Fault Detection and Diagnosis

MSc course in
Advanced Control and Systems
Engineering

Dr Tim Breikin
Control Systems Centre
University of Manchester
[email protected]

FDD,
Slide 1 Lecture 1
 Lecture 1 (2 hours)

Introduction/Motivation

FDD terminology

FDD Data and data analysis

Classification of FDD methods

FDD,
Slide 2 Lecture 1
 Introduction
General Issues:
 Teaching
 Lecture Notes
 Recommended Literature
 Internet Resources
 Examination
 Questions

FDD,
Slide 3 Lecture 1
 Recommended Literature
1. Isermann R. Fault-Diagnosis Systems – An Introduction from
Fault Detection to Fault Tolerance. Springer-Verlag, 2006.
2. Gertler J. Fault detection and diagnosis in engineering systems.
Marcel Dekker, 1998.
3. Chiang L. H., Russell E.L. and Braatz R.D. Fault detection and
diagnosis in industrial systems. Springer-Verlag, 2001.
4. Simani S., Fantuzzi C. and Patton R. Model-based fault diagnosis
in dynamic systems using identification techniques . Springer-
Verlag, 2003.
5. Fault diagnosis in dynamic systems : theory and applications /
edited by R. Patton, P. Frank and R. Clark. Prentice Hall, 1989.
6. Patton R. J., Frank P. M. and Clark R. N. (Editors), Issues of Fault
Diagnosis for Dynamic System, Springer-Verlag, 2000.
7. Barron R. (Editor), Engineering Condition Monitoring, CRC Press
1996.
FDD,
Slide 4 Lecture 1
 Motivation
• Economical and environmental risks connected with
possible destruction of industrial plant
the recent Buncefield explosion (December 2005) coursed
by a sensor failure

FDD,
Slide 5 Lecture 1
 Motivation
• Increased reliability, availability and safety of industrial
plant via predictive maintenance (maintenance by
condition)
unexpected deviations from normal operations are
estimated to cost the US economy billions of dollars per
year in the petrochemical industry alone
• Reduce maintenance time by increasing the depth of
the condition monitoring and diagnosis
helps to detect and isolate faults and the failed unit can be
replaced in short period of time

FDD,
Slide 6 Lecture 1
 Reliability, Safety and Availability
Reliability
Ability of the system to perform a required function under stated
conditions, within a given scope, during a given period of time
(continuity of service).
Safety
Ability of the system not to cause danger to persons, equipment or
environment (no catastrophic consequences).
Availability
Probability that the system will operate satisfactory at any point of time
(readiness for usage).

FDD,
Slide 7 Lecture 1
 Maintenance
Maintenance comprises actions that alter the system to keep it in (or
return to) an operational conditions.
Maintainability
Probability that the failed system can be made operable in a specified
period of time.
- Corrective maintenance
Actions to restore a failed system to an operational state
- Preventive maintenance
Actions to increase reliability of the system (generally
require shutdown of an operational system.

FDD,
Slide 8 Lecture 1
 Preventive Maintenance

Clock-based maintenance

Age-based maintainability

Usage-based maintenance

Opportunity-based maintenance

Condition-based maintenance
Actions are based on the condition of the system’s components.
This involves monitoring of one or more variables
characterising the system state.

FDD,
Slide 9 Lecture 1
 The course aim and content
The course includes following components of the FDD discipline:

Fault detection methods
Model-Free Methods

Physical redundancy

Limit checking
Signal model-based FDI

Fourier analysis

Correlation functions

Wavelet analysis
Model-based FDI

Parameter Estimation

Parity relations

State observation

The aim of the course is to


Introduce the ideas behind different FDD techniques

Describe how these techniques can be used

Provide practical skills for these techniques implementation

FDD,
Slide 10 Lecture 1
 Fault Detection and Diagnosis
Fault:
A defect, imperfection, blamable quality or feature in physical or
intellectual constitution, appearance, structure, workmanship, etc.

Detection:
The action of detecting, discovery (of what is unknown or hidden),
finding out.

Diagnosis:
Med. Determination of the nature of a diseased condition;
identification of a disease by careful investigation of its symptoms
and history.
Oxford thesaurus dictionary
FDD,
Slide 11 Lecture 1
 FDD Terminology
State and signals:
Anomaly
A deviation of at least one system parameter or system
characteristic from its known normal operating condition. Anomalous
state can be normal or faulty.
Fault
An unpermitted deviation of a characteristic property or parameter
of the system from the acceptable condition. A faulty system may still
be operable.
Failure
A permanent interruption of a system’s ability to perform a required
function. A failed system is inoperable.
Malfunction
An intermittent irregularity in the fulfilment of the system’s desired
function.
FDD,
Slide 12 Lecture 1
 FDD Terminology
State and signals:
Disturbance
An unknown and uncontrolled input acting on a system.
Residual
A fault indicator, based on a deviation between measurements
and model-based computations.
Errors
A deviation between a measured or computed value of an
output variable and its true or theoretically correct one.
Symptom
A change of an observable quantity from normal behaviour.

FDD,
Slide 13 Lecture 1
 FDD Terminology
Types of Faults:
Additive Fault
Influence a variable by an addition of the fault (offsets of sensors,
plant leaks, etc). Change in the plant outputs is independent of
the known inputs.
Multiplicative Fault
Represented by the product of a variable with the fault (parameter
changes within the process, deterioration, clogging, etc).
Abrupt Fault
Represents bias in the monitored signal, can be modelled as
stepwise function.
Incipient Fault
Represents drift in the monitored signal, can be modelled as ramp
function.
Intermittent Fault can be modelled as impulses.
FDD,
Slide 14 Lecture 1
 FDD Terminology
Functions:
Monitoring
A continuous task of determining the condition of a physical
system by recognising anomalies in the system’s behaviour.
Novelty detection
Determination of any deviation of the system from its known
normal operating condition.
Fault detection
Determination of fault or faults present in a system (indication
that something is wrong).
Fault isolation
Determination of the exact location of the fault (the faulty
component).

FDD,
Slide 15 Lecture 1
 FDD Terminology
Functions:
Fault identification
Determination of the kind of the fault and its magnitude.
Fault diagnosis
Determination of the kind, size and location of a fault (isolation and
identification)
Condition monitoring loop:

No Fault Yes Fault System

detection diagnosis repair

Faults in Controlled System?

FDD,
Slide 16 Lecture 1
 FDD Terminology
Detection Performance:
Fault Sensitivity
The ability of FDD technique to detect faults of reasonably small
size.
Reaction Speed
The ability of FDD technique to detect faults with reasonably small
delay after their arrival.
Robustness
The ability of FDD technique to operate in noisy environment and
to avoid false alarms.

Isolation Performance:
Are faults always isolable?
FDD,
Slide 17 Lecture 1
 Data Acquisition and Analysis
Data:
1. known facts used for inference or reckoning
2. quantities or characters operated on by computer
Oxford thesaurus dictionary


What data to record?

How often to record?

Good data or bad data?

Missing data?

FDD,
Slide 18 Lecture 1
 Data sampling
Maximal sampling period

sh=max is Shannon (or Nyquist) frequency

(maximal frequency of useful signal)

0=2/02sh (Shannon’s theorem

or sampling theorem)
where 0 is sampling period

General recommendation: sampling frequency 0 should be

20…30 times higher then the system bandwidth

FDD,
Slide 19 Lecture 1
 Missing or bad data
Missing or bad data should be replaced by an
estimated value
1. Average over all measured values (static modelling)

2. Windowing methods (simplest is rectangular window)

3. Previous or next available measurement

4. Other methods

FDD,
Slide 20 Lecture 1
 Fault Detection and Diagnosis
Classification 1 (Chiang at al.)
Statistical Methods

Principal Component Analysis (PCA)

Partial Least Squares (PLS)

Fisher Discriminant Analysis (FDA)

Canonical Variate Analysis (CVA)
Analytical Methods

Observer-based approach

Identification-based approach

Parity relations
Knowledge-based methods

Expert Systems

Pattern recognition (neural nets & fuzzy systems)

Case-based reasoning
FDD,
Slide 21 Lecture 1
 Fault Detection and Diagnosis
Classification 2 (Gertler)
Model-Free Methods

Physical redundancy

Special sensors

Limit checking

Spectrum analysis

Logic reasoning

Model-Based Methods (residual generation methods)


Analytical redundancy

Diagnostic Observers

Parameter estimation

Parity (consistency) relations

FDD,
Slide 22 Lecture 1
 Fault Detection and Diagnosis
Classification 3 (Isermann)
Fault Detection

FDD,
Slide 23 Lecture 1
 Fault Detection and Diagnosis
Classification 3 (Isermann)
Fault Diagnosis

FDD,
Slide 24 Lecture 1
 Lecture 1 summary
By the end of lecture 1 students are expected to:

Be familiar with FDD terminology

Be able to select sampling rate and work with
FDD data

Be familiar with classifications of FDD methods

FDD,
Slide 25 Lecture 1