An integration of multi-criteria decision making techniques with a delay time model for determination of inspection intervals for marine machinery systems
Introduction
The need for effective maintenance cannot be over-emphasized, especially in the marine industry where equipment failure can result in severe and potentially irreversible damage to personnel, equipment and the environment. Having an effective maintenance scheme in place can eliminate or reduce accidents on ocean going vessels used for the movement of global commodities. This will result in an increase in vessel availability, reduced downtime and, invariably, improved company productivity.
Maintenance is defined as a combination of activities to retain a component in, or restore it to, a state in which it can perform its designated functions [12]. These activities generally involve repairs and replacement of equipment items of a system that may either be performed based on the condition of the system or based on a definite time interval. Basically there are three types of maintenance; (1) corrective maintenance, (2) preventive maintenance and (3) condition based maintenance.
In the condition based maintenance methodology there are basically two approaches for monitoring the condition of an item of equipment or component; continuous and periodic. For the continuous monitoring type, the condition of equipment is continuously monitored using some form of measurement and/or diagnostic tools. The challenge of this approach is that it is quite expensive and on this basis many maintenance practitioners prefer the periodic monitoring technique which is more cost effective. However the major difficulty in the periodic monitoring approach is in the timing of the inspection interval of the condition monitoring activity because of the possibility of failures occurring between inspections [18]. In the course of monitoring the state of an item, if a defect is found a repair or replacement task is scheduled and if possible it is executed immediately in order to prevent the equipment from further deterioration. If inspections are not carried out then slowly developing defects will go unnoticed and can lead to catastrophic system failure with severe economic loss for the company. However even if inspection tasks are performed, if they are not properly timed then defects can still occur between successive inspections. It is thus obvious that the determination of the optimal inspection interval is central to the effective operational monitoring of any mechanical system. In conventional practice, the inspection interval is determined by maintenance practitioners relying on experience and/or on the equipment manufacturers’ recommendation and the results from this approach are far from optimal and also conservative [9].
Inspection tasks, as a maintenance approach for an equipment item, can only be beneficial if there is a sufficient period between the time that a potential defect is observed and the actual time of failure of the equipment. Hence the time that elapses between the point of failure initiation and the point when the failure becomes obvious is vital in estimating the inspection interval. This phenomenon is referred to as the P-F interval within the classical Reliability Centered Maintenance (RCM) frame work and is illustrated in Fig. 1. [22] Moubray defined RCM as “a process used to determine what must be done to ensure that any physical asset continues to function in order to fulfil its intended functions in its present operating context.” Different maintenance strategies such as corrective maintenance, scheduled overhaul, scheduled replacement and scheduled on-condition task are integrated in achieving this goal.
In Fig. 1, point P is the point of failure initiation, F is the point where the actual failure occurs. The time that elapses between points P and F is referred to as the P-F interval (TPF). In classical RCM, the P-F interval principle is applied in determining the frequency of the condition monitoring of equipment and it was suggested that an inspection interval (T) be set at T ≤ TPF/2 Arthur, 2005. The author however stated that one major challenge of the use of the P-F approach is that there is usually no data to evaluate the P-F interval (TPF) and in most cases the evaluation is based on experts’ opinion. [22] Moubray on the other hand, suggested five ways of determining the inspection interval based on P-F but the author concluded that: “it is either impossible, impractical or too expensive to try to determine P-F intervals on an empirical basis”.
Apart from the use of the P-F approach, the delay time concept has been employed by many authors in the field of maintenance engineering in the modelling of inspection intervals [30]. The introduction of this concept can be traced to Christer [6]. The delay time categorises the failure process of machinery into two phases; the first phase is the time period from when the machinery is new to the time that it starts showing signs of some degradation. The second phase is the time period from when it starts showing some sign of performance degradation to the time when the machinery eventually fails. The elapsed time between when the machinery first shows signs of performance degradation and when it eventually fails is referred to as the delay time. The delay time concept is illustrated in Fig. 2.
In Fig. 2, hf represents the delay time; pf represents the time of the initial machinery performance degradation and f represents the time that the machinery eventually failed. The most appropriate time to perform a maintenance inspection is within the machinery delay time and if it is performed then, the fault will be detected and if the necessary preventive maintenance, such as repair or replacement of the machinery, is executed failure will be averted. However if inspection is not carried out then the machinery degradation will continue until failure occurs at point f.
From the above, it is obvious that the Delay Time concept introduced by Christer is the same as the P-F interval principle described within the framework of the classical RCM. However the major difference is that each approach uses a different mathematical model in the evaluation of the time that elapses between the point of failure initiation and the point when the failure becomes obvious. For the delay time concept, as proposed by Christer, a statistical distribution, such as a Weibull or an exponential distribution was utilised, while the subjective technique was applied in determining the P-F interval within the framework of the classical RCM. Additionally, in the delay time approach a different mathematical modelling technique was used in the determination of optimal inspection intervals.
Christer and Waller [7] applied the delay time concept in the development of two inspection maintenance models for determining the inspection frequency for a complex industrial system. Two different models; cost function and downtime function, were constructed with the assumption that inspection is perfect. The cost function model shows the relationship between the inspection interval and the cost for performing inspection at that particular time while the downtime function model shows the relationship between inspection interval and the resulting downtime for performing an inspection at that particular time. The study was further extended by introducing a model to cater for imperfect inspection. Numerical examples were provided to demonstrate the applicability of these methodologies.
Christer and Waller [8] proposed both an integrated delay time model and a snapshot model for determining an appropriate inspection plan for a canning-line plant in a production company in order to reduce the potential system downtime. The integrated model was used to model the downtime consequences of the system for every inspection maintenance interval. The data applied in analysing the models was obtained subjectively i.e. based on experts’ estimates through the administering of questionnaires.
Wang [34] proposed a novel model for estimating delay time distribution from the combination of experts’ judgements in the face of insufficient or a lack of reliable data. The author also proposed a technique for combining experts’ opinions as well as a model for updating the estimate of delay time distribution in a situation where maintenance and reliability data becomes available. One of the most important features of the approach is the suggestion of the use of probability estimates rather than point estimates in designing a questionnaire. The author compared the delay time distribution obtained using point estimates with that obtained using probability estimates using two case studies. From the results of the two case studies it was concluded that the delay time distribution obtained using a probability estimate presented a better result than the one obtained using a point estimate. In a related paper, [35] Wang and Jia presented an integrated empirical Bayesian based technique with a delay time model for determining the inspection interval for an industrial boiler. The empirical Bayesian model was introduced for the purpose of utilising both subjective and objective data in estimating delay time distribution parameters.
Tang et al. [33] postulated that for a part of a system subjected to wear, objective data should be applied in estimating parameters of the delay time model. On this basis they stated that there is a need for continual functional inspection and repair for such systems so as to reduce unscheduled downtime and lead to an increased record of maintenance data. Taking into consideration the wearing parts of a system, a model based on the delay time concept was developed for both perfect and imperfect inspections. To demonstrate the applicability of their proposed models, two case studies were presented; a blowout preventer core and a filter element, both components of an oil and gas drilling system. Failure and maintenance data obtained relevant for both parts were used to estimate the delay time distribution parameters.
The papers reviewed were studies that had been carried out in non-maritime sectors, such as manufacturing, building and automobile industries. From the literature some limited work has also been reported with respect to the application of the delay time concept for developing inspection plans for maritime systems. Pillay et al. [25] applied the expected downtime model based on the delay time concept in order to determine appropriate inspection periods or intervals for fishing vessel equipment items. The inspection plan was developed with the aim to reduce vessel downtime as a result of machinery failure that could occur between discharge ports. To demonstrate the applicability of their approach, reliability data gathered from the winch system and complemented with experts’ opinions, was applied to the proposed model. The case study results showed that an inspection period of 12 h was appropriate for the system. In a related paper, Pillay et al. [26] utilised both the expected downtime function model and the expected cost function model based on the delay time concept, in determining the optimum inspection period for the fishing vessel. In order to obtain a compromise inspection period, the expected cost was plotted against expected downtime consequences. Arthur [3] used the delay time model in order to establish an inspection interval for condition monitoring of an offshore oil and gas water injection pumping system. The purpose of introducing the delay time concept was to produce an alternative inspection plan for the system that was more cost-effective than the current inspection regime of a one month cycle. Data was obtained from the Computerised Maintenance Management System (CMMS) and subjected to screening. From the data scrutiny, only one failure mode (bearing failure) was dominant for both the gearbox and the motor while three failure modes (bearing failure, shaft failure and impeller failure) were dominant for the pumps of the system. The author validated the observed data by comparing it with published industrial reliability data. The validated data was then used as an input into the delay time model in order to obtain the mean delay time and inspection interval for each of the components of the system. The delay time model that was proposed produced an inspection interval of 5 months against the current interval of 1 month with annual cost savings of £21,000. The approaches reviewed for maritime applications suggested mainly single criteria being utilised in the determination of inspection intervals, however in practical situations multiple criteria are generally involved in making such vital decisions. These multiple criteria are in most cases conflicting with one another and in such a scenario, the use of multi-criteria decision making tools for aggregating decision criteria into a single analytical problem becomes imperative. The purpose of this paper therefore is to apply MCDM tools in combination with the delay time concept in producing an efficient tool for the selection of the optimum inspection interval for marine machinery systems.
The paper is organised as follows: In Section 2 the proposed methodology for determining the optimum inspection interval is presented; in Section 3 the case of the water cooling pump is presented to demonstrate the applicability of the proposed methodology. Finally the conclusion is presented in Section 4.
Section snippets
Proposed inspection interval determination methodology
In this paper, the delay time model was used in conjunction with MCDM techniques in order to determine the optimum inspection interval for marine machinery equipment. The decision criteria; expected downtime D(T), expected cost C(T) and expected Company Reputation R(T) were considered and modelled using the delay time concept. The MCDM techniques are used in aggregating the expected downtime, expected cost and reputation models into a single analytical model. The weights of the decision
Case study: marine diesel engine—sea water cooling pump
The sea water cooling pump is used as a case study in this paper to illustrate the applicability of the proposed integrated MCDM techniques and the delay time model. The sea water pump is one of the equipment items of the central cooling system of the marine diesel engine. In ongoing research, FMEA analysis of the entire marine diesel engine was carried out and, from the analysis, the sea water pump failure modes were identified as being among the most critical failure modes of the marine
Conclusion
In monitoring the condition of an asset, the two options are continuous condition monitoring and periodic condition monitoring. The periodic monitoring approach is commonly used because it is more cost effective than the continuous condition monitoring approach. However the major challenge of the periodic condition monitoring technique is the determination of the most appropriate interval for performing inspection. Traditionally, maintenance practitioners rely on their experience in determining
Acknowledgements
The authors would like to thank the Federal University of Petroleum resources, Effurun, Nigeria for funding this research through the TETFUND academic staff training intervention fund. They would also like to thank the School of Marine Science and Technology, Newcastle University, United Kingdom for providing the enabling environment for conducting this research.
References (37)
- et al.
Comparative analysis of MCDM methods for pipe material selection in sugar industry
Expert Syst. Appl.
(2014) - et al.
Multi-attribute risk assessment for risk ranking of natural gas pipelines
Reliab. Eng. Syst. Saf.
(2009) - et al.
AHP-based methodology for selecting safety devices of industrial machinery
Saf. Sci.
(2013) - et al.
An integrated multicriteria decision making methodology using compromise solution methods for prioritising risk of marine machinery systems
Ocean Eng.
(2015) - et al.
Material selection based on ordinal data
Mater. Design
(2010) - et al.
A review on machinery diagnostics and prognostics implementing condition-based maintenance
Mech. Syst. Signal Process.
(2006) - et al.
Development of a risk-based maintenance (RBM) strategy for a power-generating plant
J. Loss Prev. Process Ind.
(2005) Use of multicriteria decision analysis methods for energy planning problems
Renew. Sustain. Energy Rev.
(2007)Reliability centered maintenance
Reliab. Eng. Syst. Saf.
(1998)- et al.
Multicriteria analysis: survey and new directions
Eur. J. Oper. Res.
(1981)