Professional Documents
Culture Documents
Failure analysis is, unfortunately, the most ignored and incorrectly performed element of the
maintenance engineering cycle. The elements of failure analysis must form an integral part of any
condition monitoring diagnosis or prognosis process.
The most difficult parts of any condition monitoring process are the diagnostic and prognostic
components. A generic process map is contained below outlining the overall process for
condition monitoring. This paper intends to outline the generic principles of causal tree analysis
and then describe its application in the condition monitoring diagnosis and prognosis processes.
SYSTEM
DESIGN
SYSTEM
DEVELOPMENT
DATA
COLLECTION
SEVERITY
ASSESSMENT
DIAGNOSIS
CONFIDENCE
FACTOR
ITERATIVE VERIFICATION PROCESS
PROGNOSIS NO
CONFIDENCE
FACTOR
YES SELECT ALL
TECHNIQUE
SUFFICIENT DATA RESOURCE PARAMETERS AND
FOR PCF> AVAILABLE MEASUREMENT
ACCEPTANCE LIMITL INTERVALS
NO NO
YES
SOLUTION
NO ACTION INTERIM ACTION RECOMMENDED
NEXT
MACHINE
INITIAL INTERIM SOLUTION
PROGNOSIS PROGNOSIS PROGNOSIS
Enteract Home
L Hit h k T h M M hi R li bilit Th Sh ll C f A t li 17/4/00
Enteract 2000, Texas, USA – Causal Tree Analysis in Condition Monitoring
Failure analysis differs from root cause analysis in that failure analysis is unique to equipment
or hardware, and focuses on the mechanisms of failure. Root Cause Analysis tends to involve
a systems approach.
In general it can be said that all failure modes can be placed in one of the five following
categories:
1. Design
2. Manufacture
3. Installation
4. Operation
5. Maintenance
There will always be a difficulty defining the boundaries of each category, however, if the same
rules are applied each time then confusion can be avoided.
The causal tree methodology involves four essential steps similar to those previously mentioned:
The essential steps in problem definition are also similar to those previously mentioned, being:
The cause identification process, however, is essentially a different process utilising causal tree
analysis techniques. This methodology has essentially four basic characteristics, being:
The graphical approach revolves around the development of cause and effect charts following
the four basic steps;
• There can be more than one Primary Effect in any given event
• A Primary effect is the point at which we start asking why
• A Primary effect is not universally known by everyone, it depends on each individual
perspective
• As we begin to understand how the puzzle of causes fits together, we often go from many
Primary effects to one
• The Primary Effect becomes the Problem name – the “What” that we want to prevent from
occurring.
• Each effect has at least two causes in the form of actions and conditions:
ACTION
(Happens)
Caused by Evidence
PRIMARY
EFFECT
CONDITION
(Exists)
Evidence
• Everything is a cause set in motion. Each effect is a set of conditions set in motion by an
action.
• Therefore, each time we ask why, we should look for Actions and Conditions to help us
focus on possible causes.
• Each effect must have evidence to support its existence:
CAUSE /
EFFECT
CAUSE / Caused
EFFECT by
CAUSE /
EFFECT
PRIMARY Caused
EFFECT by
CAUSE /
EFFECT
CAUSE / Caused
EFFECT by
CAUSE /
EFFECT
CAUSE / Caused
EFFECT by CAUSE /
EFFECT
CAUSE / Caused
EFFECT by
CAUSE /
EFFECT
Fault tree analysis is a well known methodology for diagrammatically representing failure
mechanisms and sequences. This method differs from Causal tress analysis in that it uses
logic
diagrams incorporating “and/or” conditions as well as events.
TOP Event
Delay Time
AND Gate
Transfer symbol
Fault Event Fault Event
1
Fault Event
Undeveloped Basic
Event Event
OR Gate
2 Transfer symbol
The diagram below illustrates how the casual tree process is modified for condition
monitoring processes by replacing the “evidence” with “Influences, Symptoms, Severity’s,
and Trends”.
ACTION
(Happens)
Influences
PRIMARY Symptoms
EFFECT Severity’s
Trends
CONDITION
(Exists)
Influences
Symptoms
Severity’s
Trends
Essentially diagnostics is the process of determining the root cause based on the symptom set
present. The trend data for this process will be measured. The generic process is outlined
below.
SEVERITY ASSESSMENT
DIAGNOSIS CONFIRMATION
Causal Tree analysis flowcharting can be used in the retrospective (diagnostic) sense in that
the method is used to look at the “CAUSED BY” relation ship between modes.
ACTION
(Happens)
CAUSED
By Influences
PRIMARY Symptoms
EFFECT Severity’s
Trends
CONDITION
(Exists)
Influences
Symptoms
Severity’s
Trends
Essentially prognostics is the process of determining the time to failure based on the symptom
set present and possible future symptom sets. The trend data for future deterioration rates in
this case will be estimated. The generic process is outlined below.
DIAGNOSIS
SEVERITY CORRELATION
TREND CORRELATION
DEVELOP PROGNOSIS
(NO ACTION)
Causal Tree analysis flowcharting can also be used in the predictive sense in that the method
is used to look at the “CAN CAUSE” relation ship between modes.
ACTION
CAN (Happens)
Cause
Influences
PRIMARY Symptoms
EFFECT Severity’s
Trends
CONDITION
(Exists)
Influences
Symptoms
Severity’s
Trends
In order to achieve the above modifications to the charting process the following element
block can be used. The example given can be either diagnostic or prognostic in the sense that
the severity indices are current, however, the trend data can either be current (diagnostics) or
projected (prognostics).
Load O M
Temp T W
Viscosity E
Clearance P
Load O M
Load O M
Temp T W
Position T M
Viscosity E Temp E
GEARBOX - PRIMARY
Clearance P Viscosity
Mode/Influence Tn S Tr P
Misalignment V Al
Soft Foot O
Foundation T W
Thermal Growth E
Load O Al Vibration O Ac
T Al Position T Ac
Temp
Viscosity E Temp E
Clearance P P Ac
9.0 CONCLUSION
The above approach will not be practical in many routine situations, however, it has proven
useful in solving complex problems that also require an accurate prognosis for maintenance
or production planning processes.
The main intent of introducing this concept at the ISO level is to advise analysts of the
requirement during diagnosis and prognosis to consider the following:
1. Pre-existing failure modes and deterioration rates
2. The influence between failure modes;
3. The role of each failure mode in the initiation of further ones;
4. Future failure modes and deterioration rates;
5. The influence between future and existing failure modes;
6. The sensitivity of detection of the current monitoring techniques to existing modes;
7. The sensitivity of detection of the current monitoring techniques to future modes; and
8. The design and variation of monitoring strategies to suit all of the above.