14 Note that, if the measured condition parameters have not changed, it does not mean that
the monitoring is a waste of time. It provides a peace of mind that the equipment may be in a satisfactory condition. Also, if the measured parameters did not show any trend
or change until failure, it is likely that a wrong type of information was collected. Thus, the brief discussion of condition monitoring above enables us to interpret condition-
based maintenance as an outcome of the measurement of the system condition, based on information called condition data with the aim of determining required maintenance
actions. Therefore, CBM is shown to be a method which attempts to provide a diagnosis and prognosis approach towards maintenance problems. These analyses describe the
processes of the assessment of equipment health for present and future, based upon observed data and available knowledge of the system Mathur et al., 2001.
Here, diagnosis is concerned with identifying the causes of failures or anomalous conditions in a system or its subsystems and determining the severity of given faults
once detected. Prognosis, on the other hand, is a very challenging task, which aims at predicting failures, based on observed data and available knowledge of the system, and
may lead to recommending preventive maintenance prior to the onset of catastrophic failures.
2.4 Components of Condition-Based Maintenance
Wang 2000 while reporting a review of the modelling of CBM decision support claims that there are two stages of condition-based maintenance. The first relates to
condition-monitoring data acquisition and its interpretation, followed by the second stage of making a decision based on the monitored information. Similar arguments also
can be found in Jardine et al. 2005 suggesting that a CBM programme should consist of three key steps:
1. Data acquisition, to obtain data relevant to system health. 2. Signal processing, to handle the data or signals collected in step 1 for better
understanding and interpretation of the data. 3. Maintenance decision-making, to recommend efficient maintenance policies.
15 Hence, we may conclude that the theory and implementation of condition-based
maintenance must have these components. The first component, data acquisition, plays an important role in this approach, in which the condition of the equipment needs to be
known. In general, condition monitoring can be divided into two categories Kelly and
Harris, 1978, namely monitoring which can be carried out without interruption of production, and monitoring which requires the shutdown of the unit. This categorization
is important in order to select the appropriate techniques and tools for data acquisition Shearman, 2001. Attempting to fulfil this requirement, various data acquisition
techniques and tools can be utilized within condition monitoring Williams et al., 1995. Vingerhoeds et al. 1995 discussed additional kinds of condition monitoring, i.e. off-
line condition monitoring and on-line condition monitoring. The diagnostic system for on-line monitoring enables quick and reliable fault diagnosis for the warning that
occurs. However, many challenges still exist in practical applications of on-line monitoring, such as rapid data processing, diagnosis procedure and the high operating
cost. To overcome such problems, off-line monitoring approaches have been used, in which the data is measured on-line but the analysis is carried out on a regular basis off-
line.
However, a key difficulty in condition monitoring is to detect changes that are not necessarily directly observed and that are indirectly measured together with other types
of noise. In a study of the condition monitoring of a component which has an observable measure of condition called ‗wear‘, Christer and Wang 1995 classify the
information collected during the monitoring exercise into two categories, namely direct and indirect information. They define the direct information as the measurement of a
variable that can directly determine the state of the system, for example the thickness of a brake pad or the depth of a crack. Normally, these direct methods may be based on
visual inspections or other types of monitoring such as non-destructive sensing which may not be economical to use for some equipment Jantunen, 2006. Indirect
information is defined as the associated information that is influenced by the component condition, which cannot be directly observed; for example, the information gathered in
oil analysis or vibration frequency analysis.
16 Another categorization is found in the work of Martin 1994 who distinguishes
between two different types of fault, namely soft and hard faults see Figure 2-4. The
difference between these types is important, as soft faults lead to predictable situations, hence being amenable to condition monitoring, while hard faults are basically
unpredictable; but there is a view that even hard faults must exhibit some changes before the occurrence of the failure Martin, 1994.
Figure 2-4: Hard and soft failures
One example of such a categorization is that a large amount of recent industrial research has put more effort into systems consisting of mechanical plant and equipment such as
power turbines, diesel engines and other rotating machinery, rather than electronic or electrical systems. The reason for this is that failure in mechanical systems tends to
occur slowly, so that if condition monitoring is performed, it will provide an opportunity to assess the deterioration and to compute the expected remaining life of a
system or machine, while in electronic systems, failures tend to occur without any warning or delay time.
Thus, the categorization of condition monitoring helps us to understand more about the process of deterioration, as developing prognosis models can be varied according to the
type of observed condition-monitoring data in conjunction with the defect time for the potential failure, which is not a deterministic point Moubray, 1997; Christer and Wang,
1995. Given that the initial defect time for any potential failure is not certain, it provides us with a broad view of the state of the art in fault-diagnosis techniques.
Moubray 1997 has chosen the techniques to detect these potential failures by monitoring measurable parameters
or faults as ‗equipment condition monitoring‘.
17 Others Andersen and Rasmussen, 1999
have referred to it as ‗information about technical health‘. It has been reported that major improvements have occurred in the
technology, practice and use of equipment condition monitoring over the past sixty years Mitchell, 1999. An example is the development from the mechanical
instruments that were used 20 years ago to capture a simple low frequency dynamic waveform to today‘s high-performance digital instrumentation. Methods of equipment
condition monitoring can be classified according to the monitored parameters that were influenced by the potential failure Moubray, 1997. To support his argument, Moubray
divides condition-monitoring techniques into six categories:
1. Dynamic effects, such as vibration and noise levels. 2. Particles released into the environment.
3. Chemicals released into the environment. 4. Physical effects, such as cracks, fractures, wear and deformation.
5. Temperature rise in the equipment. 6. Electrical effects, such as resistance, conductivity, dielectric strength, etc.
However, irrespective of the condition monitoring techniques used, the key elements of condition monitoring are the same: the condition data that becomes available needs to
be converted into a meaningful form and appropriate actions must be taken accordingly. As examples in this discussion, a few condition monitoring techniques that are popular
in industry have been selected from the survey conducted by Higgs et al. 2004. For other references to such methods, see Moubray, 1997 and Williams et al., 1995.
2.5 Condition-monitoring techniques