometrical and the technological parameters. A task planning scheme for an arc elding robot must employ the folloing parts. First, elding scheme definition; second, elding procedure specification for each individual eld; third, motion planner. The motion planner is used to produce a motion scheme that can be executed by a robot controller [4]. The motion scheme is generally achieved by the method of teaching programming. Hoever, for complicate eaving elding, especially for Multi-Layer and Multi-Pass elding, hich has been reported in [5] [6]. There is a draback using this method, that is that numerous teaching control points are needed in the programming. Based on a method of the robot trajectory planning, this paper focus on addressing this draback by utilizing an algorithm for the elding torch eaving control. This algorithm includes the novel formulation to calculate the tracking position, and can produce an accurate and continuous eaving path along various eld lines including a complicated combination of straight lines and arcs. The teaching programming can easily determine these eld lines hich contain fe teaching points. Therefore, this algorithm can reduce the programming time for the robotic elding application. With a eaving elding robot, the programming of a eaving path is usually performed in to steps. First, teaching program determines a eld line. This eld line is both the centre of the eaving path and the offset benchmark of Multi-Layer and Multi-Pass elding. The eld line is discretized into a set of 3D (three-dimensional) positions. Second, the set of 3D positions is calculated through the algorithm and is mapped from Cartesian space to joint space ith the inverse kinematic model of the elding robot[3]. The robot controller updates the moving reference coordinate system ithin each cycle time to achieve the expected motion path. The remainder of this paper is organised as follos. Section II describes the real-time moving reference 3

2 ELEKTRONIKA IR ELEKTROTECHNIKA, ISSN , VOL. 2, NO. 2, 205 coordinate system determining the eaving path and introduces the tool frame from actual TCP (tool centre point)-orientation. Section III is devoted to a formal statement of the eaving control algorithm. In Section IV, a lot of experimental validations for the algorithm are presented. Finally, Section V summarizes the main contributions of this paper. II. PROBLEM OF COORDINATE TRANSFORMATION A. Reference Coordinate System The use of Cartesian coordinates for the description of positions of a robot path is an important step in simplifying programming. The next step is the use of Cartesian reference systems. Reference systems consisting of position and orientation offsets are often called frames or transformations in robotics. As shon in Fig., usually, a six DOF (degrees of freedom) robot have four reference systems, such as the orld frame W, the base frame B, the object frame O and the tool frame T. The common root of all reference systems in the applications ith a single robot is called orld frame W, hich is usually chosen in the base of the robot, i.e. the base frame B. In many elding tasks the tool tip of a robot has to move to several different positions on the same elding object. When defining the positions of the robot eaving path relative to orld frame W, all positions on the elding object can be translated together by modifying the position of the reference system. As shon in Fig. 2, the eaving path and eld line are relative to orld frame W. The points beteen p and p2 can be calculated by K. Zhang interpolation algorithm [7]. Because a straight line can be determined by to teaching control points, the eld line can be achieved by teaching easily. Hoever, the eaving path cannot be achieved by using the same method. The positions of the eld line are corresponding one to one ith the positions of the elding path, just like the vector q and q. All positions of the eaving path can be achieved by modifying positions of a number of the reference coordinate system, such as frame, frame2, frame3,, under the condition of not changing the target command position in the robotic program. This coordinate transformation can be described by the follo equations: here q q q q, o q q q, () (2) q q, (3) is the position vector of the eaving path relative to the orld coordinate system. vector. q O q is called the tracking is called the tracking position, i.e. the origin of the robotic activated reference coordinate system. Without offset and rotation, q q O ith offset and rotation the relation expression of them ill be described in the Section III. The left superscript W refers to the reference coordinate system of the vector, the right superscript denotes the starting point of a vector (blank space shos that the start point is the origin of reference coordinate system) and the right subscript denotes the ending point of a vector. When the activated reference coordinate system of the robot continually moves from frame to frame2, to frame,, a continuous eaving path based on the eld line can be obtained. Fig.. The reference coordinate system of robot. Fig. 2. Coordinate transformation ithout offset and rotation. B. Definition of the Tool Frame A. P. Pashkevich has described in detail the definition of the eld frame for the linear and circular eld joints in [8]. In order to suit various eld joints, a similar approach as used to definite the tool frame in the present study. For the linear eld joints, a moving frame ith the specific TCP-orientation is defined, as shon in Fig. 3. For the circular eld joints, a moving frame is calculated from the changing TCP-orientation to ensure the tangency of the elding path and the X T-axis at every point, as shon in Fig. 4: the X T-axis is directed along the eld joint; the Z T-axis specifies the elding torch approaching direction (hich is normal to the eld joint); the Y T-axis completes the right-hand oriented frame and thus shos the eaving direction. Three coordinate axis unit vectors can be computed by rotation matrix T R from actual TCP-orientation T n, T o, T a T R cacbcc - sasc -cacbsc - sacc casb sacbcc casc - sacbsc cacc sasb, -sbcc sbsc cb (4) 4

3 ELEKTRONIKA IR ELEKTROTECHNIKA, ISSN , VOL. 2, NO. 2, 205 here T n is the XT-axis unit vector, T o is the YT-axis unit vector, T a is the ZT-axis unit vector. T R is the tool frame rotation matrix relative to the orld frame. The variable a, b, c is three Euler angles of TCP-orientation. The symbols in the matrix are described as follos: sa sin(, ca cos(, sb sin(, cb cos(, sb sin( c), cc cos( c). (5) T v v x y v v z path, vx vy vz here T a has the same meaning as in (4). The binormal vector is the unified unit vector normal to the path pbi velocity vector and Z T-vector. So, the tracking vector can be calculated by the folloing equation (7) q q A p bi, (8) here the parameter A denotes actual oscillation amplitude. The parameter refers to actual amplitude of oscillation function. The range of is,. There are three eaving types including sine oscillation function, triangle oscillation function and trapezoid oscillation function. For sine oscillation, is defined as sin( ), here is phase angle. For triangle oscillation, is defined as follos: Fig. 3. Moving tool frame ith specific TCP-orientation. Fig. 4. Moving tool frame ith changing TCP-orientation. III. THE ALGORITHM FOR THE WEAVING CONTROL if 0, 2, 2 3 if, 2 2, if, For trapezoid oscillation, is defined as follos: if 0, 3, 3 2 if,, if, 3 3, if,, if 2, (9) (0) A. The Weaving ithout Offset and Rotation Without offset and rotation, q q O. As long as the tracking vector q is calculated, the eaving path can be achieved by activating the reference coordinate system of arc elding robot hose origin is the tracking position q O. In the eld procedure, the direction of oscillation is along the binormal vector p hich is determined by the path velocity unit vector vpath bi equitation is as follos here vpath is defined as and the Z T-axis unit vector T a. The pbi vpath T a, (6) B. Offset of the Weld Line Welding operations planning can be classified into four types according to the number of eld lines and the number of layers as shon in Table I [9]. TABLE I. FOUR TYPES OF ROBOTIC WELDING OPERATIONS PLANNING. Single-Layer Multi-Layer Single-Pass Single-Layer and Single-Pass Multi-Layer and Single-Pass Multi-Pass Single-Layer and Multi-Pass Multi-Layer and Multi-Pass When the eld line moves along Y T -axis of the tool frame, the Multi-Layer and Single-Pass elding is achieved. When the eld line moves along Z T-axis of the tool frame, this eld operation is defined as Single-Layer and Multi-Pass eld. If 5

4 ELEKTRONIKA IR ELEKTROTECHNIKA, ISSN , VOL. 2, NO. 2, 205 both Y T axis and Z T-axis have an offset, this eld operation is called Multi-Layer and Multi-Pass elding [0]. The offset vector can be obtained by the folloing equation T qos vpath pbi T a qos, () here the three unit vectors v path, pbi, T a have the same definition as in (6). The vector qos denotes actual y-offset and z-offset for Multi-Layer and Multi-Pass elding relative to the orld frame. The vector T q denotes actual y-offset and z-offset for Multi-Layer and Multi-Pass elding relative to the tool frame, it is defined as os T. T q os x t y t z t (2) 2 c ( c ) v x Rz ( c ) vy vx s vz ( c ) vz vx s vy ( c ) vx vy s vz ( c ) vx vz s v y 2 c ( c ) vy ( c ) vy vz s vx, ( c ) v 2 z vy s vx c ( c ) vz (7) here the parameter is the rotation angle around the path velocity unit vector. The parameter is the rotation angle around Z-axis unit vector of the tool frame. vx, vy, vz, T ax, T ay, T az are the elements of the vector vpath and T a. The symbols c, s, c, s are defined as: Generally, x-offset is not implemented, because x-offset is a starting-point offset and is not a eld path offset. As shon in Fig. 5, the tracking position ith eld line offset can be obtained by follo equation qo q qos A pbi qos. (3) c cos( ), s sin( ), c cos( ), s sin( ), (8) C. Rotation of the Weaving Path As shon in Fig. 6, besides the eld line offset, it is necessary that rotation around the coordinate axes of the tool frame ensures to adapt to various eld joints. In this paper, the eaving path rotates around the X-axis and Z-axis of the tool frame, rotation around Y-axis is not implemented. For X-rotation, namely rotation around path velocity vector, the tracking position ith the eld line offset and X-rotation can be obtained by the folloing expression O x bi os q R ( A P ) q. (4) For Z-rotation, the tracking position ith the eld line offset and Z-rotation is described as follos O z bi os q R ( A P ) q, (5) Fig. 5. Coordinate transformation ith offset. here qos represents the eld line offset. Rx denotes 3 3 rotation matrix around the path velocity unit vector v path. Rz denotes 3 3 rotation matrix around Z T-axis unit vector T a of the tool frame. The rotation matrix x as follos: R, 2 c ( c ) v x Rx ( c ) vy vx s vz ( c ) vz vx s vy Rz is defined ( c ) vx vy s vz ( c ) vx vz s v y 2 c ( c ) vy ( c ) vy vz s vx, ( c ) v 2 z vy s vx c ( c ) vz (6) Fig. 6. Coordinate transformation ith rotation. IV. EXPERIMENTAL VALIDATION A. Experiment of Three Weavings To validate this algorithm, a softare package as developed on six DOF robotic control system hich as developed by Jiangsu Automation Research Institution. The experimental environment as shon in Fig. 7. The parameters of the Experiment are described as follos: The eaving amplitude A (0 milimeters); the eaving frequency (0.5 Hz); the eaving type is set sine, triangle and trapezoid respectively; the parameters of offset and rotation are set to zero. The eld line is obtained by teaching to points (i.e. the start point and end point of the eld line). The experiment data of three eavings ere shon in 6

5 ELEKTRONIKA IR ELEKTROTECHNIKA, ISSN , VOL. 2, NO. 2, 205 Fig. 8 Fig. 0, including the time history of the tracking position, the 2D (to-dimensional) diagram of TCP positions. Fig. 9. Tracking position ( and TCP position of triangle eaving ithout offset and rotation (. Fig. 7. The experimental environment. Fig. 0. Tracking position ( and TCP position of trapezoid eaving ithout offset and rotation (. It can be seen from Fig. 8(, Fig. 9( and Fig. 0( that hen A = 0, the maximum eaving amplitude of the first element X of the tracking position is 0. The result of eaving can be seen from Fig. 8(, Fig. 9( and Fig. 9(. The X coordinate of the eld line is 049, the oscillation function of the algorithm can accurately produce sine curve, triangle curve and trapezoid curve respectively. Fig. 8. Tracking Position and ( TCP Position of sine eaving ithout offset and rotation (. B. Weaving Experiment ith The Weld Line Offset Due to page limitation, only sine eaving data records ere provided. The parameters of the Experiment ere described as follos: The eaving amplitude A (0 millimeters); the eaving frequency (0.5 Hz); the eaving type as set sine; the Y-offset (-80 millimeters); the Z-offset (30 millimeters); the parameters of rotation as set to zero. The eld line as obtained by the same approach of experiment ithout offset and rotation. As shon in Fig. 2 Fig. 3, including the time history of the tracking vector and the tracking position, the 2D diagram of TCP positions. 7

6 ELEKTRONIKA IR ELEKTROTECHNIKA, ISSN , VOL. 2, NO. 2, 205 c) Fig. 2. Tracking vector (, Tracking position ( and TCP position of sine eaving ith Z-offset (c). It can be seen from Fig. (, Fig. (, Fig. 2( and Fig. 2( that hen the eaving have an offset of the eld line, comparing ith the tracking vector, the tracking position produce an offset element, such as Y-offset is 80 in Fig. (, Z-offset is 30 in Fig. 2(. The result of eaving can be seen from Fig. (c) and Fig. 2(c). The Y coordinate of the eld line changes from -440 to The Z coordinate of the eld line changes from 778 to 808. The oscillation function of the algorithm can accurately produce to sine curves. The photo of the eaving experimental result is shon in Fig. 3. c) Fig.. Tracking vector (, Tracking position ( and TCP position of sine eaving ith Y-offset (c). Fig. 3. The experimental result photo. C. Weaving Experiment ith X-rotation or Z-rotation The parameters of this Experiment ere described as follos: The eaving amplitude A (0 millimeters); the eaving frequency (0.5 Hz); the eaving type as set sine; the X-rotation (30 degrees); the Z-rotation (30 degrees); the parameters of offset is set to zero. The eld line is obtained by the same approach of experiment ithout offset and rotation. The sine eaving data records ere provided. As shon in Fig. 4 Fig. 5, including the time history of the tracking position and the 2D diagram of TCP positions. 8

7 ELEKTRONIKA IR ELEKTROTECHNIKA, ISSN , VOL. 2, NO. 2, 205 Z-axis. Fig. 4. Tracking position ( and TCP position of sine eaving ith X-rotation (. V. CONCLUSIONS The experimental results shoed that the algorithm for the elding torch eaving control satisfied the control of three eaving paths based on a complicated eld line, and achieved the complex spatial offset and rotation. It is possible to implement the Multi-Layer and Single-Pass elding path for both the linear and the circular joints by setting eaving parameters in robotic program. By using the developed softare package, it is also possible to reduce the programming time for the robotic elding application. The particular contribution of this paper describes a moving reference coordinate system based on the theory of coordinate transformation, hich is a key issue in the kinematic control of the elding robotic control system. When the robot activates the corresponding reference coordinate system, the expected eaving path can then be achieved by implementing robotic motion commands. The algorithm in this paper had been applied in the softare module of JARI robotic control system. This application reduced greatly the teaching time in arc elding robot programming. In the future, the research ork ill focus on the analysis of saving the teaching programming time and promoting efficiency. Fig. 5. Tracking position ( and TCP position of sine eaving ith Z-rotation (. As shon in Fig. 4( and Fig. 5(, the tracking position as computed by (2) and (3). The result of eaving ith X-rotation can be seen from Fig. 4(, the eaving plane has rotated 30 degrees around X-axis hose Y coordinate is The result of eaving ith Z-rotation can be seen from Fig. 5(, the eaving path has rotated 30 degrees around REFERENCES [] T. S. Hong, M. Ghobakhloo, W. Khaksar, Robotic elding technology, Comprehensive Materials Processing, vol. 6, pp , 204. [Online]. Available: [2] P. Sicard, M. D. Levine, An approach to an expert robot elding system, IEEE Trans. Systems, Man and Cybernetics, vol. 8, 988, pp [Online]. Available: [3] J. Y. Wang et al., A sing arc system for narro gap GMA elding, ISIJ International, vol. 52, no., pp. 0 4, 202. [Online]. Available: [4] G. Bolmsjo, G. Nikoleris, Task planning for elding applications, in Conf. Proc. IEEE Int. Conf. Systems, Man and Cybernetics, Systems Engineering in the Service of Humans, 993, pp [Online]. Available: [5] S. B. Pan, B. H. Jia, X. W. Zeng, Welding operation of thick plate on the steel constructure, Electric Welding Machine, vol. 42, no. 8, pp. 5, 202. [6] R. Y. Li et al., A study of narro gap laser elding for thick plates using the multi-layer and multi-pass method, Optics & Laser Technology, vol. 64, pp , 204. [Online]. Available: [7] K. Zhang, Z. Liang, E. Li, D. Yang, Research on an arc elding robot controller ith open architecture and the sing-elding interpolation algorithm, Int. Conf. on, IET Advanced Technology of Design and Manufacture (ATDM 200), 200, pp [8] A. P. Pashkevich, A. B. Dolgui, K. I. Semkin, Kinematic aspects of a robot-positioner system in an arc elding application, Control Engineering Practice, vol., pp , [Online]. Available: [9] D. W. Kim, J. S. Choi, B. Nnaji, Robot arc elding operations planning ith a rotating/tilting positioner, International journal of production research, vol. 36, pp , 998. [Online]. Available: [0] M. K. Jouaneh, Z. Wang, D. A. Dornfeld, Trajectory planning for coordinated motion of a robot and a positioning table. I., IEEE Trans. Path specification, Robotics and Automation, vol. 6, pp , 990. [Online]. Available: 9