Archive for May, 2008

Got Control? Part 2

Steve Meyer
May 08th, 2008.

So when is a PC (personal computer) a PAC (programmable automation controller) or a PLC (programmable logic controller)? They are all the same. They all have microprocessors. They are all programmable. They are all available with hardware that is rated for industrial environments, shock, vibration, temperature extremes and various other requirements. They are all connected to real world devices such as sensors, switches, etc. They all execute control based on a programmed binary model of a real world manufacturing process.

They have differing abilities in terms of data storage and throughput. In previous generations of PLC and CNC, the memory for those systems was very expensive and early memories were hand wound wire and discrete magnet memory. With the advent of mass manufactured memory for the PC, the industrial platforms have had to engineer hardware with different components to take advantage of competitive costs. A hard disk drive module was available for some PLC systems where extensive process information  was needed.

In today’s market one vendor offers an industrial computer platform that replaces the PLC, the HMI (human to machine interface) and uses a touch screen interface, using the PC platform to reduce control system costs in many applications by doing several functions with the same hardware. Sounds like a Tablet PC engineered for the plant floor. Nothing unreasonable about that.

Some suppliers suggest that applications which require a lot of information to be acquired and stored at the same time as the control system is executing exceeds the bandwidth of the PLC and requires a different kind of control platform. But I’m not sure how to describe that situation exactly, but I would know it when I see it.

For instance if you are making hard disk drive platters, the exact temperature, pressure, time duration and chemical concentration of each plating step may be required to be tagged to each platter. At millions of platters per year that’s a lot of information.

So what describes the need for data? Sometimes its an external regulatory requirement, like the FDA or Department of Agriculture requiring milk producers to record sterilization of facilities or store batch information on pharmaceutical production. Yeah, that sounds like something you would want to keep tabs on in the event of a problem later on, or even as a means of improving plant productivity and machine utilization.

Is it possible that we can describe the need for different control solutions as a function of the bandwidth of the application? Is the speed of the process and the amount of data needed in given timeframe, a means of defining a different model of control? Some information used for traceability is not needed on a real time basis and could theoretically be streamed in a lower speed thread with a different priority. The control application itself might have a higher priority than the data gathering part of the application.  And it can have varying bandwidth within the range of a machine or process operation.

A multi-ton roll of paper for a printing press has much lower bandwidth than chocolate covered candy bars moving down a packaging line at 35 miles an hour. The candy bar wrapper is running so fast, it takes a high speed camera to diagnose problems because it is far quicker than the human eye can discern.

Within the printing operation are registration functions that are required to insure that colored rollers line up in order to make the images from four (or more) different print sources. This registration function is usually far quicker than a normal controller can respond to and is usually handled with a dedicated module that can discriminate at 50 microsecond time frames.

Not only do the two applications have differing bandwidths, but the bandwidth varies from starting conditions to the bandwidth needed at maximum speed.  So maybe by looking at the controls world with an eye toward control system and data requirements as bandwidth propositions, we can construct better boundaries for applying the control system hardware that is available. Just a thought.

1 Comment

Got Control?

Steve Meyer
May 03rd, 2008.

The modern era of manufacturing is largely an outgrowth of controls. And that migration has had a long history since the relay logic and Computer Numerical Control systems of the 1950’s.  (CNC for any younger readers that might not know what some of these crazy acronyms actually stand for). Would high speed automation of manufacturing be possible without control systems?

Probably not. And since all business is ultimately governed by return on investment, what are the implications on manufacturing processes when control systems cost in the tens of thousands of dollars as they did in the Seventies? Or over time, as the cost of control has decreased, are we justified in putting a controller on everything?

Is the alphabet soup of industrial control the basis of real distinctions in functionality? Or is it a matter of keeping the domain of a particular field of application in the hands of a few suppliers of proprietary solutions? In the early days, I think a lot of new control frontiers were driven by the demands of particular industries, Military and Aerospace applications of CNC’s where speed and precision were required, and high costs were acceptable. This stimulated the creation of a whole universe of equipment which, as with all things electronic, have become more affordable over time.

There’s the SCADA (sequential control and data acquisition) market that has been largely architected by the use of “loop controllers” and PID (proportional, derivative and integral) control models. In an early generation there were mainframe computers talking to remote hardware in process plant applications with huge wiring and infrastructure costs. The amount of time needed for the hardware to respond to changes in the process was a critical concern.

Scada applications seemed very different from PLC (programmable logic controllers) which were the electronic successor to the giant relay logic cabinets of the sixties and seventies. PLC’s promised the ability to do what relays did faster and cheaper and with programmability that would insure the systems were never obsolete. But the PLC’s kept getting faster and cheaper, so it became economically justifiable to replace and update them. And the ladder logic programming language that documented the labyrinth of relays is still the most popular language for programming control systems.

So it was a most peculiar event a few years ago when the new (at the time) Honeywell process controller platform was an Allen Bradley Contrologix. And now, (roughly) seven years later, Rockwell seeks to reinvent itself as a process control supplier.

So excuse me if I wax philosophical, but looking backwards, which you can never do except as you get older, its interesting to see that all the forms of control are based on relatively similar mathematical models of real world systems whose models execute on microprocessor based controllers. And our primary means of performance improvement is throw faster and faster processors at the problems. But processors just keep getting cheaper thanks to those magicians of silicon in the semiconductor industry. So is there a practical limit where the business opportunity is no longer profitable?

And looking forward, where are we headed? More and more applications are becoming data intensive. Many industries are able to solve the mechatronic challenges, control system challenges, but find their business governed more by the data system because of regulatory demands for traceability or the long term data needed to impact the precision of a critical process. Semiconductor, Pharmaceutical, Food and Beverage industry, Aerospace, you name it.

So is it a control system or a computer? What are the real differences? Reliability, speed, resistance to harsh operating conditions? In the early generations of CNC and PLC had very costly memory making data storage impractical. Hard disk drives today are offering 1Gigabyte of storage for 33 cents. So when is a control system a computer by another name? And where are we going over the next 5 to 10 years in the controls arena?

More to come.

No Comments