New Control

1
New Control Fault DetectionMethods for HVAC
Systems John E. Seem, Ph.D.

• Agenda
• Sequencing Control
• Adaptive Feedback Control
• Fault Detection Diagnostics

2
JCI Research Tim Salsbury, Ashish Singhal, John
Seem

• Diagnostics/Commissioning
• Test Signals setpoint changes, manual overrides
• No test signals observe normal operation
• Control
• Local Loop Adaptive
• Supervisory Sequencing, Demand limiting,

3
Feedback Control System
Chilled Water Supply
Control
Setpoint
Signal
Controller
Valve
Cooling
Temperature Sensor
Coil
4
Split Range Control
H C
C C
Split Range
Feedback Controller
Setpoint
Heating Valve
Cooling Valve
Problem Heating, Cooling,
Open
Heat Transfer
0
Closed
0
0.5
1.0
Time
Control Signal
5
Finite State Machine
H C
C C
Finite State Machine
No Heating for 5 minutes Tsupply gt Tset
Heating
Cooling
No Cooling for 5 minutes Tsupply lt Tset
Heat Transfer
0
Time
6
PI Controller
Setpoint
Loads

e
PI
u
Process
Controller
-
Gain
Integral Time
7
Problem
Although PID Controllers are common and well
known, they are often poorly tuned.
Åström and Hagglund (1988) Automatic Tuning of
PID Controllers
8
Potential Solutions

• Autotuning Tuning on Demand
• Adaptive Control Continuous Tuning

9
Background PRAC

• Continuous tuning for PI controllers
• Features
• Easy to use
• Near-optimal performance
• Robust
• Low requirements
• Laptop Tuning Tool
• PRAC in over ½ million controllers

10
Research Objective PRAC

• Correct PRAC problems
• Under-capacity systems restarted
• Extremely high controller gain
• Periodic disturbances
• On-off staged equipment
• Add rules logic to PRAC

11
Control System Simulation
12
Periodic Step Disturbances
13
Worst performance Step Periodic
14
Worst performance Step Periodic
15
Field Test Heating Coil
16
Field Test Heating Coil
17
Problem HVAC Control Systems

• Occupants detect many problems
• Data Overload
• Alarm Systems operator set levels
• Tight Alarm Levels Too many false alarms
• Loose Alarm Levels No alarms
• Trend Graphics
• Continuous Monitoring

18
HVAC Fault Detection Research

• Active Research Area
• IEA Annex 25 34
• Approaches
• Whole Building Level
• Component Level
• Difficulties
• Not many sensors
• Custom built systems
• Many control methods
• Low computational resources
• Many faults Planning, Design, Installation,
  Commissioning, Equipment, Control System

19
Whole Building Fault Detection
Which buildings have abnormal consumption?
20
Chiller Fault
21
AHU Fault DetectionWith Dr. John House National
Resources Canada

• Research Objective
• Detect leaky valves, stuck dampers,
• No additional sensors

22
Approach for AHU Fault Detection

• 1) State Machine
• Mode of Operation
• Steady-State Conditions
• 2) Model Based Residuals
• Mass Balances
• Energy Balances
• 3) Control Performance Indices

23
Residual Generation
Closed
H C
C C
Closed 5 minutes
Closed 5 minutes
Control
24
Simulation Results
25
Control Performance Indices
What control systems have poor performance?
Communication Trunk
Digital Controllers (Connected to Sensors and
Actuators)
26
Feedback Control System
Chilled Water Supply
Operator or Supervisory Control System
Setpoint
Control Signal
Digital Controller
Air Flow
Cooling Coil
Temperature Sensor
Wear f (actuator movement) Performance f
(setpoint - measured temperature)
27
Performance Indices
1) Compute EWMAs of performance indices
Smoothed value at time t
Actual value at time t

• Performance Indices
• error
• error
• process output
• setpoint
• change in controller output
• controller output
• duty cycle
• starts, stops, and reversals

Smoothing constant
2) Select smoothing constant from
Sampling period
Desired settling time
28
Field Test 24 dual-duct VAV controllers
No Problems
2 Problems!
29
Normalized Flow Errors
Electric Actuator was
1.2
Incorrectly Installed
1.0
Electric Motor had
Defective Capacitor
0.8
0.6
maximum flow rate
EWMA of flow error
0.4
0.2
0.0
0
5
10
15
20
25
30
Summary

• Stop fast switching H?C ?H ?C ?H ?C
• Tune feedback controllers
• Detect fix faulty systems