CN105773628A

CN105773628A - Wax spraying control method of LED chip waxing robot

Info

Publication number: CN105773628A
Application number: CN201410779990.3A
Authority: CN
Inventors: 王佐勋
Original assignee: Qilu University of Technology
Current assignee: Qilu University of Technology
Priority date: 2014-12-17
Filing date: 2014-12-17
Publication date: 2016-07-20

Abstract

The invention designs an LED chip wax spraying movement control system and designs a wax spraying movement control method according to the system. By the wax spraying movement control method, a shortest movement path is planned according to the control system and target points required to be subjected to wax spraying in a wax spraying process. A movement track of a two-joint robot is planned by a five-order polynomial for each section of movement paths, real-time learning is implemented by a neural network self-adaption controller, and then tracking control to the movement track is finished. By the method, movement control of wax spraying in an LED chip waxing process is realized well, wax spraying precision is improved, and production efficiency and product percent of pass of LED chips are greatly improved.

Description

A kind of LED chip waxing machines people's wax spray control method

Technical field

The control method of the present invention is exclusively used in LED chip production process the motor control of wax spray.

Background technology

The control of LED chip waxing machines people's wax spray is before LED paster, is first sprayed onto on pasting boards by a certain amount of wax, and wax spray robot needs the most accurately wax to be sprayed onto predetermined position, this motion path and movement locus of being accomplished by studying wax spray robot.The motion of robot has been carried out some researchs by expert both domestic and external, and the modified Tabu search algorithm with disturbance and mutagenic factor that utilizes having sets up plan model, optimizes mounting order, planning attachment path.The designed robot having utilizes sensor senses oneself location, and plans the path of oneself, arrives impact point in an optimal manner.Some algorithms adopting ant group algorithm and shuffled frog leaping algorithm to blend, the path of planning paster robot, it is achieved optimum control.Some employing B-spline methods, it is achieved that multiobject track optimizing, have been finally reached time optimal, the purpose that flatness is optimum.What have proposes a kind of new robot task spatial control and learning strategy, not only achieve the ultimately uniform boundary of all signals of closed loop system, and in stable control process, it is achieved that partial nerve network weight converges to optimal value and the dynamic local of unknown closed loop system is accurately approached.Some application radial base neural nets approach the kinetic model of underwater robot, finally achieve the tracing control of track.Have for Uncertain nonlinear chaos system, propose a kind of adaptive Gaussian filtering new method based on dynamic neural network identifier, technology on-line tuning dynamic neural network identifier weights are controlled by sliding formwork, and obtaining, optimal controller is designed on the basis of dynamic neural network model, it is achieved the track of chaos system is followed the tracks of.Some use RBF (RBF) neutral nets and High-gain observer devise a kind of adaptive neural network control algolithm. and this algorithm not only achieves the ultimately uniform boundary of all signals of closed loop system, and along period tracking path implementation, unknown closed loop system determined study dynamically.What have has unknown nonlinear function and the Nonlinear Second Order System of unknown dummy coefficient nonlinear function for a class, proposing a kind of Neural network robust self adaptation Output Tracking Control method, the neural network weight adjustment law of selection is possible to prevent the parameter drift in Self Adaptive Control.Propose a class and there is external interference and probabilistic mechanical arm track following kinds of robust control problems.Controller is made up of self_adaptive RBF (radialbasisfunction) nerve network controller and PD controller. and adopt the RBF algorithm based on neuron sensitivity and triumph neuron concept, determine initiating structure and the parameter of neutral net online.

The present invention is directed to the motor system of wax spray robot in LED paster process, adopt the movement locus of the shortest path planning method planning robot of Optimum Theory, the motion for wax spray robot adopts higher order polynomial to plan the motion in joint.Under the guidance determining theory of learning, adopt a kind of neural network control device of RBF design, control the motor process of robot along the tracing of the movement planned, so can ensure that the stationarity of robot motion, rapidity, reach the saving time and save the purpose of energy.

Summary of the invention

First the present invention devises a set of kinetic control system of wax spray technique suitable in LED chip production process, and the kinetic control system of waxing machines people is mainly made up of 2 motors, 2 encoders, 2 leading screws and workbench, and wax sprayer is arranged on the table.In motor control process, first carrying out path planning, utilize shortest path planning principle, as shown in Figure 1 from O point, through n stage, the decision-making in each stage is to select next point, with (x_i, X) and represent that state, xi represent location point, X is the some set having not gone through, at state (x_i, X) decision-making set X in, depend on plan x_j X, it is thus achieved that benefit be x_iTo x_jDistance d_i,j, then proceed to next state (x_j, X { x_j).Use f_k(x_i, X) and represent that, from O point, the point in X is each once, eventually passes back to the short line of O point.X is a vertex set, and first prime number of X is k, d_i,jIt it is the distance of xi to xj.Being computed, finding out shortest path is: wax spray robot needs from O point, respectively at A, B, C, D, E, F, G, H, I, J point wax spray, returns O point.Then with the movement locus of 5 rank multinomial planning two-articulated robots each section, movement locus is divided into boost phase, constant velocity stage, decelerating phase, during acceleration as far as possible steadily, retarded motion is just 0 to target location hourly velocity, and the time used by whole process to be lacked as far as possible.Plan the track of OA, AB, BC, CD, DE, EF, FG, GH, HI, IJ, JO motion path successively.Then neural network control Algorithm Learning is adopted to follow the tracks of the movement locus of planning, definition filter function tracking error, select to control input function and control the motion of robot, be determined inquiry learning with the neural network node of N=5*5*5*5, draw output controlled quentity controlled variable.

By implementing this control system and this control method, can effectively control the wax spray process during LED chip produces, make wax spray robot stable movement, target can be searched out accurately, decrease the vibrations in robot kinematics, improve the service life of robot, improve quality and qualification rate that LED chip produces, substantially increase the efficiency that LED chip produces.

Accompanying drawing explanation

Accompanying drawing 1 is wax spray point schematic diagram

In figure: O is robot stop position, A, B, C, D, E, F, G, H, I, J are 10 wax spray points.

Accompanying drawing 2 is wax spray robot Motion Control Platform schematic diagram.

In figure: 1 is bearing B, and 2 is wax sprayer, 3 is wax sprayer erecting bed, and 4 is leading screw B, and 5 is motor A, 6 is encoder A, and 7 is encoder B, and 8 is motor B, and 9 is leading screw A, and 10 is bearing A, 11 is servo-driver A, and 12 is panel, and 13 is servo-driver B, and 14 is workbench.

Embodiment

With reference to Figure of description, the control method of the present invention is done following description in the control of LED chip production process wax spray.Concrete operations are:

1) path planning.Adopting the shortest path dynamic programming method in principle of optimality, the motion path of robot is planned, finding out optimal path is: O → A → B → C → D → E → F → G → H → I → J → O.

2) trajectory planning.LED chip wax spray robot from some action moving to another point is: motion, stopping, wax spray, move again.It is directed to this course of action, the movement locus of planning robot, in order to ensure the motion of quick and stable, and the movement velocity arriving operating point is 0, boost phase, constant velocity stage, decelerating phase is divided, with the movement locus of 5 rank multinomial planning robot's doublejointed motor process.

3) Motion trajectory of doublejointed robot is exactly that two square joints are applied 5 rank multinomial planning, the method adopting vector summation after allowing respectively respectively, adoptsCook up the movement locus of robot.

4) design neural network control device.Design neural network control device initially with gaussian radial basis function (RBF), then definition filter tracks error r (t), select to control input function, design neural network function, finally design right value update rule。

5) movement locus of tracing control wax spray robot.The number of network node choosing neutral net isN=5*5*5*5, on-line tracing learns, and along with the output in error and error rate self-adaptative adjustment neural network learning process during study, enables quickly to follow the tracks of the movement locus of wax spray robot.

Claims

1. the key point of the present invention is in that to devise the wax spray kinetic control system of a kind of LED chip waxing machines people, it is directed to the impact point needing wax spray in this control system and wax spray process, plan the shortest motion path, it is directed to each section of motion path, propose the movement locus with 5 rank multinomial planning two-articulated robots, adopt with neural network adaptive controller real-time learning, complete the tracing control to movement locus.

2. planning shortest path according to claim 1, adopts the shortest path planning method of principle of optimality, finds out the shortest path of 10 wax spray location points.

3. the Motion trajectory of two-articulated robot according to claim 1, adopts 5 rank multinomials to plan the motion in each joint, adopts the mode of vector sum to obtain the movement locus of two-articulated robot.

4. neural network control device according to claim 1, designs neural network control device initially with gaussian radial basis function (RBF), then definition filter tracks error r (t), selects to control input function, design neural network function, finally design right value update rule.

5. tracking and controlling method according to claim 4, the number of network node choosing neutral net isN=5*5*5*5, on-line tracing learns, and along with the output in error and error rate self-adaptative adjustment neural network learning process during study, enables quickly to follow the tracks of the movement locus of wax spray robot.