AS_2

CONTROLLO

tecnica

Automazione e Strumentazione

Marzo 2017

Control engineers working in a processing plant can have differ-

ent origins, backgrounds and strengths. It is their job to analyse

and improve existing systems, and to design new systems to meet

specific needs. When designing a system, or implementing a con-

troller to augment an existing system, it is necessary to follow

some basic steps like modelling the system, analyse such model,

design the system/controller and finally, implement and test the

controller. For decades, the first two steps have been based on

using transfer functions, frequency-domain analysis, and Lap-

lace transform mathematics. For single control loops lineal sys-

tems - like those from the electromechanical areas from which

these classical control techniques emerged - this approach is well

suited. As an approach to the control of hydrocarbon and chemi-

cal processes, which are often characterized by multi-loop, non-

linear systems and large doses of dead time, such classical control

techniques have some limitations.

Process Simulation

The key benefits of process simulation are related to the improved

process understanding that it provides. By understanding the pro-

cess more fully, several benefits follow naturally. These include

enhanced profitability, safer designs, improvements in control

system design, improvements in the basic operation of the plant,

and improvements in training for both operators and engineers

[2]

. Using a first-principles dynamic model, control philoso-

phies can be designed, tested, and even tuned prior to start up

(see

Figure 1

for a dynamic simulation model example).

Using adequate data connectivity protocols (like OLE for Pro-

cess Control, OPC) rigorous dynamic models are nowadays even

used to checkout

distributed control systems

(DCS) or other

standard control systems configurations. All of these features

make dynamic simulation ideally suited to control applications.

However, process control systems design is, unfortunately, still

often left until the end of the plant design cycle. This practice

frequently requires to elaborate a control strategy in order to

make the best of a poor design. Dynamic process simulation,

when involved early in the design phase, can help to identify the

important operability and control issues and influence the design

accordingly. Clearly, the ideal is not just to develop a working

control strategy, but also to design a plant that

is inherently easy to control.

With current availability of powerful comput-

ers and dynamic simulators, it is possible to

approach process control system design, which

involves the fast solution of sets of differen-

tial equations without the need to move to the

non-intuitive frequency-domain mathematics

but remaining in the time-domain to solve the

differential and algebraic equations together.

In this way, engineers and operators are capa-

ble of realizing about the interactions between

the process, the control system and the load

variables in a virtual environment, identical to

the one in the real plant. Traditional boundaries

between process engineering and plant opera-

tions are dissolving as the ability to simultaneously analyze simu-

lation and plant data expands

[3]

Bridging the Gap with a Process Simulation for Control

Engineers Hands-On Course

[4]

Simulation for Advanced Classical Control

In order to bridge the gap between the strong and deep knowl-

edge about classical control techniques of nowadays control pro-

fessionals and the time-domain analysis that dynamic (and even

steady-state) process simulation is providing in real-time, Inpro-

cess proposes a hands-on simulation course, where the right com-

bination of theory and practical exercises is fulfilling such profes-

sional needs. To match course length with usual availability for

self-training of industry professionals, the content of the sessions

has been spread along twenty-five hours (three working days)

The initial course lessons are focused on breaking the barriers that

a new user might have with a commercial process simulator. Get-

ting used to the GUI and the basic steps needed to build a sim-

ple steady state case of a real distillation column, are followed as

course introduction

The column steady state model is subjected to a sensitivity anal-

ysis of the dependent vs. independent variables, exporting the

results to an external spreadsheet to calculate the steady state gains

and the process non-linearity, in the way

Figure 2

shows. Still

in Steady State, another sensitivity analysis allows students to cre-

ate a regressed equation to relate product quality as a function of

column pressures and temperatures.

Students that by then feel comfortable with a Steady State model

of the process learn how to transition it to its Dynamic version,

by incorporating the process information that is irrelevant for a

steady state solution, like equipment dimensions, equipment

heights, and piping equivalent lengths. Final control elements

(control valves and alike) are as well included, together with

some variable monitoring means (strip charts). Such an open-

loop version of the plant is step tested for mathematical robust-

ness and for variable response monitoring prior to control loop

“installation” for automatic mass-energy balance regulation by

using basic PID and split range controllers. Some basic rules of

PID tuning are introduced and tested in the model.

Figure 1 - Example of a Process Simulator model ready to run in dynamics mode