Can anybody help me regarding the error of proportional controller? i am working in natural gas industry. where at custody transfer point known as sales meter station, we reduce the pressure of gas by pressure control valve having proportional only controller. Our sms's are located far apart and scattered on different areas. My problem is that load passing through is not constant, but it changes. By changing the load an error exist. If I have the value of proportional band, set point, span then can i have the idea at sitting in my office that what will be the error from the set point.

Can any body have some type of graph which will give me the idea that at what proportional band what will be my error? b/c we want to set our pressure exactly on the set point and we can't afford integral controller.

 

am afraid the solution you ask for is more expensive than adding integral action to the controller.

your question requires knowledge of the process gain including that of the valve itself to generate answer.

if you have the pressure, you can calculate the flow errors.

 

> Can anybody have some type of graph which will give me the idea that at what proportional band what will be my error?

No (if you want a longer answer, I'm sure someone will give it to you - but it will still be "no". You ~might~ be able to figure it out yourself based on practical experience, but it will never be accurate and it will change every time an engineer fiddles with the control or the process changes)

>b/c we want to set our pressure exactly on the set point and we can't afford integral controller.

You can't have it both ways. P-Only controllers cannot remove offset errors, you require integral action.

 

> b/c we want to set our pressure exactly on the set point and we can't afford integral controller.

The money your company is losing monthly by not having adequate control will be much higher than the cost of a good PID controller.

 

>> Can anybody have some type of graph which will give me the idea that at what proportional band what will be my error?

> No (if you want a longer answer, I'm sure someone will give it to you - but it will still be "no". You ~might~ be
> able to figure it out yourself based on practical experience, but it will never be accurate and it will change every
> time an engineer fiddles with the control or the process changes)

Actually i am searching what will be the steady state error of proportional controller ,off set .........

>> b/c we want to set our pressure exactly on the set point and we can't afford integral controller

 

Actually i don't want to eliminate the off set. i want to know what will be off set.

 

You can ESTIMATE the offset (not calculate it precisely) if you realise that the proportional band (PB) value is the change in the controller PV needed to drive the output from 0 % to 100 %. But you also need to know the starting point - and this can be very dependent on the type of controller. It will also depend on the controller action.

In a lot of cases, the offset is effectively set to 0 when the controller is switched from manual to auto, and you would need to have someone keep track of the PV and output at that time. Then any changes in the process needing a higher or lower output will result in an offset - the relationship between the offset and the shift in controller output depends on the PB.

If the PB is 50 %, the PV change will be 50 % of the output change. So if the output is say 65 % and the PV 75 % when first switched to Auto, the offset with an output of 70 % will be about 80 % for a direct-acting controller, or about 70 % if it's reverse acting.

So yes you can get an approximate value for the offset but it is not straightforward, and will be affected by a lot of factors including control valve nonlinearity. As others have said, if maintaining the PV at setpoint, install a P+I controller.

 

briefly, such a chart can be constructed, but it will be a family of curves that are a function of the various pressures, temperatures, valve type, valve position, and flow rate, basically a set of curves that describe a multi-dimensional space that you can identify only if you know all of the parameters at a given moment.

With integral action in addition to your proportional action, you hold the pressure more of less constant. This fixes most of the unknowns and the unknowable parameters of you installation and allow you to have a predictable measurement.

 

> Can anybody have some type of graph which will give me the idea that at what proportional band what will be my error?

No

 

My question is still there.... I am not talking about the erro. my concern is about steady state error; we are not losing our money by the small off set. i just want to know the value of off set. suppose I went to the regulator station, adjusted pressure at a point let it would be 100psi. but at the very next it would not be at 100psi. it would be at 98psi or 97 psi. and the main reason behind not installing the p I controller is the reset windup...

 

The "error from the setpoint" IS the offset.

 

>...I just want to know the value of offset..

But that is exactly what integral feedback attempts to discover empirically. It slowly add on an additional feedback offset until the error in the controlled parameter goes to zero. You just want a back door to integral control without calling it integral control. As others have said their is no way other than empirically to find out what the offset is for any given operating condition. If you had a complete mathematical description of the components involved in your control system then you would not need any feedback, proportional or otherwise. You could run open loop because you would know exactly what control output was necessary to achieve the desired controlled parameter.

>..and the main reason behind not installing the p I controller is the reset windup...

What you are describing is the result of a poorly implemented integral controller. A more intelligent integral controller could know enough to stop integrating during times when windup might be a problem. In fact one could even design a very intelligent integral control that would do exactly what you are intending to do manually. That is, it could gather long-term empirical data on operating conditions and offsets and weigh all of them when coming to a decision as to what control output offset to use in addition to the proportional control. When properly implemented windup can be completely removed.

Robert Scott
Real-Time Specialties
Embedded Systems Consultant
Hopkins, MN

 

But I am using very common type of wizard controller involving flapper nozzle technique of a very renowned company Emerson fisher control valve (type 4150 4160) and i am sure that the mathematical and experimental detail of this controller would be available. I basically want an expert opinion... according to an instrument Engineer "off set" will be 22% of the max. error...

Do you guys believe on this...

 

so for a set pint of 10 barg, a 7.8 barg discharge pressure results.

is that adequate for your metering station?

Fig. 15 of your suppliers maintenance manual shows the components needed for adding reset and anti-windup. Reset would allow control within the accuracy of the measurement say less than 1%.

usually anti-windup is required in more complicated process applications, not simple pressure control like this, for steady state demand.

good luck, authorize the expenditure and be done with it.

 

>the very next it would not be at 100psi. it would be at 98psi or 97 psi.

This is the very reason why the answer is NO. Asking the question again is not going to get you a different answer. Its still <b>NO</b>

 

A mathematical model of the controller is very easy - the output is proportional to the offset. In algebraic form,

Out = Out0 + (offset)/PB x 100

where Out0 is the output when the offset is zero (at the instant the controller is switched to auto) and offset = PV - PB0.

with PB0 the value of the PB at the instant the controller is switched to auto.

The mathematical model of the process is another matter - how far the valve must open will depend on the required flow, the actual valve characteristic, and the valve pressure drop - which in turn depends on the upstream pressure as well as the regulated pressure sensed by the controller. Since it's a feedback system, you can't simply write a single equation describing the behaviour. Even "knowing" the PB value indicated on the dial won't necessarily tell you much - the accuracy of the dials (and reading them) is not very good.

I think you need to discuss this with a Fisher rep and find out what exactly is involved in fitting an I element. Then look at what the actual requirements of the application are - is it absolutely necessary that you know the offset? Or is it just that someone somewhere thinks that things will be much tidier if all controllers are at set-point? How much is the current lack of I costing you? Then you'll have a financial justification if it is warranted.

But a lot of guys have spent a lot of time trying to help you with this, and it doesn't seem to be getting through.