Slashdot Mirror
Creating a Homebrew Industrial Process Monitor?
pionzypherm asks: "I work at a glass plant for a major beer company. My job entails monitoring the furnaces that melt the glass. I have been working on a project on the side, collecting data from various sources and compiling it into an easily used form for the higher ups. I've finished two of our three furnaces, but one remains. This furnace uses technology from the early nineties. There is no networking, the hardware is completely closed and unavailable for any screen scraping. Two of the items I'm looking to monitor (and would appear to be the easiest starting point) are two valves for a gas and oxygen line which will provide data on a portion of our energy usage. I was thinking of a microcontroller board or something similar tied in to monitor the positions of the valves. I'm unsure where to begin though. What books, microcontroller boards or alternatives would you recommend for someone new to this? What suggestions would you have for such a project, and what pitfalls might I run into?"
You might try to ask this question on some of the robotics yahoo groups. They are filled with people that do this kind of thing for a hobby and spend a great deal of time thinking about such things not to mention that they do their work with small home brew or cheap microprocessor systems.
People that make their own CNC machines know a LOT about monitoring position of things etc. This might be your best bet for initial and longer term answers and help about how to accomplish what you wish to do.
One piece of advice though is think through what you want to ask. When you ask, explain the system in some detail, your thoughts on what might be monitored, how, and what your end goal is with your monitoring. They may have suggestions that go beyond your knowledge scope if you explain more about the system so they can think about the problem with all the requisite information.
Support NYCountryLawyer RIAA vs People
If all you need is monitor the position of a valve, you can homebrew the position sensor yourself using a plastic disc, a magic marker, and an LED. The sensor sends pulses as the valve turns, these get picked up by a small microcontroller board. Get an experiment board that supports RS-232 serial. Then, you write some minimal code on the microcontroller to process the events into some reasonable data format, and send them over serial to a PC. Do your real logging and data processing there, instead of on the micro board.
There are tons of options out there (I havent worked in the field for ~ 10 years). Assuming you can access something to get the readings off you could get a 90-30 PLC to pull the data points. The Cimplicity MMI is a great software package for monitoring, alerting, reporting etc. Try calling GE Fanuc and just tell them what you are trying to do and the can give you plenty of options.
Are they not willing to fund it/hire someone to do it?
You might be making this too complicated.
Let's say you misjudge the tolerances and your fancy little project gets turned into cinders/melts inside the furnace?
Why can't you monitor the volume of gas flow and then calculate the energy? I assume these gases are stored in a tank or something like that. It should be comparatively easy to attatch some type of flow sensor upstream of the furnace.
Spending the money up front for a reliable, standard solution will save a ton later when your homebrew breaks or some other poor bastard has to support it. There's been too many times I've opened a a panel where my first words are "WTF?".
Especially if you're working with oxygen. Get yourself a nice little flow meter (micromotion makes a good one). Then you can get both volume, and (presumably) valve position. If the valve is electrically actuated, you can use the information for a host of alarms.
Either way, if the information is valuable enough to record, its worth the money up front.
I am billdar, and I approve this message.
Put a toad in the furnace and then ignite it. If the toad does not jump out then clearly the furnace is not heating quickly enough. For day to day management use a dragon. They are very good with high temperatures and will be able to help out with your energy bills by giving your furnace the odd blast. One safety tip with the dragon though is that if your name is George its probably best not to let the dragon know that.
Do not try to read the dupe, thats impossible. Instead, only try to realize the truth
What truth?
There is no dupe
If you just want the position of two discrete valves I would suggest finding a used PLC on Ebay. Single box types (like an Allen Bradley SLC 150) that work with discrete IO only can be had for a little bit of nothing. Your biggest concern with costs would be the programming software so I would stick with brands offer it free of charge.
Without knowing what kind of budgetary firgure you are working with to implement this it is hard to get much more specific.
"Sacrifice for the good of The State" - The State
There are people that do industrial automation. There are large companies that make every type of sensor you can imagine to monitor anything you want. There are industrial controllers to control automatic assembly lines. This stuff is all off-the-shelf. It's not "homebrew".
If you're asking this on Slashdot, you're looking in the wrong place.
Do it like a professional would do it. It's a furnace. Stuff can go wrong. Monitoring it with a half-assed homebrew approach is probably worse than simply observing it carefully and worrying about it all the time.
You should probably install an electronic flow control valve with a flow sensor. Use a microcontroller, PLC, or some such thing to monitor your sensor & control your valve. Monitoring it visually (via camera hooked to an embedded computer) is doable, but way harder than what I just described above. You also run into problems with high temperatures getting to your camera. Take a look at some of the solutions Freescale has to offer. You can order a development board, so you can breadboard something together. Should cost around $200. http://www.freescale.com/webapp/sps/site/homepage. jsp?nodeId=02430Z
For a SR Design project for Electrical & Computer engineering, we were tasked to do exactly this with a freescale microcontroller. We needed to precisely monitor the amount of certain gasses put into an oven we use to bake chips.
Hardest part about all of this is getting to know the specific PLC or microcontroller you're using. PLC's are easier (generally) to program than most microcontrollers, but not quite as versatile in number/type of interfaces.
You need to have a very good understanding of Assembly, C, & how sensors work.
You also need to be able to read & understand the mind numbing manuals & technical documents describing the sensors & microcontrollers you choose to use.
-Steve
If you are doing serious industrial work do not attempt to roll your own unless you absolutely have to. The money you save on professional instrumentation will be wasted in downtime and glitches.
There are plenty of professional solutions out there, from gas flow monitors to automated valve systems to integrated industrial process monitoring and control systems. If you are looking to control fuel and oxygen supplies then you need to get stuff that is blastproof so that a stray spark can't set anything off.
Start off with a major supplier like Grainger Industrial Supply. There are tons of components there that might suit your situation. Particularly look at their process monitoring section.
Sapere aude!
One technique you might think about, coming from a "first do no harm" strategy to avoid being blamed if something goes wrong with the equipment, is to try to not add new sensors or valves and etc to the equipment, but try to take advantage of the increases in computing power to simply read it and recognize the situation exactly as a human would.
Check out this project: http://www.eissq.com/DialADC.html It describes software that uses a webcam pointed at an anologue needle gauge to recognize the position of the needle. Why not, as much as possible, set up passive sensors that don't touch or interact with the equipment in any way, feed them into a cheap multi-gigahertz computer, and process everything that way ? If the furance has a big accident, it would be hard to blame your apparatus.
...let's be clear about this: all you're doing is monitoring and gathering data - there is to be no feedback signal from the homebrew rig to control the valves. There's a whole field devoted to control theory, one that is best not trifled with, especially with industrial processes that can potentially cause fatalities.
If you really want meaningful data from those process streams, you're much better off installing calibratable (calibrable?) flowmeters on those lines that cover the performance range of the process fluids you're working with. If you've got the flow, you don't need the valve position, unless it's for a secondary indication to validate the valve's performance (e.g., position vs. Cv vs. measured flowrate). The flowmeters can be hooked up to provide data for remote collection, or more simply, display data for periodic local reading. Here's a mess to start with. Whomever you buy from, you'll need to develop specifications defining the operating range, operating conditions (pressure, temperature, humidity), power requirements, tolerances, calibration frequency, etc.
Science never settles, never rests.
I read that as "Creating a Hebrew Industrial Process Monitor?"
Really wouldn't know where to begin with that one.
sic transit gloria mundi
There are a whole lot of things that go into handling an industrial system. If you're really going to try to do this on your own, you've got a lot of reading ahead of you; the cost of faults in an industrial system is typically prohibitively high. You're going to need a deep familiarity with modern methods. You're going to need to be familiar with direct hardware control, realtime coding (which is harder than most people think,) constant test polling, and all sorts of stuff most programmers never, ever have to deal with.
This is not something you can take lightly. If you're going to do this, you have to get it right, the first time, and that means your test cases and regression tests have to be diamond-hard, your specification has to be absolute, and you have to know your timing will not fail. These are difficult issues, but with the appropriate know-how, this can be done.
Here are some places to start:
Embedded Control:
Realtime:
Integration:
Testing:
StoneCypher is Full of BS
I build CNC and automation equipment, so I can pretty safely say that from what you describe this is a brain-squishingly trivial project. Probably one that can be done over lunch - After you spend five years climbing the learning curve, which is not at all trivial.
I would just ask someone who does do this for a living out for lunch, it'll take them ten minutes. I do this when I need coding done. The price of a few beers to get the occasional patch or script written is a lot more efficient than many years learning coding to do it myself the one time a year the need comes up.
The learning curve on automation hardware is at least as steep as learning Linux, and with crappier documentation. Coding guys usually seem to underestimate the complexity of the physical engineering and design side, and think they are always bright enough to just pick it up and do our jobs. There is more to hardware engineering than the butt-crack guy with a monkeywrench, just like there is more to coding than script kiddies.
In short - unless you want to go into this as a hobby or career change, just treat a hungry engineer to lunch and call it good. Even if you paid him it'd be less than the books you'd need.
Every time someone asks Slashdot a question like this, the hysteria crowd comes out of the woodwork to scream about how it's absolutely impossible for an "amateur" to do it, and you absolutely must hire a "professional," lest something tragic happen, ranging from the ever-popular "you'll lose your job!" to a bucket of dead puppies or something.
Yes, I realize that professionals are sometimes necessary, especially in situations where life is clearly at stake (pilots, medical, law, etc.) I'm sure some jackass will show up to tell me how this is an industrial furnace and that clearly means that a professional is warranted, but we have no idea what the particulars of this situation are. Just stick to the freakin' question, people.
It used to be the case that "professional" implied not only a degree of competence, but also a certain amount of integrity and experience. But that's just not true any more. All it means now is that someone gets a paycheck for doing something. Often it means that they're experts in nothing more than doing something as cheaply as possible.
For what it's worth, I'm personally fond of the Atmel AVR microcontrollers. Many, many people are also fond of Microchip's offerings in the PIC line. But for rapid development, something like the Parallax BASIC Stamp is probably the way to go. They're cheap and easy (like a good woman) and let you focus on the task at hand rather than the bit-level details of how to read sensors, etc.