Software speeds dynamic process control simulation

Engineers have used VisSim and VisSim/Real-Time software to develop a high-fidelity dynamic simulation model of DuPont's non-woven sheet manufacturing facility in Richmond, Virginia.

In a joint effort, engineers from EI DuPont de Nemours and EnTech Control Engineering used Visual Solutions' VisSim software, and its real-time add-on package VisSim/Real-Time, to develop a high-fidelity dynamic simulation model of DuPont's non-woven sheet manufacturing facility in Richmond, Virginia.

The model consists of approximately 31,500 blocks and 250 differential equations and simulates roughly a half dozen interrelated processes.

It is used by DuPont's control and design engineers to verify process dynamics during product transitions; develop and tune control strategies; and explore possible design changes to enhance control performance.

In addition, system operators train on the model to maintain proficiency and learn new procedures without impacting plant operations.

According to Hank Graeser, Senior Engineer at DuPont: "VisSim is a highly intuitive environment for developing large scale high-fidelity process models.

The DuPont Spruance model, developed in VisSim, has saved the company an estimated one million dollars to date.

We developed the model in a third of the time it would normally take using conventional methods.

VisSim's block diagram interface made it easy to document and maintain the model.

Every time we use the model for control design and off-line tuning, DuPont saves significant dollars as plant down time is reduced.

We also train our operators using the VisSim model".

As DuPont engineers drew up the specifications for the model, EnTech engineers were tasked with finding the best simulation software with which to build it.

The sheer scope of the model, which included the entire DuPont facility 15 tanks; 20 sets of pumps, lines, and valves; refiners; headbox and drainage table; vacuum devices, dryer cylinders, and scanning sensors warranted a simulation software package capable of modelling and simulating large, multi-variable dynamic processes with a high degree of fidelity.

The software had to be interactive and graphically oriented so that dynamic information could be presented in an intuitive manner.

In addition, the block set had to include a complete selection of continuous, discrete, transfer function, Boolean, arithmetic and I/O blocks.

Other key requirements included the capability to run in simulated time, real time, and continuous time; drive real-time analogue and digital I/O; stop and continue simulations; initialise all state variables; and extend the block set with custom blocks written in C, for enhanced speed and additional functionality.

Because system operators would also use the simulation model for training purposes, the ability to create realistic control panels with controller faceplates, dynamic tank levels, and built-in alarms was also important.

Based on these requirements, the simulation software that best met EnTech's needs was VisSim and the VisSim/Real-Time companion software.

According to DuPont and EnTech engineers, the DuPont model simulates the outputs of 80 sensors and transmitters and accepts input from 50 controller outputs.

In addition, the model provides high integrity dynamics as "seen" through the eyes of the actual sensors and transmitters, with a time constant in the range of about 3s.

This means that the truly fast dynamics, such as that of incompressible fluid flow, which typically have time constants of 20ms, do not have to be solved rigorously.

Instead, the equations associated with the pump curves, fluid flow, and control valve characteristic curves can be approximated by solving only the nonlinear algebraic equations.

These "algebraic loops" involve the online iteration from the last known flow and are solved using algebraic loop time constants of typically 1s, which provides an adequate safety margin compared with the high-fidelity specification of 3s.

In the resulting simulation, the time constant spread ranges from a fast value of 1 second to that of the mixing time constant for some tanks of 20 minutes.

This time constant spread, even though quite large, means that the simulation avoids some of the pitfalls of "stiff systems of differential equations" which are very difficult to solve numerically.

The final simulation model is organised in a multilayer format in which detailed simulation sub-elements collapse into "compound blocks".

There are 900 compound blocks, in about six layers, organised in an easy-to-follow, process-oriented layout.

In the testing phase, close to 200 real-time I/O channels were used to validate the model and control hardware.

The simulated process runs ten times faster than the real process on a Pentium 100MHz personal computer, at a simulation step size of 0.5s.

The DuPont model is an excellent example of how a dynamic multivariable process control model can be developed using VisSim and VisSim/Real-Time.

Graeser and other DuPont engineers have observed a "close match" between model and actual plant data.

Based on these results, they are confident that the DuPont model can be used as a "life-cycle" tool to faithfully predict the effects of future design decisions before modifications are actually made to the plant.

VisSim is supplied and supported in the UK and Ireland by Adept Scientific.

Back to top