Begin by fitting a first order plus dead time (FOPDT) dynamic model to process data.

Process is defined to include all dynamic information from the output signal of the controller through the measured response signal of the process variable. Generate process data by forcing the measured process variable with a change in the controller output signal. For accurate results: - the process must be at steady state before forcing a dynamic response; the first data point recorded must equal that steady state value - the data collection sample rate should be ten times per time constant or faster (T 0.1 P ) - the controller output should force the measured process variable to move at least ten times the noise band Use Design Tools to fit a FOPDT dynamic model to the process data set. A FOPDT model has the form:

Time Domain: P
where:

dy (t ) + y (t ) = KP u(t P ) dt

Laplace Domain:
also:

Y (s) K P e P s = U ( s) P s + 1

y (t ) u(t ) KP
P P

= measured process variable signal = controller output signal = process gain; units of y(t)/u(t) = process time constant; units of time = process dead time; units of time

KC = controller gain; units of u(t)/y(t) I = controller reset time; units of time

D = controller derivative time; units of time = derivative filter constant; unitless

Values of KP, P and P that describe the dynamic behavior of your process are important because: - they are used in correlations (listed below) to compute initial PID controller tuning values KC , I , D and - the sign of KP indicates the action of the controller (+KP reverse acting; KP direct acting) - the size of P indicates the maximum desirable loop sample time (be sure sample time T 0.1 P ) - the ratio P / P indicates whether a Smith predictor would show benefit (useful when P P ) - the model itself is used in feed forward, Smith predictor, decoupling and other model-based controllers These correlations provide an excellent start for tuning. Final tuning may require online trial and error. Best tuning is defined by you and your knowledge of the capabilities of the process, desires of management, goals of production, and impact on other processes.

Standard Tuning:

IMC (lambda) Tuning C is the larger of 0.1 P or 0.8 P Conservative Tuning: C is the larger of 0.5 P or 4.0 P

* This is an ITAE
correlation as no P-Only IMC exists

KC
P-Only*

0.202

KP

( P / P ) -1.219

PI

P 1 K P ( P + C )
1 KP 1 KP 1 KP 1 KP

P P + 0.5 P P P + 0.5 P P P P 2 P + P
0.5 P

PID Ideal

P + 0.5 P + 0.5 P C P + 0 .5 P C P + 0.5 P + P C P + P C

PID Interacting

PID Ideal w/filter

P P 2 P + P
0.5 P

C ( P + 0.5 P ) P ( C + P )

PID Interacting w/filter

C C + P

Begin by fitting a first order plus dead time integrating (FOPDT Integrating) dynamic model to process data. Process is defined to include all dynamic information from the output signal of the controller through the measured response signal of the process variable. Integrating processes are unstable so the process should already be in closed loop. If not, stabilize the process with a simple P-Only controller and generate process data with a set point step. For accurate results: - the process must be at steady state before forcing a dynamic response; the first data point recorded must equal that steady state value - the controller output movement from the set point step should force the process variable to move at least ten times the noise band Use Design Tools to fit a FOPDT Integrating dynamic model to the process data set. A FOPDT Integrating model has the form: * dy (t ) Y ( s) K P e P s * Time Domain: = K P u(t P ) Laplace Domain: =

dt

U (s)

where:

y (t ) u(t ) K P*
P

= measured process variable signal = controller output signal = integrator gain; units of y(t)/(u(t)time) = process dead time; units of time

also: KC = controller gain; units of (u(t)time)/y(t) I = controller reset time; units of time

D = controller derivative time; units of time = derivative filter constant; unitless

* Values of K P and P that describe the dynamic behavior of your process are important because:

- they are used in correlations (listed below) to compute initial PID controller tuning values KC , I , D and - the sign of KP* indicates the action of the controller (+KP* reverse acting; KP* direct acting) These correlations provide an excellent start for tuning. Final tuning may require online trial and error. Best tuning is defined by you and your knowledge of the capabilities of the process, desires of management, goals of production, and impact on other processes.

IMC (lambda) Tuning

Standard Tuning: Conservative Tuning:

* This is a Ziegler-Nichols
process reaction curve (PRC) correlation as no P-Only IMC exists

C = P 10 C = 5 P 10

KC
P-Only*

1 K * P p

1
PI

2 C + P ( P + C ) 2

K* p

2 C + P
0.25 2 + C P P 2 C + P 0.5 P
0.52 + C P P 2C + 1.5P
2 (0.5C )(2C + 1.5 P )

PID Ideal

1 2 C + P K * ( C + 0.5 P ) 2 p 1 2 C + 0.5 P K * ( C + 0.5 P )2 p

2 C + P

PID Interact

2 C + 0.5 P

PID Ideal w/filter

2C + 1.5 p 1 * 2 2 K P C + 2C p + 0.5 P

2 C + 1.5 P

2 (C

+ 2C P + 0.52 )(0.5 P + C ) P
2 C 2 C + 2C P + 0.5 2 P

PID Interact w/filter

1 2C + P * 2 2 K P C + 2C P + 0.5 P

2 C + P

0.5 P