This case study explores the Reliability Centered Maintenance + Decisions (RCMD) methodology and its application to modelling assets and their reliability. Modla was engaged to analyze a fleet of electrical motors operated by an international dry bulk and cargo vessel operator. The company maintains a fleet of vessels with significant exposure to costly downtime. As electrical motors are an integral part of the unloading system, they directly impact the profitability and competitiveness of the company. Moreover, electrical motors drive auxiliary systems such as hydraulics, cooling and ship stabilisation.
Modla’s primary directive was to find opportunities for improvement within the self-unloading conveyor system. Opportunities include improvements to maintenance plans to reduce downtime and labour costs by focusing on cost-beneficial tasks.
The project produced individually tailored maintenance plans for 96 unique electrical motors on a single vessel. Adherence to the proposed maintenance plans would deliver a 29% reduction in labour cost, a 49% reduction in downtime hours, and a 22% overall risk-adjusted cost reduction for a single vessel.
Following the initial model build, the RCMD methodology facilitated the development of tailored maintenance strategies for additional vessels in a fraction of the time and cost.
Modla uses a unique asset class modelling approach, which can be leveraged for large volumes of similar assets. The small pilot study was requested and applied to a single vessel, with the vision of applying the methodology to all vessels worldwide if proven effective. The most important issues that the study needed to address was:
RCMD is built on the well-known and proven Reliability Centered Maintenance (RCM) methodology. RCM has been proven to be a repeatable and robust methodology, evident by its application to the aviation industry.
Successful implementation of RCM increases cost-effectiveness, reliability, machine uptime, and the understanding of the level of risk an organisation is exposed to. However, the methodology is slow to deploy, resource-intensive and cost-prohibitive to most companies. RCM is typically difficult to deploy as it necessitates repeating the analysis for every asset or piece of equipment. RCM in its current form is not easily repeatable as every asset has a unique set of operating parameters, environmental conditions and failure characteristics. For organisations with thousands of assets, the process can stretch over many years. Even if a company has the resources to commit to the process, they are typically left with outdated results at the end.
RCMD addresses the shortcomings of RCM by increasing the ease of deployment and lessening the cost and resource implications. This is achieved by building asset class models as opposed to asset-specific models. The asset class model can be thought of as a superset of thousands of specific assets. The asset class model accounts for all feasible configurations and contextual considerations. During the analysis of a specific asset, a governance layer or decision set is applied to the asset class model. By answering plain English questions about the asset, the governance layer can produce a specific asset representative of the actual asset under consideration.
The initial development of the asset class model is resource-intensive, but once built users with little to no asset-specific knowledge can apply the governance layer and produce outputs within minutes. The outputs are of comparable quality to those produced by reliability engineers and subject matter experts.
Modla Technologies ensures the RCMD methodology produces quality outputs that converge on a standard accepted by industry and subject matter experts alike. This is primarily achieved by building open source asset class models. The models are open to the wider community to view and improve upon.
For the pilot study, an asset class model for electrical motors was built. The model development included building the governance layer or decision layer. The basis of the asset class model rests firmly on the RCM methodology. This means the asset model is built up from functions, functional failures, failure modes and includes applicable maintenance tasks.
The decision layer produces a specific asset by answering questions such as:
The application of each decision answer to the model increases its specificity until it converges on the actual asset under consideration. The electrical motor model forming the basis of the case study represents the first link in the evolutionary chain. The asset class model was built to cover all variations operated by the dry bulk carrier. The inclusion of contextual elements such as environmental and operational parameters are also heavily skewed to what is relevant to them. RCMD is an evolutionary process and the intent is to expand on and improve the model over time.
However, even the first iteration of the model and governance framework can produce specific assets representative of any electrical motor in their fleet. The model’s governance layer was fed information on 96 individual assets. From the 96 assets, 70 unique variants were identified. Unique variants could be missed due to missing information. For cases where the decision questions can’t be answered, the model defaults to the most probable input.
After the model build and information compilation, two scenarios were run. For the first scenario, the model was fed the current maintenance plan. This means the model could only perform maintenance tasks included in their current schedule at the prescribed intervals. After overlaying their maintenance plan, the model outputted the associated cost and risk profile of each asset.
For the second scenario, no specific maintenance plan was prescribed. The model was able to choose any task available to it and schedule it at an interval it deemed optimal. Following the optimisation, the model outputted an optimised maintenance plan and its associated risk and cost profile for each unique asset.
The optimisation runs produced eleven unique maintenance plans to be implemented. The optimised maintenance plan and its associated risk and cost implications were then compared to the current maintenance plan.
The results of the scenario modelling are outlined in Figure 1. For the 70 unique variants, 11 maintenance plans were produced by the optimised scenario. Comparison between the current and optimised maintenance plan resulted in 90 task changes e.g. interval changes, 98 additional scheduled tasks, and the removal of 580 tasks. The optimised scenario suggested a high number of tasks be removed. This primarily relates to inexpensive non-critical assets. It illustrates that performing inspections or fixed time preventative maintenance on inexpensive assets with no consequences upon failure is not cost beneficial.
The remainder of the section explores the generated outputs in the context of the two scenarios.
For each asset, the analyses produce a risk profile. The risk profile is presented as a Probability Density Functions (PDF). Figure 2 show an example PDF comparing the associated risk between the current and optimised maintenance plan for an 18-year-old electric motor.
The risk profile is front-loaded as the ageing asset is nearing its end of life. The probability of failure then drops off, as it is unlikely that the asset will survive past this point to be able to fail. The optimised maintenance plan significantly dampens the risk exposure but does not completely flatten it.
The risk remains relatively high, but it does represent an economic optimum where the risk of failure has been balanced against increased maintenance tasks and frequencies. Also, the model accounts for maintenance effectiveness shortcomings such as imperfect workmanship and potential failures going undetected during inspections.
Accounting for business-specific maintenance regime characteristics is another method employed by the RCMD methodology to tailor results from the asset class model.
All cost and consequence information are business specific. Thus. as standard the methodology accommodates the ability to build up costs and consequences however a business deems fit. For the example showed in Figure 3, a relatively simple costing strategy was employed. The costs associated with maintenance tasks and consequences of failure were broken up into an equipment, labor, downtime and a contractor component. However, more buckets such as environment and safety consequences can easily be included.
For the entire fleet of 96 motors the analysis and optimized maintenance plan predicted a reduced downtime losses by 49.2%. Even an optimized maintenance plan struggles to eliminate downtime given an average asset age of 14.75 years.
The optimized plan recommended employing contractors to perform thermography and vibrational analysis on critical motors. The cost is attributed to contractors as the business does not have the internal capability. Even with the addition of contractors, the direct maintenance costs (excluding downtime) have been reduced.
The figure above explores the total incurred costs over the full modelling window of 10 years. It is immediately evident that the optimized maintenance plan has dropped off non cost beneficial fixed time preventive maintenance activities at year 5 and 9. The mechanics of the model allows comparisons to be made between any number of scenarios. This is useful to evaluate the economics of preventative replacements and their risk mitigation impact.
As standard the model produces a Failure, Mode Effects and Criticality Analysis (FMECA). In addition to outputting a FMECA before optimization, a FMECA is also produced for the optimized maintenance plan. This allows for an easy comparison to assess how criticalities have changed across assets and components, and what remains high risk.
The Figure above shows a FMECA generated for the same 18 year old asset depicted in Figure 2. The probability of failure calculated below is not a point measurement but represents the overall probability of failure across the modelling window of 10 years.
For each asset a task list is produced. As standard the task lists are not specific to any Computerized Maintenance Management System (CMMS), but can easily be transformed into any load sheet. Currently the task list does not go down to the level of work instructions and only provides high level tasks and the intervals at which they should be performed.
Lastly the analysis informs spare holding requirements at a high level. The tool does not perform a detailed analysis in terms of trading off holding costs, economic order quantities and lead times. However, it provides an indication of how many spares the business will need at any given point in time. The spare holding recommendation is split between spares required for corrective (unplanned) events and planned events.
Spares required for fixed time replacements are based on the actual optimized maintenance plan. In comparison, spares required for corrective or secondary tasks (corrective action following an inspection) is determined probabilistically. In year one there is a minute probability of a bearing failure, thus the analysis recommends holding at least one spare in anticipation of the event, however unlikely it is.
The pilot study built a RCMD model of an electrical motor and produced optimized maintenance plans, risk and cost profiles for a fleet of 96 assets operated by a dry bulk and cargo carrier. The study represents the first chain in the evolutionary process. The initial build was more resource intensive than a traditional RCM analysis. However, the asset class approach means the model can now be applied and tailored to any vessel in their fleet with little to no lead time.
In comparison a traditional RCM analysis would be repeated verbatim with no cost or time saving. The RCMD approach also means any additional time spent on the model translates to improvements instead of rework. Over a few generations the model accommodates a larger number of variants operated in differing environments.
The current version of the model passed initial quality reviews and produced outputs comparable to reality. If implemented the effectiveness of the maintenance plan can be reviewed against predictions. Part of Modla's quality control involves performing root cause analysis on any discrepancies between the predicted and actual results. This filters back into the model to change assumptions and include new data and expert opinions as they come to light.