Maintenance Optimisation

Research related to efficient scheduling of maintenance has become common practice since maintenance has become a core activity within all major in- dustries and can be transformed into a MO problem (Paz and Leigh, 1994, 51). The purpose of optimisation would be to find the balance between factors such as costs, objectives and constraints. Solutions to MOproblems can con-

sider multiple criteria and aim to minimise costs while maximising reliability or other performance indicators. Sharma et al. (2011, 7) state that the most appropriateMOMconsiders aspects such as maintenance costs, objectives and reliability measures simultaneously.

Factors such as safety, health, environment, maintenance cost, failure cost, opportunity cost and replacement cost are considered within mathematical models when deriving the optimum balance between the costs of maintenance and most appropriate time to perform maintenance, or other benefits associ- ated with maintenance (Dekker and Scarf, 1998, 111). A number of tools such as reliability maintenance models, simulations, statistical methods, mathemat- ical programming and decision-making tools have been used to realiseMO. The use of these tools can combined or they can be used separately. (Dekker and Scarf, 1998, 111) states that four key aspects are involved in MO:

• describing the relevant technical system, what its function is and how important it is;

• modelling system deterioration;

• providing management with the options available regarding the system and all other relevant information; and

• determining the objective function and the technique used for optimisa- tion.

2.2.6.1 Maintenance Optimisation Classification

MOMsare classified as quantitative or qualitative. QualitativeMOMssuch as

TPM,RCMandPAMare considered to be subjective and based on experience (Arthur 2005, 252; Scarf 1997, 495). Alternatively quantitative MOMs origi- nate from applied mathematics and operational research which are determinis- tic and stochastic such as Markov decision, Bayesian models and mixed integer linear programming (Garg and Deshmukh, 2006, 206). However, criticism that these techniques are theoretical and not useful in practical real-life scenarios has been presented by Scarf (1997, 494) and Arthur (2005, 251). Arthur (2005, 252) further states that quantitative MOMs often require data which is not available and advocates the use of simpler quantitative approaches that use well-founded assumptions and arrive at an approximately optimal solution.

The objective of MO is often to maximise reliability or to minimise cost within predefined constraints. Common decision factors include optimal main- tenance intervals, optimal delay time, spare parts optimisation, HR optimisa- tion, opportunity cost and redundancy Hilber (2008, 8).

2.2.6.2 Maintenance Optimisation – Task Intervals

timisation techniques within the context of on-shore wind farms. Firstly the Modelling System Failures technique is considered. The Modelling System Failures technique presented by Davidson (1994) considers equipment failure distribution, the delay in repairs, spares management and the availability of resources in the process of investigating equipment failure patterns with aim of optimising maintenance. The process firstly requires a suitable statistical dis- tribution, representative of the equipment failure characteristics, for the asset in question.

A suitable method, such as probability plot, regression analysis or Maxi- mum Likelihood Estimation (MLE), then needs to be selected that can be used to calculate parameters of the previously selected statistical distribution. These parameters that were calculated during the previous step are then used to develop Reliability Block Diagrams (RBD) with the aim of modelling asset failures. Monte Carlo simulations can then be used to simulate the optimi- sation of key variables such as maintenance costs, spare parts management, equipment reliability and availability, and so on. (Andrawus, 2008, 49).

The second quantitative optimisation method is the DTMM model which considers the consequences of failures, inspection costs and inspection inter- vals when optimising inspection intervals. The time that an incipient failure progresses from inception to a complete failure is the key consideration when determining the best maintenance intervals.

During theRCMprocess, a subjective approach which is based on experience is taken when determining P-F intervals, as seen in figure 2.20 (Rausand and Hoyland, 2004, 395). Co n d iti o n Time S P F

S = Point where defect initiates

P = Point where defect can be identified (Potential Failure) F = Point where component fails (Functional Failure)

P-F Interval

Figure 2.20: Potential-to-Functional failure intervals (Adapted from Andrawus (2008, 55))

The frequency at whichPdM activities are conducted is determined by the P-F interval withPdMactivities generally conducted at an interval that is less than (P-F interval)/2. Scarf (1997) highlights the quantitative mathematical model called theDTMMto determine the optimal interval between inspections by considering factors such as costs, risks and performance. The delay-time is the amount of time that elapses from the defect being detected to the time that a functional failure actually occurs and is synonymous to the P-F interval. Andrawus (2008, 143) presents the comparative results, considering data requirements, analysing robustness, practicality and benefits, after applying MSF andDTMMtechniques to a case study. Both techniques exhibit different potential benefits and the author concludes that the two techniques are in fact complementary, not conflicting. However Andrawus (2008, 143) concludes his comparative study between the two techniques by stating that the DTMM

technique could be incorporated into the RCMprocess. RCM is regarded as a qualitative technique that can be used for MO, while DTMM is quantitative and combining these methods can provide a balance between often subjective qualitative assumptions and the hard-to-obtain extensive dataset.

2.2.6.3 Concept of the Delay-time Maintenance Mathematical

Model

The DTMM assumes a Poisson process of defects rate of arrival (α), delay- times that are exponentially distributed and has a mean of (1

γ), and perfect inspections where all expected failure modes are detected. The rate at which defect arrives assumes the complete failure of equipment or defects that were detected during inspection.

Period under consideration

2 3 Inspection Interval F1 F2 _t i ti 0 T

For example, assume that all the main bearings of wind turbines located within a wind farm are inspected at regular and evenly spaced intervals of ∆ within the interval [0, T ]; where T is a multiple of (∆) as depicted in figure2.21. Furthermore, two instances of defects occur (defects arrival) and are noted as F1 and F2 and underpin the principles of the DTMM. The inchoate failure F1 happens between inspection intervals ∆ and 2∆; however it is detected during the next inspection at 2∆ and then repaired, alternatively catastrophic failure F2 occurs at failure ti prior to the next inspection at 3∆.

Additional detail on the concept and derivation of theDTMM can be found in studies conducted by Baker and Christer (1994) and Baker et al. (1997). Andrawus et al. (2008) presents a case study where the FMECA and DTMM

are combined to quantitatively optimise the PdM inspection intervals for key subsystems within 600 Kilowatt (KW)wind turbines.

2.2.6.4 Improvement Processes

Ahuja (2009, 747) and Borris (2006, 154) state that various improvement methodologies exist such as Six Sigma, 5S, Define-Measure-Analyse-Improve-

Control (DMAIC), Kanban (which means “signboard” in Japanese and is a

scheduling system for lean manufacturing and just-in-time manufacturing), design for manufacture and assembly (DFMA), FMEA, TPM, Just-in-Time (JIT), Total Quality Maintenance (TQM) and Quality Function Deployment (QFD).

Focused improvement is regarded as the systematic identification and elim- ination of losses (Ahuja, 2009, 722). According to Borris (2006, 11) cross- functional teams are often used to investigate issues and to find permanent solutions during focused improvement projects. The problem under investiga- tion is often evaluated with a cost-benefit analysis to justify the action (Borris, 2006, 11).

Six Sigma is methodology for systematically improving processes or prod- ucts and relying on statistical and scientific methods. Six Sigma also has an improvement procedure called DMAIC. DMAIC is similar in function to its predecessors in manufacturing problem solving, such as Plan-Do-Check-Act and the Seven Step method (de Mast and Lokkerbol, 2012, 604). Misra (2008, 176) also mentions the DMADV method which is Define, Measure, Analyse, Design and Verify.

Borris (2006, 154) also states that methods such as RCM and Six Sigma

DMAICfollow the same improvement methodology which would consider iden-

tifying and clarifying the problem, evaluating the cost and risk of the problem, understanding the root cause and finding an appropriate cost-effective solu- tion, justifying the expense, planning and completing the corrective actions,

confirming that the applied solution is effective, and finally setting up a system for checking, training, and maintaining the new standard (Borris, 2006, 154).