what you will learn in this seminar - pronamics control … tuning for first order processes...
TRANSCRIPT
Lambda Tuning for First Order Processes
Why Process Variability is Bad
The Role of the Control Loop in Reducing Process Variability
Current Industry Status – Control Loop Performance
Process Dynamics – First Order Processes
Lambda Tuning the Control Loop – First Order Processes
Examples of Proper / Improper Controller Tuning
What You Will Learn in This Seminar
Lambda Tuning for First Order Processes
Variability Reduces Production Rate:
Increased Off Grade
Reduced Operating Efficiency
Equipment Limitations
The Impact of Process Variability
Tightened Distribution
Setpoint Response and Regulation
Slow variability cycles
Fast variability cycles
Variability Increases Operating Cost:
Raw Material Application is Unnecessarily High Due to Conservative Targets
Raw Material Usage Inefficiency
Lambda Tuning for First Order Processes
Improved Process Performance
Product Quality Improvements
Production Increases
Chemical, Material, Energy Savings
Reliability and Maintenance
High Process variability Leads to Excessive Equipment Wear and Premature Failure
The Benefits of Reduced Variability
Chemical, Material and Energy Cost
Productivity and Yield
Product Quality
Operating Average ShiftHard Spec.Limit
OperatingAverage
Hard Spec.Limit
OperatingAverage
Minimizing variability creates the potential for
production increases and operating cost
reduction through target shifting
Lambda Tuning for First Order Processes
The Role of the Control Loop
The average installed cost of the control loop is $40,000. The controller has a key role in minimizing process variability The primary function of the control loop is to respond to process upsets/disturbances to keep the process on target. It is particularly important to minimize variability in the key processes – those that affect process efficiency and product quality.
CausticWater
MAKE-DOWN
CHEST
Dilution Header
Slow variability is
attenuated by the feedback
control loop Disturbances
Process
Dynamics
PID
ControllerProcess
Variable
PVCO
-
+S
+
+ES
L
SP
Lambda Tuning for First Order Processes
Current Industry Status
The majority of control loops do not deliver maximum value in reducing variability, and a significant percentage of loops act to increase process variability.
There are some straightforward indicators that the control loop is not functioning properly.
The Loop is in Manual Mode
The Output Signal is Fixed at 0% or 100% (i.e. Saturated)
There is Dead-Time in the Set Point Response of Flow and Pressure Loops
The Set Point Response is Very Slow or Oscillatory
There are Slow Cycles in the Process
The Loop Components Wear Out Frequently
The Trend Lines are Very Smooth
Lambda Tuning for First Order Processes
Current Industry Status
The E&I mechanic fixes control loop problems identified by the operators. The Operator often does not flag control related problems until the situation has become serious.
The majority of E&I Mechanics use ‘guesswork’ tuning. Control loop optimization cannot be achieved “guesswork”.
Successful tuning is not well defined?
The Loop is Still in Auto Mode the Day After I Tuned It
The Operator Says it’s OK, so I Can Go Home Now
The Process Variable Stays Close to Set Point
Lambda Tuning for First Order Processes
Current Industry Status
The Primary Reasons for Poor Loop Performance:
Loop Health Problems
Control Valve Backlash and Stiction
High Sensor Noise
Sensor Reliability
Poor Control Strategy Design
The Strategy Does Not Attenuate Known External Disturbances Effectively
The Strategy Does Not Account for Highly Non-Linear Dynamics
Unnecessary Control Loops Result in Interaction
Loop Design Resulting in Poor Process Dynamics
Excessive Process Dead-Time Due to Sensor Positioning
Process Gain is Too High Due to Conservative Engineering Practices
Process Time Constant is Too High Due to Unnecessary Filtering
Poor Controller Tuning
Lack of a Structured Tuning Method
The Process and Control Objectives are Not Clearly Defined
Lambda Tuning for First Order Processes
Process Dynamics and PID Controller Tuning
Optimizing controller tuning is a straight-forward and inexpensive first step in process optimization.
Controller Tuning – General Approach
Define the Process and Control Objectives of the System
Measure the process dynamics by conducting open loop bump tests.
Determine if there are Outstanding Control Loop Health Issues
Estimate the Degree of Non-Linearity in the Control Loops
Develop a tuning strategy that minimizes variability in the key process loops. Remember: It is not possible to minimize variability in all control loops. Strive to minimize variation in the key processes.
Implement / Validate the Tuning
Lambda Tuning for First Order Processes
Measuring the Process Dynamics
• The simplest method for determining the dynamic response of a
process is to conduct a Bump Test. A Bump Test is performed by
placing the controller in manual mode and stepping the controller
output. The dynamics are determined by measuring the process
response to the output step.
• Multiple and bi-directional output steps should be conducted.
Several open/closed steps should be performed at intervals such
as 1%, 2% and 4%.
Lambda Tuning for First Order Processes
Process Dynamics – First Order Plus Dead-Time
The process response is fitted to a process model. A first order response is described by the Process gain (KP), the Time Constant
(tP) and the Dead-Time θP .
Steady State Process Gain, Kp How Far PV Travels
Process Time Constant, tP How Fast PV Responds
Dead-Time, Өp How Much Delay Before PV Responds
DCO
DPV
95.0 86.5
Time to
Steady State
t t t t
98.2
63.2
KP = DPV
DOP
Process Gain
qP
Open Loop Bump Test
dPVp + PV = Kp CO(t θp)
dtt
Lambda Tuning for First Order Processes
Process Dynamics – Good Targets
An important insight is that the PID control performance is largely dictated by the process dynamics. Control loops with slow dynamics will usually require slow tuning. Attempting to speed up the PID controller beyond the inherent loop dynamics usually ends up increasing process variability.
Process Type Kp,
%Span/%CO
Flow, Hydraulic 0.5 to 2.0
Pressure,
Hydraulic
0.5 to 2.0
Concentration 0.2 to 1.0
A low process gain limits the control range. Investigate valve sizing. A high process gain limits the process resolution. Investigate pump and valve sizing.
A high time constant can limit
control loop performance. Check
sensor filtering and valve response
time.
Dead-Time is very destabilizing and
should be minimized. Check sensor
positioning, valve response.
Lambda Tuning for First Order Processes
Process Dynamics – Example
Process Gain = 6.37 lps / %CO or 2.55 %SPAN/%CO
Time Constant = 0.025 min or 1.5 seconds
Dead-Time = 0.0042 min or 0.25 seconds
Lambda Tuning for First Order Processes
Tuning is the manipulation of the controller Proportional (P), Integral (I) and Derivative (D) terms in order to produce the desired response
KC - Proportional Gain, %CO/%SPAN
tI - Integral Time, seconds or minutes
Disturbances
Process
DynamicsController Process
Variable
PVOP
-
+S
+
+ES
L
SP
Kc
CO=CO + Kc e(t) + e(t)dtbias
t I
Controller Tuning – What is it?
Lambda Tuning for First Order Processes
Controller Tuning – Popular Methods
PV
P
V
Time
PV
Fast Response
Robust
Slow Response to Disturbances
Higher Variability
Robust
Lambda Tuning (LOOP-PRO)
Growing popularity
Guesswork Tuning
Most Popular method
QAD
Still Taught in colleges
Very Fast Response
Not Robust, Watch out when process
dynamics change
Lambda Tuning for First Order Processes
Lambda Tuning (IMC) – 1st Order Process Loops
Lambda (l) Tuning Results in a 1st Order Response to a Set Point Step. There is No Overshoot. The Lambda Value is the Closed-Loop Time Constant.
The user gets to decide on the Lambda value. The Lambda value should be based on achieving the process objectives.
The process will reach a new Set Point in 4 Lambda (l) periods.
Typically, Lambda should be at least 2 times greater than the process time constant or 2 times greater than the process Dead-Time – whichever is bigger.
Setpoint Step
Process
Time
l
63.2
86.5 95
98.2
l l l
Once the process dynamics are
known, the tuning constants can be
easily calculated to achieve the chosen
Lambda value
Lambda Tuning for First Order Processes
Lambda Tuning – 1st Order Process Loops (Procedure) 1. Conduct a series of open loop output bump tests to determine the
process dynamics.
a) Identify loop health problems such as backlash/stiction. If severe then fix problem before retuning
b) If the process dynamics will result in poor control performance then upgrade dynamics before retuning
2. Select a Lambda value. The process objective should be the most important criteria in Lambda selection. On most loops, the Lambda value should be considerably slower than the process time constant and process Dead-Time to ensure a minimum level of robustness.
3. Calculate the Lambda tuning constants using the formulas shown below.
4. Enter the tuning constants into the PID controller and confirm the Lambda tuning by conducting Set Point step tests.
)P
P
C
KpK
ql
t
PI tt
Lambda Tuning for First Order Processes
Lambda Tuning – 1st Order Process Loops (Example)
) SPAN
CO
CO
SpanKpK
P
P
C
%
%25.0
sec)26(*%
%5.1
sec3
ql
t
sec3 PI tt
3 %Span
2 %Output
DPV
DOP KP = = = 1.5
%Span
%Output
qD = 2 seconds
t = 3 seconds
Select λ = 6 seconds
Tuning Calculations
Lambda Tuning for First Order Processes
Control Performance – What Can Be Expected
Hydraulic pressure and flow loops will typically have fast process dynamics. A small Lambda value (fast response) can be safely selected. Typical Lambda values of 3 to 10 seconds are achievable.
Concentration (consistency) loops often have significant process Dead-Time, limiting the Lambda selection (and control capability). Typical Lambda values range from 10 seconds to 2 minutes.
Temperature loops often have high process time constants and Dead-Time and a large Lambda value (slow response) is required.
Lambda Tuning for First Order Processes
FC
4g
pm
FC
5
gp
m
Time
Setpoint Response
Process
Setpoint Response
Process
LC
1%
Ou
tpu
t Output Bump
Time to reach desired blend
Different flow responses mean that the fibre ratio is not maintained constant during a production rate change
Lambda Tuning – Defining the Process Objectives The tuning is optimized when it supports the process objectives. Often the tuning of several loops has to be coordinated to achieve the best overall result.
The E&I mechanic needs to understand how to calculate tuning constants to achieve the desired response.
The Lambda tuning method is recommended. It allows the tuner to select a speed of response.
Hardwood
HWD
Softwood
SWD
Lambda Tuning for First Order Processes
Lambda Tuning – 1st Order Process Loops (Example)
Loop KP (%SP/%OP)
τ (sec)
qD
(sec)
FC4 0.5 5 1
FC5 1.5 2 1
Process Dynamics Loop λ
sec Kc %OP/%SP
TI
sec
FC4 10 0.91 5
FC5 10 0.12 2
Lambda Tuning
FC
4g
pm
FC
5
gp
m
Time
Setpoint Response
Process
Setpoint Response
Process
LC
1%
Ou
tpu
t Output Bump
Desired blend maintained over
entire response
Lambda Tuning for First Order Processes
Lambda Tuning – 1st Order Process Loops (Flow Chart)
Conduct Set Point Response Tests to Confirm Tuning. Measure Final
Variability.
Review Flow Chart / Define Process and Control Objectives
Select Lambda Value. Calculate Tuning Constants. Install Tuning
Constants
Measure Baseline Variability. Compare Auto/Man Variability
Conduct Open Loop Bump Tests
Is loop health acceptable? No - Repair/replace equipment
Yes
Are process dynamics acceptable? No - Reconfigure / re-engineer loop to upgrade dynamics
Yes
Lambda Tuning for First Order Processes
Summary – Achieving “Good” Loop Performance
Design in good process dynamics. Unnecessarily high process gains, Dead-Times and time constants will undermine control loop performance
Position the sensors correctly to minimize Dead-Time, sensor noise
Select the control valve carefully. Do not oversize
Do not apply excessive filtering
Maintain the loop in good health. If the sensor or the control valve is flawed, then the loop performance will be poor
Ensure that the sensor is properly calibrated
Minimize backlash and stiction in the control valve
Use a structured, scientific tuning method. The Lambda tuning method is preferred because it is relatively fast and robust.
Select Lambda values that support the process objectives and minimize variability in the key processes
Ensure that the tuning is robust - able to deal with a range of operating conditions without causing the process to cycle