25 Ağustos 2007 Cumartesi

Sensor-guided robotic laser welding for 3-D seam welding (CRAFT project SenseLasBot) - Ir. M.W. de Graaf

Project background

Current laser welding techniques put high demands on the manipulator that is used. For most applications, accuracies in the order of 100 µm have to be reached at velocities up to 250 mm/s. It is not feasible to program industrial robots that accurate in a production environment. Sensors are needed to increase the positional accuracy and ease-of-use of a robotic welding cell.


OBJECTİVES
The main objective of this research project is the development of a fully automated robotic laser welding cell by integrating a Nd:YAG laser, an industrial 6-axis robot, laser welding optics, an optical seam-tracking sensor and process sensors (Figure 1). Both a commercial seam-tracking sensor (Falldorf GmbH) and the developed integrated welding head of Iakovou will be considered. The system uses sensor information for teaching of the welding trajectory. Furthermore it measures the seam trajectory in real-time during the robot motion and compensates for positioning errors, hereby making robot programming easier and faster.

Seam tracking
An optical seam-tracking sensor is used to accurately measure the seam trajectory in the product to be welded. The commercial sensor uses the principle of optical triangulation (Figure 2) to measure four coordinates (3D position and 1 rotation angle) of the weld joint with respect to its coordinate frame. The integrated welding head of Iakovou can even measure six coordinates (3D position and complete orientation). To be able to use the seam-tracking sensor while the robot is moving, the sensor measurements and the robot joint position measurements have to be synchronized. Algorithms are being developed that can teach an unknown seam trajectory while the robot is moving and track an approximately known seam trajectory during laser welding. For this development and subsequent testing, a simulation environment has been realized, where the effect of non-ideal behavior like sensor errors, robot geometric errors etc can be simulated.

System architecture

The system components are shown in Figure 3. There is a separation between tasks that have to be carried out in real-time (hardware control during welding, seam tracking) and tasks that don’t necessarily have to be in real-time (seam teaching, tool calibration, product frame calibration). The software framework is designed in such a way that it is independent of the hardware used. Different robots and sensors can therefore be used

LASER
A Graphical User Interface, called 24-Laser has been developed. With this GUI the robot can be programmed and welding jobs can be carried out. Furthermore a number of dedicated tasks have been implemented
·
Laser and sensor tool calibration

·
Manual robot motion control

·
Sensor-guided seam-teaching using point-to-point movements

·
Real-time seam-tracking

·
Product frame calibration



24-Laser communicates with the robot and sensor through commonly used Ethernet-sockets. Only a part of the software, necessary for communication and real-time control of the hardware, is located at the robot controller, making this GUI robot and sensor independent.

Research objectives

·
Development of intelligent seam-teaching algorithms for easy robot-programming.

·
Development of real-time seam-tracking algorithms for faster robot-programming and to compensate for product clamping errors, etc.

·
Development of a Graphical User Interface for work-preparation and job-processing

·
Implementation of easy and reliable calibration procedures for the calibration of laser and sensor tool center point

·
Integration of the software framework on multiple robot and sensor platforms
Gönderen zaman:

Hiç yorum yok:

Yorum Gönder

Kaydol: Kayıt Yorumları (Atom)
Cascode stage
or “collector follower”
Jean-Paul Brodier
All microprocessors from the 8051 family have inputs and outputs that are ‘quasi-bidirectional’. This means that when power is first applied, the ports behave as inputs with a logic high level and a weak pull-up.,

Glitch
a relay or some other load such as When driving an optocou-pler or LED, there is a problem
at power on: the NPN transistor in the common emitter connec­tion (Figure 1) causes an unde­sirable excitation of the load from the moment power is applied until the microprocessor has had the chance to turn the output low. In addition, logic high outputs are seldom able to deliver enough current to drive the transistor into saturation because they have been designed to be active low.
Figure 1. An NPN transistor drives a load.
To solve both of these problems in one hit, we have to make the active level logic low. This can be done in three different ways: use an emitter follower as a buffer stage (Figure 2a), an inverter in a common emit­ter circuit (Figure 2b) or an inverter/open collector circuit (Figure 2c). The disadvantage of solution 2a is the fact that the voltage to the load is reduced. In the case of a relay with a 5-V coil there is the risk that the resulting voltage is too low. The disadvantage of examples 2b and 2c is that they require more parts.
Collector follower
That leaves the open collector buffer in the form of an IC type 7404. This solution, however, also has a few disadvantages. You do not always need all of the 6 buffers in one IC. Also, the SMD version can only handle 12 V. This is too low and dan­gerous if we happen to supply the load from an unregulated voltage.
The solution presented here com­bines in one transistor the advan­tages of the emitter follower (inactive when power is first applied) and open collector (higher power supply voltage, lower current). This circuit has been known since the valve era by the name cascode (drive via the cathode). The goal was to reduce the Miller-effect of the internal (parasitic) capacitances. Not having the option of reduc­ing the capacitance between the internal electrodes, a lower volt­age was used instead. The cas-code circuit is often used in pow­erful transmitters (tens of kW) to minimise the Miller-effect. This circuit was also used to limit tran­sistor conduction and to keep the dissipation within bounds, which increased the life of bipolar tran­sistors. This was in the IGBT and VMOS era.
The transistor conducts only when the output from the micro­processor is low (refer Fig­ure 3). The base current is lim­ited by resistor R. This current is determined by the current flow­ing through the load. When the power is switched on, both the base and emitter see the same potential, VCC, so the transistor remains blocked. One thing we have to keep in mind: we may not exceed the current rating of the microprocessor output because it has to cope with all the current flowing in the emitter of the transistor.
In the case of the quite common 80C51, this maximum current is typically 3.2 mA (two LS TTL loads). This is sufficient to drive an LED without overloading the 5-V regulator, or for driving a PNP power stage at the high side (Figure 3b). The parallel Philips PCF88574 I2C interfaces can handle 25 mA. For the Atmel AT89Cx051 as well as for the Philips P89LPC9xx the limit is 20 mA. For the latter type the cascode circuit or ‘collector fol­lower’ is even more interesting when the outputs are configured as open-drain because the nom­inal voltage is only 3.6 V. In all cases we have to make sure that the maximum dissipation of the
Figure 3. Cascode driver stage with discrete transistor.
package is not exceeded. 24 V is sufficient to energise its are determined by the power
Should this be the case, then the half Watt relay coil, which in PNP (or VMOS) transistor.
number of open collectors turn can drive a load of 16 A at The cascode transistor can be a required will probably justify 230V. ‘digital’ type with integrated resorting to a 7404. For loads driven from the positive base and emitter resistors.A current of around 20 mA at side, the voltage and current lim-


pot as interrupt generator

In battery-powered, microcon­troller driven circuits, as well as with microcontrollers operating in cars, it is desirable to switch the micro into power-down mode once a task has been completed. An interrupt request is then required to wake up the micro. This circuit allows an interrupt to be generated in a simple way using a common potentiometer. In the example circuit, the pot may also copy its spindle position to the ADC. This enables the pot to be used for continuously variable settings (like volume) as well for getting the micro out of its power-down mode.
IC1A is configured as a differen­tiator with R3 preventing oscillation by keeping the gain down to 10 times. Because the opamp oper­ates off a single-rail supply voltage, an 18k/10k potential divider (R1/R2) is able to create a virtual ground level at +1.75 V. This can be done because the LM358 can handle input levels of up to 3.5 V when supplied at 5.0 volts. IC1A supplies a brief High pulse at a falling input voltage, and a similar Low pulse when the input voltage rises. In order to get a High pulse when the potentiome­ter spindle is turned cw or ccw, IC1B is set up as an inverter. Next, each opamp output drives the base of a BC547 transistor. The 5 V-to-0 V transitions at both collector outputs are shaped and combined into a usable interrupt pulse by three NOR gates IC2A, IC2B and IC2C.If the potentiometer spindle is turned very slowly, it is possible that the circuit does not respond
That is why an LED has been added that lights briefly when a pulse is generated. Finally, a tip: a 100-pF capacitor may be connected in parallel with R5 for additional suppres­sion of self-oscillation.

elektor time standart


Elektor Time standard (1988)
Jan Buiting

The Elektor Time Standard and associated Slave Unit were spin-offs of another hugely successful project, the DCF77 Receiver / Locked Frequency Standard. The receiver was published in the January 1988 issue, the Time Standard and Slave display in the next two issues. All units were housed in then very fash­ionable and (expensive!) Ver-obox two-part ABS enclosures which had also been used for a number of Elektor test instrument designs published between 1984 and 1987. The Time Standard box was designed to process seconds pulses received from the VLF (77.5 kHz) DCF77 time standard transmitter in Mainflingen, Ger­many, and display time (with atomic accuracy) and date on an LC display. The circuit was based on then extremely popular 8052AH-BASIC microcontroller from Intel, a device, we can safely claim, that made it to fame & glory thanks to Elektor Electron-
ics. The 40-way DIL chip con­tained a BASIC interpreter capa­ble of executing ‘tokenised’ code from an external EPROM. This, we were told by our resident designer Peter Theunissen, made writing the DCF77 time signal decoding routines ‘a doddle’ using his specially adapted BASIC computer and interpreter. For example, when concerns were raised (by myself) that not all of Europe was in the time zone served by DCF77 (i.e., CET or GMT+1h), a menu option was quickly added to allow users to select between UTC and GMT+1h. As a relative novelty, a ready-made self-adhesive front panel foil with built-in membrane keys was designed into the proj­ect. This expensive item had been produced specially for Elektor. However, when the article went into print (using a rather glum page layout and black & white print), there were yet other con­cerns regarding the range of the DCF77 transmitter. This is offi­cially claimed as “approximately 1,000 km by groundwave propa-

gation”. A quick use of a com­pass and a map of Europe sug­gested that the signal would only cover the south-eastern part of the UK, possibly including Greater London. For a couple of months we waited with baited breath for readers’ responses, only to receive two enthusiastic reception reports, one from the East coast of Ireland and another from Riyadh, Saudi Arabia! The latter report came from a reader work­ing at a chemical laboratory. I remember he wrote that DCF77 could be received for a few min­utes a day only, synchronising the clock, usually around nightfall despite heavy ‘static’. A huge wire antenna was used (nothing like the 1-inch ferrite rods we used in our lab, which is less than 100 km away from Mainflingen). Although the BASIC program list-
ing for the Time Standard was freely distributed to interested readers (on paper, in an enve­lope, by snail mail!), only very advanced readers were able to compile the program into tokenised code and burn it into an EPROM. Most other readers had to rely on a ready-pro­grammed 27C64 supplied through our Readers Services. Apart from displaying time and date at atomic accuracy, the Time Standard was also capable of outputting time/date information

in the form of ASCII character strings for other (intelligent) equip­ment to use, for example, a timer or switching clock. Although sales figures of the PCB and EPROM were in the hundreds, I never heard from anyone actually hav­ing enjoyed the wonders of the ASCII output so extensively described in the article. The Slave unit published in March 1988 was connected to the Time Standard via screened (micro­phone) cable, the idea being that one or more Slave units could be installed on walls in rooms at some distance from the main clock unit. Central timekeeping deluxe for offices, labs, schools and workshops, but at what an expense and design effort! Not too many PCBs were sold for this extension of the Time Standard.