1

Control Systems Overview for Automated Robotic Welding Machines

D. A. Wykstra, Senior, Calvin College Engineering

Abstract Control Systems are overviewed and discussed for two basic divisions of robotic welding machines. The small, single operation machines are discussed and contrasted with large, multi-cell robotic machines. The control methods used for both types of machines, along with the different types of sensors used to provide system feedback are discussed. The basics behind the control theory, as well as the actual controlling entities are explained. Finally, applications and examples of both types of machines and their basic operations are given and explained. Index Terms Industrial Control, Manipulators, Multisensor Systems, Process Control, Programmable Control, Relay Control Systems, Robot Sensing Systems, Robots, Welding.

III. SMALL WELDING STATIONS These stations are generally used to assemble smaller subparts on an assembly line or in individual stations. These are usually operated by a single worker, and contain only one or two weld guns, as they assemble smaller parts than their larger counterparts. They are also limited in their motion range to a single axis, usually the y-axis. The relative simplicity of these stations means that the operator needs little training, and there is enough safety control in place that it is difficult for one to injure themselves on them if they are functioning correctly. Many require that the operator simply load the required parts into the machine and activate the cycle. At that point, operation is controlled by the PLC and the welding cycle commences. Once it is finished, the station waits for another activation signal before repeating the process. A. Control Overview The control system for these stations is relatively simple, just as the operation is. The system generally consists of a PLC, a set of sensors, pneumatic cylinders, weld gun(s), and a water system, usually to cool the welding guns. These components function together to create a control system capable of fully automated operation cycles.

I. INTRODUCTION

n this paper, I will discuss the control systems of some different types of robotically controlled welding machines. These have found uses in all areas of manufacturing, the biggest market for which is automotive. Not only are they used in direct automotive assembly, but they are also used by most suppliers for the automobile makers, where they are used to assemble parts and fixtures which are then sent to assembly plants. Since robotic welders and welding machines are used in so many different types of applications and areas, there are many different directions a paper on this topic can go. This paper will be directed mostly at the applications these machines find within the manufacturing community, particularly in the automotive supply and assembly world. II. BASIC THEORY The theory behind robotic welders, be they simple, single weld stations or massive multi-robot fixture assembly stations, is relatively simple. Behind each of these machines is the need for the owner to be able to assemble and weld a part or group of parts together more accurately, efficiently, and consistently than human workers are capable of doing. This is accomplished by using a combination of sensor arrays, weld guns, pneumatic cylinders and water cooling systems. These are all usually controlled by a PLC (programmable logic controller), while a rare few are controlled by more powerful systems, although those will not be discussed here. This PLC is custom-programmed and coordinates the actions of all these various systems.

Fig. 1.Basic control system block diagram for a small welding machine.

B. Control Methods As mentioned previously, these individual components are controlled by a PLC. With this setup, there is a surprising amount of flexibility and customization that can be accomplished. This is due to the fact that the program running in the PLC and controlling the station is custom written by the same people who design the system of components on the station. The program in the PLC runs constantly, but only begins to control anything when it receives an activation signal from the operator of the station.

This signal is typically input by a simple button the operator hits when the parts to be welded are in place. When this button is hit, the system of sensors mentioned earlier is checked by the PLC, as they are placed so that they can sense whether the part is loaded properly and make sure that any previous operations that should have been performed on the part have been. The pneumatic cylinders can have many different functions. Some of them are used to check the position of parts or features of parts, such as holes in the part. More often though, they are used to hold a part in place by serving as clamps, move a part in the midst of a welding cycle so more welds can take place, or most importantly and commonly, to move the welding gun down onto the part and then lift it back off when the weld has finished. Last is a water system used to cool the weld guns and transformers providing power to the guns. All of these components are controlled by the PLC via a series of input and output cards connected to the PLC processor. Signals from the sensors are sent into the input card(s) and processed by the PLC, which then decides, based on the program, which outputs should be activated. This controls what each of the water solenoids, cylinders, and weld guns on the station does, such as activating water flow, extending and retracting cylinder pins, or activating the welding guns. C. I/O Cards/Modules The input and output (I/O) cards or modules (the terminology depends on the manufacturer) are the components that are used as the interfaces between the input components, such as the sensors and switches, the output components, such as the pneumatic cylinders and weld gun firing controls, and the PLC itself. These devices come in a number of configurations and operation modes. Many of the different options available in these devices are the operating voltage, either 24VDC or 120VAC D. Interfaces The user interfaces for these small stations are usually quite limited, as there are few options to be changed. Some of these stations dont even have an interface other than a cycle start and emergency stop button, along with a couple of pilot lights to indicate a status, such as faulted or stopped, or possibly cycle running or clear to enter. Other stations will have small screens with a few buttons that allow different modes of operation to be selected, but based on their limited size and abilities; there are few things to be changed on most of these types of stations. IV. LARGE ROBOTIC WELDERS These systems are by far the more interesting of the two breeds of machines; they are also more expensive, complicated, capable, and may even require multiple workers to operate them, depending on the application. These systems often contain multiple robot arms or cells for their operation, and are capable of far more complex and comprehensive operations than their smaller counterparts. While some of the basics remain the same, such as PLC control and sensor

arrays, they are far more impressive machines, both in design and operation. A. Expanded Capabilities When compared to the smaller machines discussed earlier, these machines cover many more situations and applications, are more difficult to design, and to many people may seem over the top in terms of complexity and abilities. While the systems discussed earlier were limited to a mere one or two weld guns and a single axis of motion, these larger systems are under no such limitations. The only limits on this class are the investments the owner wants to make in time and money. These larger stations also step outside the convention of single machines, as they often employ multiple individual robots and weld stations in their operation. The systems discussed earlier were limited in scope to one or two different welds which would take up to 3 pieces and assemble them into a single part. These stations are capable of welding more than a dozen parts together, requiring only that an operator load the required parts into the various machines at the start of a cycle. Once the parts are loaded and the cycle is started, the system can manipulate the various parts and perform the necessary welds to assemble a finished product without human action during the cycle. They also can have more than 10 different welding guns to weld the various parts together into a single fixture. The smaller machines also typically do not move in more than one axis, requiring simple tack welds to complete the part being assembled. These larger systems expand to using multiple robotic arms to accomplish the desired end. These arms have up to 6 degrees of freedom and can be fitted with grippers, weld guns, or interchangeable tools on their ends. All of this new movement and the new flexibility of the robot arm fittings means that these stations can move, sort, and drop finished parts onto an assembly line without the input of a human operator all in all a very impressive achievement. These new capabilities give these stations incredible new abilities when compared to the smaller, more restrictive stations discussed earlier. B. Controls Overview While the basic concepts behind the controls for the larger systems are similar to the ones behind the small systems, the sheer complexity of the system increases almost exponentially when systems become as large as those being considered here. The set of sensors described in the earlier machines consisted of less than a dozen different sensors, most of which are common proximity switches. In these systems, the sensors can number in the dozens, possibly more, depending on the particular system. The sensor selection is also larger, increasing to include such things as photo eyes, limit switches, pressure switches, and even visual cameras to check the accuracy of the work done by the robots. Although the latter does not work in the fashion described above (a camera cannot communicate directly with a PLC it must work through an advanced user interface), it is still used in the same fashion as the sensor array. The inputs must also include connections to the robot controllers that run each individual robot arm. The outputs from the PLC also function in the same manner as described above, although they too have increased in scope with the complexity of the machine. The outputs

must now include the pneumatic cylinders discussed above, along with motor starters and outputs that allow the PLC to communicate with any robot controllers that run the individual robot arms. C. Control Methods Surprisingly to some, the methods used to control these more complex stations are really no different than those used to control their smaller cousins, albeit more complicated. In addition to the PLC, there may be many other devices such as robot controllers, motor controllers, nut or stud feeders, or other control devices used to run the individual components of the station. Since these larger stations may incorporate devices such as those mentioned here that are located up to 20 feet away from the PLC, many of these systems use something called remote I/O. This is where I/O modules, much like those discussed earlier, are located in their own enclosure at the device and are connected through an adapter and a single line to the PLC. These modules operate identically to the modules that are attached directly to the PLC and are used to interface with the sensors and outputs located at a device. Any sensors and outputs needed to control any remote devices on these large stations are connected into these remote I/O modules and their signals are relayed through a single connection to the PLC. This avoids the myriad inconveniences and potential problems that would be associated with running dozens of individual cables over long distances. D. Interfaces The interfaces of these systems are also of increased complexity, with more possible modes of operation, more options of operator control, and more capabilities. Instead of being simple screens and buttons as discussed above, these interfaces are full-fledged computers contained in a touch screen interface console. In addition to providing a way for the operator to have better control over the station, they can also provide an additional amount of processing power to handle more intensive input devices, such as visual cameras. Interfaces such as these often cost in the range of $10,000 for some of the most expensive models, but if someone has a machine that can use their power and flexibility, they are very much worth the extra cost. They are able to hold much more in terms of program size and memory, and can perform in cycle diagnostics on machines to determine if and when any operational problems are arising, and alert the operator through use of its own screen or even by turning on a strobe light or a buzzer near the operator on the machine. They are also capable of interrupting the normal operation of the machine they are part of when given the proper input by an operator. This can be done of a variety of reasons, but most common is to use this option to perform minor, short time maintenance on a machine. This way the maintenance team does not need to go through and shut down all power to a machine in order to say, replace the welding tips on a weld gun, keeping maintenance time and costs down. Another good feature of these interfaces is that they will, like the interfaces on some of the smaller machines, inform you of the machines status and inform the user of a problem. Additionally, these interfaces are often programmed to give the user or maintenance team the option of telling the

machine how to respond when a problem is found. So if the operator were working at a machine and the interface informed them of a problem with the part, the operator could use the interface to suspend system operation and reach into the machine to remedy the problem and replace the part. The depth of the abilities of these types of interfaces are only limited by the programs of the designer and the flexibility the designer wants to give to the operator, as there are some safety concerns with allowing the operators that much freedom in some situations. E. Operations The operations these systems are capable of far outpace those of simple machines or even human workers. They can also take much longer to complete a cycle or part than the smaller machines would due to their expanded abilities. While many of the smaller machines are only designed for a simple operation, these stations are designed with a more comprehensive assembly in mind, and can thus spend more time completing a cycle than their smaller counterparts. However the finished product is well worth the wait, and is still finished quicker and more accurately than could have been done any other way; at least with current technologies F. Applications The systems discussed in this paper, be they the simple machines or the expansive multi-robot stations, are widely used in most manufacturing environments, where they are capable of assembling and operating on various parts faster, more efficiently and more accurately than humans ever could. This of course raises some questions as to the morality of replacing humans with machines, which will be addressed fully in the final draft. Manufacturing environments have been transformed over the past few decades by the ever advancing tide of automated technology, of which the systems described here are an integral part. They are used in everything from simple bracket assembly through assembling vehicles, and as technology increases, so will the scope of these automated systems abilities. V. SENSOR TYPES SYSTEM INPUTS In this section I will explain the function and operation of some of the common sensors found providing feedback to the PLCs on machines like the ones described in this report. The most common of these are reed switches and proximity sensors (conductive, capacitive, and ultrasonic), and although there are some others, such as flow sensors and pressure sensors, the operation of these are well known to most readers. A. Proximity Sensors Inductive Inductive proximity sensors are designed to operate by generating an electromagnetic field and detecting the eddy current losses generated when ferrous and nonferrous metal target objects enter the field. The sensor consists of a coil on a ferrite core, an oscillator, a trigger-signal level detector and an output circuit. As a metal object advances into the field, eddy currents are induced in the target. The result is a loss of energy and smaller amplitude of oscillation. The detector circuit then recognizes a specific change in amplitude and

generates a signal which will turn the solid-state output ON or OFF. The active face of an inductive proximity switch is the surface where a high-frequency electro-magnetic field emerges. A standard target is a mild steel, 1mm thick, square form with side lengths equal to the diameter of the active face or 3X the nominal switching distance, whichever is greater.

Fig. 2.Internal diagram of an inductive proximity sensor, with all parts labeled

In the absence of a target, the oscillator is inactive. As a target approaches, it raises the capacitance of the probe system. When the capacitance reaches a specified threshold, the oscillator is activated which triggers the output circuit to change between on and off. The capacitance of the probe system is determined by the targets size, dielectric constant and distance from the probe. The larger the size and dielectric constant of a target, the more it increases capacitance. The shorter the distance between target and probe, the more the target increases capacitance. The standard target for capacitive sensors is the same as for inductive proximity sensors. The target is grounded per IEC test standards. However, a target in a typical application does not need to be grounded to achieve reliable sensing. 1) Shielding Shielded capacitive proximity sensors are best suited for sensing low dielectric constant (difficult to sense) materials due to their highly concentrated electrostatic fields. This allows them to detect targets which unshielded sensors cannot. However, this also makes them more susceptible to false triggers due to the accumulation of dirt or moisture on the sensor face. The electrostatic field of an unshielded sensor is less concentrated than that of a shielded model. This makes them well suited for detecting high dielectric constant (easy to sense) materials or for differentiating between materials with high and low constants. For the right target materials, unshielded capacitive proximity sensors have longer sensing distances than shielded versions.

Fig. 3.A metal target approaching an inductive proximity sensor (above) absorbs energy generated by its oscillator. When the target is in close range, the energy drain stops the oscillator and changes the output state.

To determine the sensing distance for materials other than the standard mild steel, a correction factor is used. The composition of the target has a large effect on sensing distance of inductive proximity sensors. If a target constructed from one of the materials listed is used, multiply the nominal sensing distance by the correction factor listed in order to determine the nominal sensing distance for that target. Note that ferrous-selective sensors will not detect brass, aluminum or copper, while nonferrous selective sensors will not detect steel or ferrous-type stainless steels. B. Proximity Sensors Capacitive Capacitive proximity sensors are designed to operate by generating an electrostatic field and detecting changes in this field caused when a target approaches the sensing face. The sensors internal workings consist of a capacitive probe, an oscillator, a signal rectifier, a filter circuit and an output circuit.

Fig. 5.Comparison of construction between shielded and unshielded capacitive proximity sensors

Unshielded models are equipped with a compensation probe which allows the sensor to ignore mist, dust, small amounts of dirt and fine droplets of oil or water accumulating on the sensor. The compensation probe also makes the sensor resistant to variations in ambient humidity. Unshielded models are therefore a better choice for dusty and/or humid environments. Unshielded capacitive sensors are also more suitable than shielded types for use with plastic sensor wells, an accessory designed for liquid level applications. The well is mounted through a hole in a tank and the sensor is slipped into the wells receptacle. The sensor detects the liquid in the tank through the wall of the sensor well. This allows the well to serve both as a plug for the hole and a mount for the sensor. C. Reed Switches A reed switch gets its name from the use of two or three

Fig. 4. Internal diagram of a capacitive proximity sensor, with all parts labeled

thin metal pieces, called reeds, with plated contacts at their tips and spaced a small distance apart. The reeds are typically encapsulated in a sealed glass tube filled with inert gas. A field from a magnet or an electromagnet deflects the reeds, making or breaking switch contact.

will stay closed as the magnet draws nearer and will remain closed until the magnet is out of range. One important quality of the switch is its sensitivity, the amount of magnetic energy necessary to actuate it. Sensitivity is measured in units of Ampere-turns, corresponding to the current in a coil multiplied by the number of turns. Typical pull-in sensitivities for commercial devices are in the 10 to 60 AT range. VI. PLC OUTPUTS After the PLC processor itself, the most necessary parts for any welding station to function properly are the outputs. These outputs are what drive the actual function of the machine, and control how the machine operates. While there may be more diversity in the types of outputs found on a machine, there is one main type that drives the main functions of the machine. A. General Outputs There are many different outputs from the PLC that fit into this category, since these outputs are relatively few in number on any given machine, but are myriad in different type. One of the first and most common outputs that fit into this category is the outputs to control the weld control module. This is a separate control system that is not outlined in this paper that is responsible for taking command outputs from the PLC and controlling the weld guns and their firing sequences individually. Typically there are four control signals coming out of the PLC and going to the weld control, but on the biggest machines, there can be as many as 7 outputs tied to the weld control. Another type of output in this category is a motor starter. This type of output is necessary for machines on which there are large, high current motors that need to be started. These outputs are used because the PLC output cards generally can have a voltage of 120VAC, but at a low current, usually under 1A. At this power, the output card is not able to provide the current needed by a motor when it is starting from standstill, especially if it is under a heavy load. The current drawn would at best short out the PLC card and at worst melt the card and start a fire in the PLC. These motor starters are a certain type of relay which is connected to ground when not active and is able to handle very high currents without having the contacts melt into place under high loads. These motor starters are generally very expensive relays, costing upwards of $500 for some types. Another type of output closely related to the motor starter is the relay. Relays are common in welding machines, although there are only a few in most robotic cells. They are most often used to allow the PLC output card to control a circuit that runs on a different voltage than that native to the output card. It effectively isolates the PLC from this other circuit, allowing the PLC to control operations at a different voltage, as well as providing external power to another circuit, allowing the PLC output card to enable a circuit that requires more current than it can provide, as in the case of the motor starter.

Fig. 5. of the two types of reed switches discussed here

Internal diagram

Two-reed devices have normally open contacts which close when actuated. Three-reed versions have a pair of normally open and normally closed contacts. Operation of the switch causes these parts to change to the opposite state. An applied field makes the reeds magnetic so their ends attract. Removal of the field lets the springy metal reeds return to their original positions. The movement of the magnet relative to the reed switch determines how the switch toggles. Moving the magnet perpendicular to the side of the switch causes one switch closure per pass. Moving the magnet parallel to the switch provides as many as three closures with the maximum magnet travel. Another option is to spin the magnet near the switch. When the pole axis and the switch axis are parallel, the switch closes. When the axes are perpendicular, the switch opens. There are two or more closures with each revolution depending on the number of poles. Another typical scenario for a reed switch employed in position sensing is to either move the magnet near the switch or slide ferrous metal between the switch and the magnet, thereby toggling the contacts. Alternatively, the reed can be biased closed with one magnet sitting just inside the actuation distance. A second magnet can then open the switch by canceling the field of the first. Typical commercial-grade reed switches handle currents in the milliamp range on up to about 1 A of either dc or ac current. However, special designs can get to around 10 A or more by incorporating separate magnetic and contact units, thereby allowing optimization of the contact area. Reed switches frequently get incorporated into sensors and into relays. One of their advantages is a built-in hysteresis with regard to magnet position. For example, a magnet coming within range forces the switch contacts closed. They

B. Solenoids Pneumatic solenoids are by far the most common types of outputs on a welding machine. These are the outputs that are used to control the actual motion of almost all aspect of the machine. They are used by the PLC sending an output signal to the solenoid, which the changes its position, allowing airflow to enter any cylinders attached to that solenoid. This allows the pin of the cylinder to extend, providing motion. After the objective of the motion has been achieved, the solenoids are deactivated, which then exhausts the pressure from the cylinder and retracts the pin. Depending on size, these solenoids can handle over 200 psi and can be connected to over a dozen different cylinders. 1) Operation Examples This is used in many machine operations and is one of the best ways to achieve machine motion. It is used to clamp and unclamp the parts being operated on, where activating the solenoid and extending the pin on a cylinder closes a clamp and holds the part being operated on in place. Deactivating the solenoid releases pressure on the cylinder and retracts the cylinder pin, releasing the clamps hold on the part. The other most common operation performed by these solenoids is movement of the machine weld guns. They are used to descend the weld gun tips onto the part being operated on, after which the guns are fired and the guns are then raised up again. They are also used to move the guns (term: index or slide) from one position to another, allowing multiple welds to be performed on a part by a single gun. This is an operation that sounds simple, but has cost saving ramifications, since it reduces the number of weld guns needed on a machine. VII. CONCLUSION Automated welding machines are complex machines that run on rather basic principles. While the technology to control them advances and changes, the basic control methods have not changed, and will continue to remain the same for the foreseeable future. In this paper, I have outlined the basic operational methods of robotic welding machines, from the smallest, most basic machines to the large, complex, multi-million dollar machines. I have also given a basic explanation of the operations of many of the different components involved in the machines. VIII. MORAL ISSUES One last thing to remember when discussing this is the effect of automation on the people in the workplace. While developing these welding machines is a very interesting thing, one also needs to keep in mind that the extensive automation they are a part of isnt necessarily a good thing. These machines sometimes take the jobs of people, some of which have a hard time finding new work after they are laid off. This can cause problems when done on a large scale. However, on the opposite side is the argument that many of the jobs taken by these machines are ones that people cannot do properly or efficiently anyway, which leads to more accurate work, and in the long term, lower production and consumer costs, making things more affordable to people.

Another thing to consider is that many of the machines take jobs that are very dangerous for people to do directly, which then frees up some workers to pursue a safer and possibly more satisfying job. Yet another thing is that while these machines may take some jobs, they still require people to operate them, maintain them, and manufacture them. Thus, in the long run, there may not be the net loss in jobs that some people argue exists. I am not purporting to be the moral expert on this subject, simply offering a little non-technical food for thought for the readers of this paper. I am not saying one side or the other is right in this debate, as I do not have the expertise to do so. However, this is another important consideration in the ever increasing complexity of the world we now live and work in. IX. REFERENCES
[1]. Anonymous, Switch Tips: Reed Switch Basics Published on MachineDesign.com http://www.machinedesign.com/ASP/viewSelectedArticle.asp? strArticleId=57254&strSite=MDSite&catId=0 [2]. Apex Controls, Inc., Staff and in-house documentation