Weighing Control - continued

You haven't mentioned how quickly you can stop the flow.

In an ideal world you'd just open the thing all the way up, add your measured weight plus the "in flight" estimate, and as soon as the sum equals your target, then BANG! close the door. You'd be perfect every time.

That's the ideal world...

I'd make an estimator for the "in bin" plus "in flight", then have a simple proportional controller from that estimate to your flow target. I assume that the flow is measured, and that you have to control it, so I'd wrap a PI or a PID controller around that flow value.

If you do it all correctly then for most of the time your flow target will be maxed out. When you get close to the right amount, things will rapidly shut down, then you'll get a few dribs and drabs out to bring the weight right up to the dot.

If your flow control isn't super-quick, either because you can't achieve much bandwidth with the thing operating in a linear mode, or because you're demanding more speed out of the valve (or whatever actuator there is) than is available, then you'll need to ramp back the proportional gain from estimated total to flow target.

I bet you'll end up filling the bin quickly and accurately, though.

It's either an analog output card controlling a vibratory feeder controller or Profi-Bus communication to a VFD for auger motors. Shut off is however long it takes the vibratory to settle or ramp down time for the motors. The augers have doors but we have to wait for the motor to stop or else if feeds product and binds the door.

We have some loss of weight systems that work pretty much that way.

Today I tried ramping down, when it gets to the estimated ramp down point, I record the weight from the load cell and then decrement it each pass by

RogerN

Trying to tune things without bringing the line down must be a royal pain in the behind -- fortunately the loops I close are all in products, so I get to have something to play with all by my oneself in a lab.

Now I'm thinking about trying to us a flow rate as a set point for the PID and base it on the weight set point - current weight.

Since the "inFly" varies from around 2 X flow for fast flow to flow/2, I may try to set maybe flow_SP := Lbs_Err * 0.2 for high flow rates to something like flwo_SP := Lbs_Err * 1 for low flow rates.

This week the line programmer from Germany is supposed to be there to upgrade our software so I may not get much testing done. He used to support us good but since he's went from independent to employee, he hardly responds when we need help. That's why I'm trying to dig into the problems and learning to fix the line without depending on others.

I'd kind of like to find out if I can get Siemens data into a visual basic program and plot or graph the data. I use that a lot with Rockwell in trying to see what's going on in a machine.

It's fun to take a machine that runs bad, operators hate, and make it run better than the others.

RogerN

I'd try to model it somehow so you can test different approaches more quickly. I'm battling a difficult control problem myself for a backburner project where I need to control the rate of liquid going into an evaporator by using the temperature reading from a thermocouple. I already ordered a book which should be more hands-on than the theoretical books I have from my EE study.

AFAIK it is possible to dump data from a Siemens S7 PLC into a CSV file which you can load into Excel. A guy I work with every now and then is a real wizzard when it comes to Siemens industrial automation gear. I could get you in touch with him. He has projects all over the world.

The braking analogy is good, but the disk drive seek servo analogy is bette r. Closed loop HDD servo's have for decades used two PID loops to achieve min seek times and zero position error. This is the same as your requirement fo r max flow rate and zero weight error at target. The main limitation is th e rate of change of velocity or acceleration in your system controlled by s ome current.

The reason this is done is to make two second order feedback loops and avoi d the effects of a 3rd order system instability created by the 2nd derivati ve of position or weight count in your case.

Create a setpoint profile for Velocity and use PID control for velocity fee dback with a transition from Velocity control (or rate of weight transfer) to Position Control ( or accumulated weight control) within the de-accelera tion or braking limits of your system.

You are still counting weight transfer (Position) but control starts with a velocity error loop then switches to position error loop when your positi on error or weight remaining is < X. Then your PID weight loop takes over from the PID velocity loop for weight/second. This is how the better close d loop servos work for max speed and zero error. For zero overshoot, the PI D loop on position error must be critically damped.

Disturbances and target error, such as mechanical stiction of mass flow con trol must be considered.