Industrial machine software

martin1016

I have been working on a machine of our own design and need software written for it. GUI for interface purposes with analytical tools and for control of teh motion control system. System will be 4 axes with image analysis of the captured and processed data. There would be a need for some image enhancement tools such as sampling the live image to produce the best quality image to perform the analysis. Output would be formating info into reports etc.

Machine would be stand alone but would need capabilty to connect to customers network for obvious reasons. I am considering a Linux platform and have been trying to get my head around Ubuntu to see how an idiot like me can cope with it, Windows would be the obvious choice but I am not a fan of Vista, XP would be fine but wonder about moving to Windows 7 from XP. The unit will have a standard PC, workstation spec more than likely with a graphics card, nothing like that for extreme gaming but something to get a crisp image displayed on it.

So I would be interested to hear any of your comments, I have not posted it in the sticky thread yet as there are some other options that I could use such as a modified 3rd party solution. However I am leaning towards having more control ourselves and keeping it local.

Anyway - comments please, language recommendations, OS etc or if someone has an interest let me know we'll see where it goes. Sorry its not all sexy web stuff, just boring engineering type application.

BTW, I for obvious reasons I may not be able to answer all your questions here but feel free to ask.

fergalr

If you are still looking for feedback on this, I would suggest that you look at your initial post again, and try find a way of making clearer what it is you are after.
I understand that if you are working on a proprietary design you want to keep things somewhat under wraps, but, speaking just for myself, I found it very hard to extract enough of an idea of what it is you are building from your post, to attempt to give you any meaningful advice on your questions such as choice of operating system, language etc.

professore

Where are you based?

copacetic

Maybe have a look at labview as an easy way to develop machine control, analysis and gui at one time. Sounds like you are doing something it was pretty much made for.

stylers

There is an open source linux project called EMC which is for control of CNC machinery, perhaps it may suit ?. There is even an ubuntu distro with it already installed to try out - it tweaks the kernel slightly to get exact realtime control..

pauldiv

Look at National Instruments as they make all sorts of specialised cards for data aqusition and control applications. Labview is a great programming tool and is used a lot in SCADA applications.
Their website and printed catalogues will have case studies of some pretty impressive solutions.

martin1016

Thanks for the comments guys - I suppose its not that clear to someone on the outside looking in – so I will give it a bash again..

Basically we will place a sample in the machine when the table is in the load / home position. This could be anything from a medical syringe in packaging to your ipod, mobile phone to toaster. Components could be possibly powered and cycled while under inspection. From here the operator will be able to manually drive to locations on the sample to inspect. I want the option to be able to program the system to move to specific location to carry out the inspection at which point the operator should note the fault, tag it and output it to a file or proceed to the next inspection point.

At all times there will be a live image feed of the sample on the monitor, this image will not be suitable for processing as the nature of the inspection will require a sampling of the image to produce a crisper image and improve detail. Different rates of sampling will be available to sharpen the image over a varying number of frames – more frames sampled and averaged the better the image. From this image it should be possible to apply specific tools - measurement, greyscale calculations, tag faults etc and out put to a report.

There are essentially 3 major components in the software, control and programming of the motion control systems 3 axes minimum with addition of another 2 if required. Motion should be via buttons / sliders in the software and also via a hardware joystick for ease of use. Labview could do this and the GUI could interface with labview in the background I suppose. Control of the hardware in the imaging system, ie power intensity, zoom, contrast etc. Finally there would be the analysis of the captured image, measurement of features, tag issue or output images to reports or files, would be nice to give a report wizard or something similar.

The software needs to look solid and slick, one comprehensive interface that should be easy to use. Plan is, assuming we get what we need at our target prices, to market internationally, hence systems could be installed anywhere in the world. As I said windows would be the easy choice but with virus and security issues I am not sure it is the right one. The system would be stand alone and may or may not be networked. Files would be removed via network or USB and hence the concern about security as if not online maintaining an anti-virus package would be difficult. Linux would eliminate much of the security concern, lighter package and could lead to expansion of an embedded system or thin client setup if hooked to a server but that’s getting ahead of myself.

Located in Naas but I live on de Northside so Dublin is local. Hopefully that explains it a little better. Our background is installation, service and configuration of high end automation but this is a new step for us in developing our own system so it is a steep learning curve at present but we’re figuring it out as we go.

Our stuff won't be as big as this so no messing at lunchtime

DublinDilbert

martin1016 wrote: »

At all times there will be a live image feed of the sample on the monitor, this image will not be suitable for processing as the nature of the inspection will require a sampling of the image to produce a crisper image and improve detail. Different rates of sampling will be available to sharpen the image over a varying number of frames – more frames sampled and averaged the better the image. From this image it should be possible to apply specific tools - measurement, greyscale calculations, tag faults etc and out put to a report.

All the motion control stuff is fairly straight forward, and there's lots of ways to do it...

I'm not sure what you mean by "nature of the inspection will require a sampling of the image to produce a crisper image and improve detail"? Cause you can do sub-pixel estimation, but its not going to improve the detail as such.

Are you planning on using a high resolution camera? (which could be quite expensive? )

A rule of thumb in vision is that the camera should operating a resolution 10x greater than the smallest defect your trying to identify. This can be archived using a high enough resolution camera, then with some optical zoom also.

Really high resolutions systems would use a "line-scan" camera, and move the object underneath it on the x-y table, not unlike how your computer scanner works. Even a cheap computer scanner will operate at 600dpi...

Also in this type of application control of the lighting is key, and it will vary depending on the type of defect your trying to pick up....

martin1016

The method of inspection is what I do not want to expand on in the forum. Its not dramatically new but I would rather keep it internal at present.

The system will use an industrial usb camera, (could be CCD or CMOS yet, weighing up the pros and cons of each at the the moment) to capture the image which will not be created using regular light. This is why I need to sample and average frames. Image intensity can be adjusted but in a different way than you would with a standard camera.

I am hoping to achieve detail recognition less that 10 microns, image magnification will be in teh order of 2000x or more depending on the final geometric configuration we decide on.

DublinDilbert

martin1016 wrote: »

The method of inspection is what I do not want to expand on in the forum. Its not dramatically new but I would rather keep it internal at present.

The system will use an industrial usb camera, (could be CCD or CMOS yet, weighing up the pros and cons of each at the the moment) to capture the image which will not be created using regular light. This is why I need to sample and average frames. Image intensity can be adjusted but in a different way than you would with a standard camera.

I am hoping to achieve detail recognition less that 10 microns, image magnification will be in teh order of 2000x or more depending on the final geometric configuration we decide on.

I presume you mean a "Fire-wire" camera rather than an industrial usb camera. You should also be able to get "fire wire" servo drivers, so both can be controlled from the PC. If your not using regular light, you may need to synchronise the capture to the light source...

You seem to have the vision side of things sorted, and coming from an automation background the motion control shouldn't be a problem. I guess your looking for a PC application that will tie it all together...

I've done a lot of PC development and have worked on vision projects, PM me if you need more information / advice.

Cantab.

Why re-invent the wheel?

4-axis control gets very complicated. The tooling/wear-simulations/machine learning aspects is a massive black hole of information.

Leave it to the professionals.

martin1016

Cantab. wrote: »

Why re-invent the wheel?

4-axis control gets very complicated. The tooling/wear-simulations/machine learning aspects is a massive black hole of information.

Leave it to the professionals.

Future plans are for 6