Depalletising robot

M

Thread Starter

Michael Griffin

I am studying automating a process which may include removing parts from a cardboard carton and either placing them on a transfer nest (for further handling by another system) or placing them in the final location directly (if this is feasible). I intend to have someone else (a robot integrator) perform the system design and installation (it sounds like fun, but I don't have the time). I have not personally been involved in a project exactly like this one before, and I am looking for some suggestions which may help me evaluate ideas. The sort of information I am looking for is: A) I believe there are special material handing robots which are intended for this application, but what are their limitations (speed, accuracy, repeatability, etc.)? The working volume is larger than the robots that I am used to are capable of. B) I believe the scenario I am envisioning (see below) would seem to be fairly simple, but there may be problems I am overlooking. Is there anything I am considering which is likely to cause trouble? C) To avoid vision guidance, I believe we could have the operator push each box to a particular position (e.g. up against a set of stops). However, the system must accomodate different size boxes for different models (in different batches). Are there any *simple* but *reliable* methods of ensuring the operator has positioned the box correctly? I would rather not use rams to push the box into position. A set of sensors (e.g. ultrasonic) to detect each box is in position may work. A set of manually operated locking bars to hold the box in position (they won't close if the box is out of position) is another possibility. The locking bars would have to be adjustable for different models of dunnage though. D) Is some sort of collision sensing advisable to avoid crashing too hard into an unexpected obstruction (i.e. some foreign material was left in the cell)? For example, someone may forget to remove the lid from a box before loading it into the enclosure. E) Is there any particular brand of robot which is especially widely used for this type of application? What robots are used for this application? How fast are they? What are the typical costs? I don't need to point out that I want it to be cheap, simple, reliable, minimal maintenance, run for years needing only occasional lubricating, etc. The robot controller should have no moving parts (no fans, hard drives, etc.), and should be capable of withstanding a typical factory floor enviroment. The system would have the following general requirements: 1) The parts to be handled are plastic and are approximately 300mm to 600mm across, and no more than 25mm thick. Each part weighs no more than a few hundred grams. The parts are quite rigid, and all models could be picked up by a common suction gripper (or some such similar means) at the centre of the part. 2) The dimensions vary according to the part number (within the above mentioned range), but only one model needs to be handled during a batch (i.e. the parts are not mixed). 3) The parts are oriented horizontally in piles (i.e. stacked one atop the other) in the cardboard cartons. Each 'pile' is separated from the rest by cardboard dividers. There are different numbers of columns of parts in each box for different part numbers, but each part number has a specific known way it is packed. The robot would have to reach up to one metre down into the box (with the dividers in place) to retrieve parts. Some sort of sensor will likely be necessary for the robot to know the correct height to unload each part. 4) The boxes range in approximate size from approximately 0.5m x 1m x 1m high, to 1.5m x 1.5m x 1m high, with the larger ones being more common. The smaller containers could quite likely be eliminated, as they are intended for easier manual handling. 5) There are two scenarios. In one scenario, the system must unload one part every 20 seconds. In the second scenario, the system must unload two parts in succession for a total of two parts in 20 seconds. 6) The application may require the robot to drop the part onto a transfer nest to allow another mechanism to operate on it. In this scenario, an accuracy of plus or minus several millimetres should be good enough. 7) If accuracy (or perhaps more appropriately, repeatability) of roughly a few 10ths of a millimeter is possible for the entire system (not just the robot), then it may be feasible to avoid the transfer nest, and allow the robot to place the part in the final location directly. The robot would likely need to accomodate vision guidance for this to work. I am envisioning having a guarding enclosure containing a gantry robot, with interlocked access doors allowing an operator to push full cartons into the enclosure, and remove empty ones. The robot would be disabled while the doors are open. There is the possiblity of having a partitioned enclosure, with the robot allowed to work in one side, while the other side is being loaded. The working enclosure would probably be a metre or so deep, and several metres long, so that several boxes could be worked on in succession before needing attention. ********************** Michael Griffin London, Ont. Canada [email protected] **********************
 
W
Michael, We have an application here at our plant which does more or less what you describe. Some points follow: 1. With good programming the handling of different parts or boxes should be no problem. you could have push button selectors which the operator would select upon loading a box. Box types could be automatically sensed quite easily with optical sensors. Likewise a signal can be generated when the robot arm is working on the last layer. This can be used to warn the operator that intervention will soon be required. 2. Most modern robots will meet the accuracy and cycle times you need. You're application as described is not complex. 3. Many different types of collision detectors are available. This is not a problem. 4. There are several robots available that are suitable for this type of application. Most of them are built so that minimal maintenance is required. They are also robust enough to withstand most factory floor environments. Controllers for a robot of this size would probably require a cooling fan. Programming can be done through a portable PC arrangement or laptop. This means that you only need a hard drive during commissioning or while making program changes. Clearly costs are dependent on whether you require an Escort or a Lexus but you should be able to source a suitable arm for $15K US upwards. Your problem may lie in getting an arm with suitable reach at these prices. 5. A vacuum pick up arm would probably handle this application although many types of end effector or grippers can be developed at low cost. 6. You may have problems in getting down into the bottom layers of the boxes. Unless your dividers are rigid they may foul on the parts as they are lifted up through the box. You should consider different packaging to simplify the task. We got our parts delivered in cheap, clear plastic trays or dividers. These place each of the parts in a position on a matrix for each layer. We unload the trays from the box and place them into a simple elevator which we manufactured in-house. This allows us to elevate each layer to the same pick up position. The elevator is built close to the size of the trays so that when the door is closed it pushes them into the correct position. The robot knows when a layer is finished and lifts the tray before ejecting it. 7. The accuracy you require should not be a problem. With a good nest design you should be able to place directly. Feel free to contact me direct if you have any other specific questions.
 
J

Joe Jansen/ENGR/HQ/KEMET/US

<snip> < The sort of information I am looking for is: < A) I believe there are special material handing robots which are <intended for this application, but what are their limitations (speed, <accuracy, repeatability, etc.)? The working volume is larger than the <robots that I am used to are capable of. Given the requirements below, I would recommend looking at Adept. (www.adept.com). I have also worked with FANUC robots before. Both are quite good, although the design of the Adept systems may make reaching into the box easier. < B) I believe the scenario I am envisioning (see below) would seem <to be fairly simple, but there may be problems I am overlooking. Is there <anything I am considering which is likely to cause trouble? < C) To avoid vision guidance, I believe we could have the operator <push each box to a particular position (e.g. up against a set of stops). <However, the system must accomodate different size boxes for different <models (in different batches). Are there any *simple* but *reliable* <methods of ensuring the operator has positioned the box correctly? I would <rather not use rams to push the box into position. A set of sensors (e.g. <ultrasonic) to detect each box is in position may work. A set of manually <operated locking bars to hold the box in position (they won't close if the <box is out of position) is another possibility. The locking bars would have <to be adjustable for different models of dunnage though. Would you consider placing the boxes onto a conveyor of some sort, and letting them be squared up with adjustable guide rails, and an end stop that it would just run up against? This is something that should be easily solved by the integrator. Just tell them ahead of time what you are and are not willing to have the operators do / pay for. < D) Is some sort of collision sensing advisable to avoid crashing <too hard into an unexpected obstruction (i.e. some foreign material was <left in the cell)? For example, someone may forget to remove the lid from a <box before loading it into the enclosure. Yes. A device called a "Wrist Watch" is an example. Basically, there is a chamber that is pressurized by using an air pressure regulator. If the "End Of Arm Tooling" (EOAT) encounters a force that is greater than the pressure setting, it will "trip" the device, releasing the air pressure which allows the EOAT to swing semi-freely away from the collision. This is also tied into an Emergency Stop on the robot controller. The wrist watch is detented to insure alignment / repeatability when resetting the collision. < E) Is there any particular brand of robot which is especially <widely used for this type of application? What robots are used for this <application? How fast are they? What are the typical costs? I don't need to <point out that I want it to be cheap, simple, reliable, minimal <maintenance, run for years needing only occasional lubricating, etc. The <robot controller should have no moving parts (no fans, hard drives, etc.), <and should be capable of withstanding a typical factory floor enviroment. Adept and Fanuc are both good choices for this. Both are extremely capable of repeatability down to less than +/- .01mm. Fanuc is capable of getting within a few thousandths of a mm repeatability, and Adept is used for high speed circuit board assembly. Both are built for industrial use. < The system would have the following general requirements: < 1) The parts to be handled are plastic and are approximately 300mm <to 600mm across, and no more than 25mm thick. Each part weighs no more than <a few hundred grams. The parts are quite rigid, and all models could be <picked up by a common suction gripper (or some such similar means) at the <centre of the part. < 2) The dimensions vary according to the part number (within the <above mentioned range), but only one model needs to be handled during a <batch (i.e. the parts are not mixed). < 3) The parts are oriented horizontally in piles (i.e. stacked one <atop the other) in the cardboard cartons. Each 'pile' is separated from the <rest by cardboard dividers. There are different numbers of colums of parts <in each box for different part numbers, but each part number has a specific <known way it is packed. The robot would have to reach up to one metre down <into the box (with the dividers in place) to retrieve parts. Some sort of <sensor will likely be necessary for the robot to know the correct height to <unload each part. < 4) The boxes range in approximate size from approximately 0.5m x 1m < x 1m high, to 1.5m x 1.5m x 1m high, with the larger ones being more common. <The smaller containers could quite likely be eliminated, as they are <intended for easier manual handling. < 5) There are two scenarios. In one scenario, the system must unload <one part every 20 seconds. In the second scenario, the system must unload <two parts in succession for a total of two parts in 20 seconds. Absolutely effortless for either system. < 6) The application may require the robot to drop the part onto a <transfer nest to allow another mechanism to operate on it. In this <scenario, an accuracy of plus or minus several millimetres should be good <enough. < 7) If accuracy (or perhaps more appropriately, repeatability) of <roughly a few 10ths of a millimeter is possible for the entire system (not <just the robot), then it may be feasible to avoid the transfer nest, and <allow the robot to place the part in the final location directly. The robot <would likely need to accomodate vision guidance for this to work. Most likely. Both systems can use an integrated vision guidance system. This would probably need to be set up where the robot would pick the part, present it to the vision station to check the orientation of the part on the EOAT, and then drop it into your tooling. Repeatability that you mention is well within the robot's capability. (as mentioned above, +/- a few thousandths mm) < I am envisioning having a guarding enclosure containing a gantry <robot, with interlocked access doors allowing an operator to push full <cartons into the enclosure, and remove empty ones. The robot would be <disabled while the doors are open. There is the possiblity of having a <partitioned enclosure, with the robot allowed to work in one side, while <the other side is being loaded. < The working enclosure would probably be a metre or so deep, and <several metres long, so that several boxes could be worked on in succession <before needing attention. A gantry is probably more than you need. I would suggest an articulated arm. The movement can be coordinated such that it can reach down into the box. Check the websites for local integrators. For Fanuc, I would say use an A-510 in an inverted mount position ( or whatever is equivalent today. It has been about 4 years since I left the integration business). As far as guarding, you seem to have the right idea. Operator entrance can be protected using light curtains, gates, etc. Or, as you mention, have the operators load / unload a conveyor on one side of a fence, and the conveyors can carry them through a hole in the fence. A light curtain would guard the hole from people sneaking through, and the system would simply disarm the light curtain as a box came through. Any other questions, feel free to ask. I would definitely recommend submitting your idea to a couple of integrators for quotations. --Joe Jansen
 
M

Michael Griffin

At 15:12 16/01/01 -0500, Joe Jansen wrote: <clip> >D) Is some sort of collision sensing advisable to avoid crashing too hard <clip> >Yes. A device called a "Wrist Watch" is an example. Basically, there is a >chamber that is pressurized by using an air pressure regulator. If the >"End Of Arm Tooling" (EOAT) encounters a force that is greater than the >pressure setting, it will "trip" the device, releasing the air pressure >which allows the EOAT to swing semi-freely away from the collision. <clip> I have seen these devices advertised, but have not used one. Have you used one, and is there any particular brand with which you are particularly happy? <clip> >Given the requirements below, I would recommend looking at Adept. >(www.adept.com). <clip> We have some Adept Modules, and these have been very reliable in our applications. They seem to be nice and simple also. These are actually a third party device which Adept sells under their own label. These however, don't have enough axis travel for what we want to do, and I don't think the controller is sophisticated enough for good vision integration. We also have an Adept 1. I've been less than thrilled with this machine. The robot hardware itself has been very good, but the controller is very complex (OS and various development software) and we've suffered from hard drive problems several times as well. I'm sure this robot is good where its capabilities are required, but it was overkill in the dispensing application it was used in. Some of the complexity problems though may be due to the original integrator (a very large and reputable company). When I tried contacting various third parties about conducting a minor modification a couple of years after installation, several of them hemmed and hawed about it, and one simply outright refused to touch it once they found out who the original integrator was. I had already looked at the software myself and decided I was better off leaving it alone and let someone else deal with the headache. ********************** Michael Griffin London, Ont. Canada [email protected] **********************
 
D

Davis Gentry

While we do not offer a fully integrated system, our controllers are often used in conjunction with robot hardware from various manufacturers to do sophisticated coordinated motion. We also have systems which have been integrated with machine vision. Any questions on the app give me a yell. Davis Gentry Applications Engineer Delta Tau Data Systems 804.795.4288
 
G

gerald beaudoin

We use a depal for unloading bulk glass jars and cans and feeding them onto a line. Don't know what your application is but the folks at Priority One Packaging are pretty good at that sort of stuff. They have a web page at "priorityonepackaging.com". Gerald Beaudoin [email protected]
 
B

Brandon Ellis

Michael: I have read your posting, along with replies you have received. I must say that, while an articulated arm robot (e.g. Adept, Fanuc, etc.) would perform the task at hand, it is an expensive way to go. Also, when it comes to speed, accuracy and payload, the cold hard mechanical truth is that articulating are and SCARA type robots do not perform as well as a Cartesian robot (i.e. X,Y,Z combination). Many people neglect to think of Cartesian robots as "real" robots. A Cartesian system can provide high speed, high accuracy / repeatability, rigidity (i.e. quick settling times), and high payload capability. Applications range from small, high speed pick and place to 35 ft X 60 ft gantry style Cartesians doing gel coating on boats. Additionally, these same systems may be used for cutting, dispensing, inspection, etc. My company, while we are a manufacturer of motion control equipment, also has a consulting group which has had years of experience designing and implementing Cartesian systems. You may contact me for more information. Thanks, Brandon S. Ellis Sales Manager Robotic Control Group 700 S. Illinois Ave. Suite A104 Oak Ridge, TN 37830 Tel: (865) 425-0301, Ext. 160 Fax: (865) 425-0268 http://www.roboticcontrol.com
 
J

Joe Jansen/ENGR/HQ/KEMET/US

Michael Griffin wrote: <clip> > I have seen these devices advertised, but have not used >one. Have you used one, and is there any particular brand with >which you are particularly happy? Someone else beat me to this one: QuickStop. They are EXCELLENT! Never had a problem with them. <clip> >>Given the requirements below, I would recommend looking at Adept. >>(www.adept.com). <clip> > We have some Adept Modules, and these have been very reliable >in our applications. They seem to be nice and simple also. These are >actually a third party device which Adept sells under their own label. >These however, don't have enough axis travel for what we want to do, >and I don't think the controller is sophisticated enough for good >vision integration. > > We also have an Adept 1. I've been less than thrilled with this >machine. The robot hardware itself has been very good, but the >controller is very complex (OS and various development software) and >we've suffered from hard drive problems several times as well. I'm >sure this robot is good where its capabilities are required, but it was >overkill in the dispensing application it was used in. > Some of the complexity problems though may be due to the original >integrator (a very large and reputable company). When I tried >contacting various third parties about conducting a minor modification >a couple of years after installation, several of them hemmed and hawed >about it, and one simply outright refused to touch it once they found >out who the original integrator was. I had already looked at the >software myself and decided I was better off leaving it alone and let >someone else deal with the headache. You can look at the FANUC robots. They are a bit spendy, though. The good side of FANUC is that if it is a production machine, they guarantee a 24hour response time to a down system. This means a tech at your plant with a part. This is a good thing for extremely high speed lines. For your application, it sounds as though you could probably make do with an operator while the thing gets fixed. AFAIK, most robots should not have any difficulty with your accuracy/repeatability issues. The bottom line, then, is price, vision integration, and of course, the ability to reach into the box. I am afraid that Adept and FANUC are all that I have worked with. FANUC the most. I know that they can do it, and will do it well. I also know that FANUC is like Allen Bradley when it comes to market position and pricing. Hope that helps. (?) --Joe Jansen
 
A

Aravind Durai

Hello Michael: I am a mechanical engineer with ABB robotics. I have been in robotics for nearly 10 years. I am well versed in the kind of application that you have at hand. I will be glad to help you out so feel free to contact me. Thanks. [email protected]
 
Top