Calibrating procedure by teach-in robotised flexible manufacturing system.

Radulescu, Corneliu ; Varga, Stefan ; Grigorescu, Sanda 等

Abstract: One of the industrial re-technology objectives is buying a great number of robotized Flexible Manufacturing System--FMS. Once the technical application is identified and the equipment is procured, a quick installation process is required. Although the robotized FMS components are executed with great precision, because of some important systematic situation errors, caused by foundation and installation imprecision, the duration of the first put in service can be extended very much, having an unjustified effect at the expenses increases. This paper presents a new calibration method based on teach-in implementation that reduces the losses caused by the difficulty of putting into service of the robotized FMS.

Key words: Flexible Manufacturing System, Robot, Calibration, Teach-in.

1. INTRODUCTION

Each robotized FMS contains at least one robot and a certain number of the peripheral components. If the robot co-operates with the peripheral components, the functional parameters must be automatically correlated. If the functional parameters are variable, they are correlated by special application software. If the functional parameters are constant, having a reference role, calibrating can correlate them. The calibration can be done using an outside-robotized FMS instrumentation, or using the robot sensorial capacity (Coiffet, 1991). In the second case, there is a teach-in implementation with many advantages because it eliminates the system's systematic errors.

The problem is to calibrate the FMS peripheral components relative situations, according to the robot that serves these components.

The calibrating methodology of the system depends on the robot's guiding devices and their peripheral components, and can be described with direct or inverse geometrical models (Hartenberg & Denavit, 1964), (Craig, 1986).

Other people use successive teaching of the application in every workplace served by robot. This method creates time losses, especially for complex applications, because a lot of precision points need to be defined.

2. THE EXPERIMENTAL DETERMINATION OF THE TRANSFORMATION MATRIX FROM A PERIPHERAL COMPONENT TO THE ROBOT GEOMETRICAL MODEL

If between the robotized FMS components, there are unknown positioning and orientation errors, the transformation matrix from a peripheral component to the robot, must be experimentally determined. In order to do this, on every peripheral component, there are some calibration points. The general form of the transformation matrix, introduce by its elements 12 unknown variables. Because one spatial point situation needs three coordinates, for one peripheral component there are 12 / 3 = 4 calibration points (O, Q1, Q2, Q3). The 4 calibration points are chosen in different planes, in order to define on every peripheral component a new frame. In the teach-in phase these calibration points must be effectively materialized. In order to do this, some pegs or markers are fixed on the every peripheral component calibrate base that will be used. After the four calibration points coordinates are established, the elements of the transformation matrix are calculated (Varga, 2001).

We have done two calibrating algorithms based on transformation matrixes used to adapt the robotized FMS, with a real positioning of the components, at application programs elaborated "off line" or "on line" in the other system, with different real components positioning using the same application.

These algorithms can be implemented in the robot software, robot that was mounted with situations errors according to peripheral work place. The algorithms allow an automatically software adaptation of what was initially defined for a similar FMS, but with different layout errors. Using the "on-line" acquisition of the calibration points coordinates, for the initial FMS and for the new FMS, the calibrating algorithms can automat calculate the transformation matrix from the every peripheral components to the robot frame. With these matrixes the initial FMS software databases are redefined, resulting a new database, valid to command a new FMS robot.

2.1 Using the calibrating algorithm to implement an "off line" elaborated application program inverse geometrical model

This algorithm is based on the premise that in the "off line" elaborated application program, the precision points coordinates are expressed in comparison with the frame, united with ideal robotized FMS peripheral component. The algorithm needs to execute the following steps:

(1) By "on line" sensing the calibration points positions materialized on real robotized FMS peripheral component are measured.

(2) Applying the relations of the direct geometrical model the co-ordinates of the every calibration points can be determinates.

(3) The elements of transformation matrix are calculated.

(4) The application program is loaded into the robot control system computer.

(5) The precision points co-ordinates, defined on the application program as a database, are transposed from the calibration points frame into the reference frame.

(6) The inverse geometrical calculus is done.

(7) The impulses numbers are deduced for every axis and every precision point, and they are memorized into a new database, because they are the reference values of the control signals on the energetic axis numerical control. This calculus must be done once for each program, after that the robot from real FMS can execute the application, with the new database.

2.2 Using the calibrating algorithm for "off line" implementation of the "on line" elaborated application program inverse geometrical model

To formulate this algorithm, two theoretically identical robotized FMS must exist, made from the same layout, but practically they are with different real situation of the components. It is supposed that on these two robotized FMS theoretically identical applications must be executed, after an "on line" elaborated program on one of them (Fig. 1). In this case the algorithm to follows reduces the calibration program. It is important that the "on line" elaborated program into the first robotized FMS, defines the application precision points by the impulses numbers proportional with this system robot joint variables. This algorithm needs to run the following steps:

(1) Based on the impulses numbers for every energetic axis and every precision point, defined for the considered application, the joint variables are determined.

(2) With the resulted joint variables we calculate for all the precision points' co-ordinates in comparison with the fixed frame, which is united with the first robot.

(3) Applying the (1), (2) and (3) phases presented in [section] 2.1, for the first robotized FMS, the transformation matrix is determined from the peripheral component of the first system to their robot.

(4) The inverse of this transformation matrix are done, and the precision points co-ordinates are transformed from robots frame in Q1 origin frame, that is united with the first robotized FMS peripheral component basis. Because the two robotized FMS execute the same application, having the same precision points P1 [equivalent to] P2, results that their co-ordinates relative to the frames, connected to the two robotized FMS have to be the identical. As a result they can be transferred from the first to the second FMS and the problem is same like the case presented in [section] 2.1.

The robot will co-operate correctly with the second FMS peripheral component, where the relative situation between the components differ from that found in the first system, only after applying the seven phases of the algorithm presented in [section] 2.1, but using the second robot and the calibration points materialized on the second robotized FMS.

These algorithms can be adapted, with minor changing, on different robotized FMS layouts. In the FMS presented in Fig. 2, one single robot serves two theoretically identical peripheral components, but with different real positioning relative to the robot. In this case the only modification that has to be operated in the calibrating algorithm presented in this section is that we can skip the program transfer phase. After the "on line" program elaboration and the application precision points definition relative to the Q1 peripheral component, the Q2 component calibration can be made, in order to adapt the joint variables to the real situation of the robot into origin O.

In the end it is mentioned that the presented algorithms depend on the robotized FMS structure. A complex guiding device, or a peripheral component with proper mobility, lead to modifying the system geometrical models, so that the parameters number taken in account increases.

[FIGURE 1 OMITTED]

3. ESTIMATING THE TECHNICAL AND ECONOMICAL EFFICIENCY OF THE CALIBRATION

It is known that the applications program transfer and the calibrating programs of the systems with more components that execute the same applications represent efficient techniques to reduce the duration of the first put into service of the robotized FMS and also the related expenses. The time gain can be estimated by the ratio between the necessary precision points into the system (as shown, there are 4 points needed for each component) and the needed precision points that define the application.

The robotized FMS manufacturer use the beneficiary staff teaching to do on a prototype system temporarily installed in their testing workshop. At this moment the prototype system is programmed for the required application. Usually this program is lost. If this program is saved and transferred to the beneficiary and it is adapted through calibration to the components' real relative positioning, after their reinstallation at the beneficiary, the application programming time can be saved, considering just the first component.

Of course, the greatest economies are not done thanks to the reduced programming time for a calibrated FMS they are done because of the rapid introduction of the FMS into the fabrication process where it starts generating profit. Instead of calibrating the FMS components physical layout (weight objects precise moving, vague foundation, etc.), our method propose the software calibration on every imprecise layout. In this manner, this new calibrating algorithms replace the hard work of many people with the easier computer work.

It is obvious that in future it is needed to elaborate, for the robotized FMS calibrating, the "tool" program, that allows to adapt the application program in a short time. These "tool" programs have to operate into the database, that have archived the information about application precision points, from the initial application program, to recognize this information, to decode them, to modify the acquisition reference values on one of the peripheral component and transform these into needed values for other components, to re-code and to re-archive into another database in the same order as they are extracted. The application program resulted for the real situation of the components from the calibrated system, can be extended to the many robotized FMS, and must be accessible in automatically mode on every application execution.

If one robot serves more theoretically identical peripheral components, but placed into different real situation, the application that has to be executed is the program used for the calibration of the peripheral component that is put into service. In future we can be use the 3D CompuGauge System--Dynalog, Inc., (bought by Grant 112 CEEX II-03/15.09.2006--Simulation, Control and Tasting Platform applied in Mechatronics, CNMP Bucuresti INFOSOC), to know the real positions of the 4 calibration points on the every peripheral components of the FMS.

[FIGURE 2 OMITTED]

4. CONCLUSIONS

The calibration of the robotized FMS serves to obtain an adjustment of the positioning parameters of the system components. Into the calibrated FMS, the components automatically co-operate, following the application programs, even if they are placed with relatively great positioning and orientation errors.

The calibration can be made by "on line" teach-in of the system. In order to do this, on the robot peripheral components 4 non-coplanar calibrations point must be materialized and there position are recognized by sensing.

The calibration algorithms give the "tool" program that allows controlling the system computer in order to automatically adjust the application precision points to the real situation of the FMS components.

By the FMS calibration a lot of time is gained, accelerating the first put into service, and also profit can be gained by accelerating the start the production.

5. REFERENCES

Coiffet, Ph. (1991) Les Robots, Hermes Publishing, Paris

Craig, J.J. (1986) Introduction to Robotics, Addison--Wesley Publishing, ISBN 0-201-10326-5, Massachusetts

Hartenberg, R.S. & Denavit, J. (1964) Kinematic Synthesis of Linkages, Mc. Grow--Hill Book Company, New York, San Francisco, Toronto, London

Varga, S. (2001), Zero Rank Calibrating Procedure by Teach-In Robotized Flexible Manufacturing System, International Journal "Robotics & Management", Vol. 6, No. 1, June 2001, pp. 16-21, ISSN 1453--2069