Guidelines for Setting Filtering and Module Execution Rate

Terry Blevins Principal Technologist

Presenters

 Terry Blevins, Principal

Technologist

 Kent Burr, Gary Law, Joe Nelson

– DeltaV Product Engineering

Introduction

Filtering and module execution period can directly impact control performance. In this workshop we will be addressing:

– Protection against 50-60hz pickup provided by analog input card and Charm analog input.

– Filtering of process measurements –configuration guideline to void aliasing and to minimize impact of process noise.

– Control execution – configuration guideline for setting execution period based on process dynamics, impact on control performance.

Guidelines for setting filtering and execution period are presented and examples used to illustrate their impact.

Protection against 50-60 Hz pickup

A/D

Converter

Hardware

Filter

DeltaV Analog Input Card

1 st Order

Configurable

Software

Filter

A/D Converter

3 rd Order

Sigma

Delta

Converter

FIR

Digital

Filter

CHARM Analog Input

2 nd Order

Software

Filter*

*DeltaV v11.3.1

The DeltaV analog input card uses a two pole hardware (RC) filter to provide -3 dB at 2.7 Hz and > -40dB attenuation at 50-60 Hz.

The CHARM analog input uses the A/D software (

FIR ) and configurable

2 nd order software filter after the A/D. By default will provide -3 dB at 2.7

Hz and approx – 70 dB attenuation at 50-60Hz.

A/D FIR Filter – 50-60 Hz Attenuation

Filtering of process measurements

Field Input of 4.5 Hz (green), AI output (blue) of Module executing at 5 Hz (200 msec) - Scaled inTime

The impact of aliasing for noise containing frequencies higher than ½ the module execution frequency

(Nyquist frequency) is illustrated in this examples.

Filtering to prevent aliasing can not be added at the module level since at this point the data is already aliased.

Example – Process Noise

Var 10 PI4735A.PV - Ind.

Primary Cleaner Feed Pressure psig

41.39

39.98

38.57

37.16

Time Series

DO1204AA.dat

05/29/2001 14:15:14

Var 10 PI4735A.PV - Ind.

Primary Cleaner Feed Pressure

Variance (E-3)

7.8138

5.8604

3.9069

1.9535

Power Spectrum (FFT)

DO1204AA.dat

05/29/2001 14:15:14

% Variance

100

75

50

25

35.75

0.00

102.40

204.80

Mean=38.1345 2Sig=1.671 (4.38% )

Var 10 PI4735A.PV - Ind.

Primary Cleaner Feed Pressure

Auto Correlation (FFT)

1.0

307.20

409.60

Sec

DO1204AA.dat

05/29/2001 14:15:14

0.5

0.0

-0.5

-1.0

0.0

3.2

6.4

9.6

12.8

Sec

0.0000

0.00

6.00

12.00

18.00

De-Trend=No, Win=None, Seg=0

Var 10 PI4735A.PV - Ind.

Primary Cleaner Feed Pressure

Power Spectrum Peaks

De-Trend=No, Win=None, Seg=0

Lower Threshold: 1.703E-3, Change Threshold: 2.044E-3

24.00

Cycle/Sec

0

DO1204AA.dat

05/29/2001 14:15:14

Total Variance: 0.69770

Peak Freq.

1 0.018785

2

3

4

5

0.039546

0.38086

0.70557

0.062533

6

7

0.19043

0.36664

Period

53.234

25.287

2.6256

1.4173

15.992

5.2513

2.7275

4

2

1

1

2

% Total

Shape Variance

3.557

1.246

0.8886

0.8553

1.294

1 0.7231

4 1.797

**Truncated**

P-P

Amplit.

0.44559

0.26373

0.22271

0.21849

0.26870

0.20090

0.31669

2 Sigma

Remain.

1.6406

1.6601

1.6631

1.6634

1.6597

1.6645

1.6555

Example – Process Noise

Var 07 19FC058 - Auto

Stock Flow to Tickler #1

GPM

2893

2722

2550

Time Series

CWAR0907001AG.dat

07/30/2009 15:32:31

Var 07 19FC058 - Auto

Stock Flow to Tickler #1

Variance

41.050

30.788

20.525

Power Spectrum (FFT)

CWAR0907001AG.dat

07/30/2009 15:32:31

% Variance

100

2379 10.263

25

2207

0.00

Var 07 19FC058 - Auto

Stock Flow to Tickler #1

1.0

163.84

327.68

Mean=2527.04 2Sig=169.5 (6.71%)

Auto Correlation (FFT)

0.5

0.0

-0.5

-1.0

0.00

0.64

1.28

491.52

655.36

Sec

CWAR0907001AG.dat

07/30/2009 15:32:31

0.000

0.00

3.00

Total Variance: 7182.2

Peak Freq.

% Total P-P

Period Shape Variance Amplit.

1

2

3

0.21661

0.23534

0.86070

4.6165

4.2491

1.1618

3

3

5

0.7221

0.9107

0.8980

20.370

22.875

22.714

6

7

4

5

0.30203

5.0003

1.1671

3.3109

0.063113

15.845

0.19999

0.85681

3

5

3

2

0.6074

0.9172

0.5105

0.4993

18.682

22.956

17.126

16.938

8

9

10

0.57133

0.65307

0.84623

1.7503

1.5312

1.1817

6

3

3

1.205

0.5172

0.6678

26.317

17.239

19.589

6.00

De-Trend=No, Win=None, Seg=0

Var 07 19FC058 - Auto

Stock Flow to Tickler #1

Power Spectrum Peaks

De-Trend=No, Win=None, Seg=0

Lower Threshold: 2.1918, Change Threshold: 2.6302

2 Sigma

Remain.

168.88

168.72

168.73

168.98

168.72

169.06

169.07

168.47

169.06

168.93

9.00

12.00

Cycle/Sec

0

CWAR0907001AG.dat

07/30/2009 15:32:31

1.92

2.56

Sec

75

50

Configuring Anti-aliasing Filter

Note: Help is providing in setting this filter based on module execution period.

 Rule 1 : If a measurement is characterized by process noise then anti-aliasing filtering should be applied at the IO channel.

Filtering Within a Module

 Rule 2 : To remove process noise the filter time constant of an analog input in a module should be no more than 10% of the process response time.

 Example: For a process response time of 5 seconds the input filter time constant should be no more than 0.5 seconds.

Response Time – Self-regulating Process

 The process dynamic of a self-regulating process may be approximated as first order plus deadtime and the response time assumed to be the process deadtime plus the process time constant.

Value

Gain =

O2 – O1

I2

– I1

Dead Time =

T2 – T1

Time Constant =

Note: Output and Input in

% of scale

T3 – T2

O2

Output

Input

O1

I1

T1 T2 T3

63.2% (O2 - O1)

Time

I2

•Most processes in industry may be approximated as first order plus deadtime processes.

•A first order plus deadtime process exhibits the combined characteristics of the lag and delay process.

Response Time – Integrating Process

 For integrating processes, the response time may be assumed to be the deadtime plus the time required for a significant response to a change in the process input.

Value

Integrating Gain =

O2 – O1

(I2 - I1 ) * (T3

– T2)

Dead Time = T2 - T1

Note: Output and Input in % of scale,

Time is in seconds

Output

O1

Input

I1

Time

T1 T2 T3

I 2

O2

•When a process output changes without bound when the process input is changed by a step, the process is know as a nonself- regulating process.

•The rate of change

(slope) of the process output is proportional to the change in the process input and is known as the integrating gain.

Example: Impact of Filtering (Cont)

Example: Impact of Filtering

Tuning

Method

Typical PI

Filtering as % of

Response Time

No Filtering

PID Tuning Setpoint Change

Gain Reset Rate Response* Overshoo

Time (sec) t (%)

1.13

3.5

7 0.2

Load Disturbance

Recovery* Max Dev

Time (Sec) (%)

11 6

10%

30%

0.92

0.90

4.8

6.7

11

17 -

16

22

7.2

7.7

60% 0.93

8.9

23 27 7.6

Lambda

λ=1.5

120%

No Filtering

10%

30%

60%

1.06

0.6

0.47

0.44

0.47

11.8

4.5

6.5

9.2

12.3

-

-

-

-

19

18

34

53

67 -

-

-

-

33

23

40

59

74

7.0

10.3

11.9

12.8

12.8

83 93 12.1

120% 0.52

16.9

Process Gain=1 , TC=4 sec, DT=1 sec

* Time to return within 2% of setpoint .

Control Execution Period

Deadtime (T

D

)

I

Process Input

Process Output

63% of Change O

Control Execution

New Measurement Available

To minimize delay introduced by IO processing, analog inputs are oversampled at a rate sufficient to support the fastest module execution rate.

To reduce controller load, the module execution rates is adjustable. The default execution rate is 1/sec.

Control Execution

 Rule 3 : Control loop execution period should be ¼ the process response time or less to achieve best control performance.

 Rule 4 : The module execution period should be 2X the Process

Deadtime or less.

Note: Executing control faster than the guideline provides little improvement in setpoint and load disturbance response. Quality of control will be degraded if execution is set significantly slower than the Guideline.

Example: Control Execution - Rule 3

Tuning

Method

Typical

PI

Module

Period

Gain

PID Tuning

Reset

0.2 sec 0.89

3.3

-

Rate

0.5 sec 1.01

3.9

1 sec

2 sec

1.31

1.0

5.4

14.3

-

5 Sec 0.22

22 -

9

12

47

Setpoint Change

Response* Overshoot

Time (sec) (%)

7 -

Load Disturbance

Recovery* Max Dev

Time (Sec) (%)

12 10

-

-

14

16

53

10

10

12

316 310 15

Module Execution Impact - Process Gain=1 , TC=3 sec, DT=1 sec

Example: Control Execution - Rule 3 (Cont)

Example: Control Execution - Rule 4

Tuning

Method

Typical

PI

Module

Period

PID Tuning

Gain Reset Rate

0.5 sec

1 sec

2sec

0.49

0.57

0.6

3.2

4.5

7.0

5 sec 0.21

22

10 Sec 0.12

0.44

-

-

-

-

Setpoint Change

Response Overshoo

* Time

(sec) t (%)

16 -

23 -

38 -

Load Disturbance

Recovery Max Dev

* Time

(Sec)

(%)

20 15

27 15.3

42 16.9

316

>600 -

330

>600

18

19

Module Execution Impact - Process Gain=1 , TC=2 sec, DT=2 sec

Example: Control Execution - Rule 4 (Cont)

Examples – Applying Execution Rules

Process Type

Liquid Flow/Pressure

Gas Flow

Column Pressure

0.1

0.1

1

Furnace Pressure

Vessel Pressure

0.1

0.2

Compressor Surge Control 0.05

Liquid Level

Exchanger Temperature

Batch Temperature

Column Temperature

Boiler Steam Temperature

Vessel Temperature

Gas composition – O2

Vessel Composition

Inline (static Mixer) pH

Vessel pH

30

2

30

30

10

30

10

0.05

10

10

Deadtime Time

Constant

30

30

300

600

30

300

12

0.4

1

10

0.5

10

0.5

300

2

60

Fast Process (sec)

Execution

Period

10

20*

60

60

20*

60

20*

0.1

0.2

2

0.2*

2

0.1

60

4*

60*

Deadtime

60

3

60

60

30

60

20

0.3

30

30

0.1

0.3

5

0.3

0.6

0.2

Time

Constant

300

180

500

600

180

600

60

5

30

5

1

5

50

600

5

600

Typical Process (sec)

Execution

Period

60

6*

60

60

60*

60

60

60*

60

40*

1

10

1

0.2

1

10

• Rule 4 applies Note: Maximum was limited to 60 sec. Faster update may be needed for operator visibility, calculations or alarming

Business Results Achieved

Control variability caused by process noise and unmeasured load disturbances can be minimize through tuning and by following the guidelines for module execution period and input filtering.

When plant throughput is limited by an operating constraint or variation from target operating conditions impacts operating efficiency or product quality, then a reduction in process variation provides direct economic benefit in plant operation.

Maximum

Maximum

$ Lost

$ Lost $/HR

Profit

$/H

R

Pro fit

Time

Time

Maximum

$ Lost

Maximum

$ Lost

$/HR

Profit

$/HR

Profit

Time Time

“Better” Control

“Better” Control

Summary

Easy to follow filtering and execution guidelines are proposed as a means of improving control performance and reducing process variability.

These guidelines are based on the process response time to changes in setpoint and disturbance inputs.

A reduction in process variation can provide direct economic benefit in plant operation when throughput is limited or variations impact operating efficiency or product quality.

Where To Get More Information

DeltaV Product Data Sheet, DeltaV S-Series Traditional

I/O

DeltaV Product Data Sheet, S-series Electronic

Marshalling

W.L. Bialkowski and Alan D. Weldon, The digital future of process control; possibilities, limitations, and ramifications . Vol No. 10, Tappi Journal, October, 1994.

Jeffrey Li, A PID Tuning Method Using MINLP with

Nonparametric Process and Disturbance Models, AIChE

2010 Spring National Meeting, San Antonoio, TX.