Reactive and Proactive Maintenance The Best Mix

Transcription

1 Reactive and Proactive Maintenance The Best Mix Introduction The amount of proactive maintenance and reactive maintenance and the balance between them have very large effects on assets maintenance and replacement costs, their availability and any consequence costs resulting from asset failure. A purely proactive maintenance policy is the ideal policy but is not realistic. A maintenance policy that has a large element of proactive maintenance based on predictive analytics should be the objective, always recognising that however good the maintenance, assets will always fail unexpectedly and so require reactive maintenance. Indeed, one of the factors considered in PAM is the balance between proactive maintenance and reactive maintenance (see Asset Survival Simulation Module in Figure 1 summarises the trade-off between proactive maintenance and reactive maintenance. The fundamental difference between them is that reactive maintenance asks questions about the past whereas proactive maintenance asks questions about the future. Figure 1 Copyright PAM Analytics

2 - 2 - The analysis presented in this paper shows that as the proportion of proactive maintenance increases, the total asset maintenance and replacement cost decreases. Before considering maintenance policies that involve both types of policy, it is best to first consider each maintenance policy alone. Reactive Maintenance Only In a purely reactive maintenance policy assets are maintained after they fail. In effect, the policy looks backwards rather than forwards after the event maintenance. A purely reactive maintenance policy has the following features: maintenance prioritisation is rule-based, for example on the severity of the consequences of asset failure very high asset replacement costs very high maintenance costs due to repeated and therefore costly maintenance very high consequence costs due to, for example, pollution and flooding takes much longer than proactive maintenance. Let: max_capacity be the maximum number of assets that can have reactive maintenance in a month no_asset_failures be the number of assets that failed in a particular month. Consider the total cost of asset failure where the number of assets that failed in a month is firstly smaller than and secondly greater than the maximum number of assets that can have reactive maintenance in a month. IF no_asset_failures <= max_capacity: total_cost_of_asset_failure = (no_asset_failures cost_reactive_maintenance) + consequence_costs. IF no_asset_failures > max_capacity: total_cost_of_asset_failure = (max_capacity cost_reactive_maintenance) + additional reactive maintenance costs and consequence costs due to some assets not being repaired immediately, leading to higher levels of service disruption, flooding, etc. Copyright PAM Analytics

3 - 3 - Proactive Maintenance Based on Predictive Analytics In a purely proactive maintenance policy based on predictive analytics assets are maintained before they fail, and so such a policy is much more likely to overcome risks before they occur. The necessary work can then be scheduled with the required priority. A purely proactive maintenance policy based on predictive analytics has the following features. asset maintenance is prioritised based on the current risk of failure of each asset so that the risk of subsequent asset failure is reduced significantly considers how the factors and operating conditions of each asset affect its risk of failure, so leading to a better understanding of the causes of failure considers the consequences of failure of each asset costs much less than reactive maintenance. The ideal scenario is that all assets that require proactive maintenance are maintained before they fail. It assumes that assets do not fail unexpectedly and that all the work can be carried out when required. Let no_assets_proactive be the number of assets that need proactive maintenance. Therefore total_cost_of_asset_failure = no_assets_proactive cost_proactive_maintenance. Reactive Maintenance and Proactive Maintenance Assume that proactive maintenance is carried out on assets at greatest risk of imminent failure and reactive maintenance is carried out on assets that failed. For any month let: no_assets_proactive be the number of assets that need proactive maintenance. no_assets_reactive be the number of assets that fail unexpectedly. be the maximum number of assets that can have proactive maintenance. be the maximum number of assets that can have reactive maintenance. Table 1 shows the four scenarios that must be considered. Copyright PAM Analytics

4 - 4 - Table 1 Proactive Maintenance no_assets_proactive < no_assets_proactive < no_assets_proactive > no_assets_proactive > Reactive Maintenance no_assets_reactive < no_assets_reactive > no_assets_reactive < no_assets_reactive > Consider the first and third scenarios in Table 1. First Scenario total_cost_of_asset_failure = (no_assets_proactive cost_proactive_maintenance) + (no_assets_reactive cost_reactive_maintenance) + consequence_costs. Third Scenario The analysis below assumes that all the maintenance (proactive and reactive) is carried out in the current and next maintenance periods. It allows for some of the assets whose proactive maintenance was postponed to the next period failing before they are maintained. Proactive maintenance does not incur consequence costs. total_cost_of_asset_failure = ( cost_proactive_maintenance) + {SUM OVER[assets_requiring_proactive_maintenance_next_maintenance_period] (PROB(asset_surviving_to_next_period) cost_proactive_maintenance) + (PROB(asset_failing_before_next_period) cost_reactive_maintenance) + consequence_costs} + [(no_assets_reactive cost_reactive_maintenance) + consequence_costs]. Copyright PAM Analytics

5 - 5 - This analysis can be generalised to cases where the required proactive maintenance is postponed by at least two maintenance periods. This can occur when a large number of assets are in a poor state or when the organisation does not have the capacity to carry out all the work when it is required. It will lead to an lengthening backlog of maintenance, with some of the assets failing before they have the planned proactive maintenance. The reactive maintenance then carried out on these assets will be more expensive and take longer than the originally planned proactive maintenance, and lead to lower asset availability. Reduction in the Costs of Asset Failure For all four scenarios, as time goes by and some or all of the assets are maintained proactively, the total cost of asset failure decreases. In particular: fewer assets will fail and therefore require reactive maintenance proactive maintenance costs less than reactive maintenance the consequence costs will be lower and asset availability will be higher the intervals between proactive maintenance will be longer than the intervals between reactive maintenance. The analysis presented above shows that reductions in the maintenance and replacement costs of assets by proactively maintaining those assets at greatest risk of imminent failure are realised very soon after a (more) proactive maintenance policy is implemented if the organisation has the capacity to carry out all the required proactive maintenance. The asset survival models show that this leads to a virtuous proactive maintenance feedback policy. The Predicted Maintenance Interventions module (see Predicted Maintenance Interventions Module in /PAM_modules.html) enables the reduction in the assets maintenance costs and replacement costs to be realised at the operational level by modelling the change in the asset management policy from reactive fail-and-fix to proactive predict-and-prevent. The Asset Survival Simulation module (see Asset Survival Simulation Module in /PAM_modules.html) enables the reduction in the assets maintenance and replacement costs to be realised at the strategic level. Copyright PAM Analytics

6 - 6 - Summary Comparison of Reactive Maintenance and Proactive Maintenance Reactive maintenance: looks backwards, i.e. after the event maintenance, rather than forwards does not consider the state of each asset and its individual risk factors when the maintenance is carried out Proactive maintenance based on predictive analytics and discrete event simulation: is based on an analytical model and so can overcome risks before they occur considers the state of each asset and its individual risk factors when the maintenance is carried out PAM helps organisations change their asset management policies from reactive fail-and-fix to proactive predict-and-prevent, so saving them very large sums of money in maintenance costs (because proactive maintenance is much cheaper than reactive maintenance), replacement costs (because the number of assets suffering critical failures and therefore having to be replaced is reduced) and any consequence costs. Risk Tolerance PAM quantifies organisations attitude to the risk of asset failure by risk tolerance. Risk tolerance is a measure of the number of times an asset can fail in successive periods before it is replaced. Frequent failures may indicate a systematic problem with the asset. Figure 2 shows that low risk tolerance and high asset availability lead to low consequence costs and high customer service levels. On the other hand, high risk tolerance and low asset availability lead to high consequence costs and low customer service levels. Figure 2 Copyright PAM Analytics

7 - 7 - Maintenance Policies There are four types of maintenance policy, and they are described below. Reactive Maintenance Reactive maintenance is the simplest asset maintenance policy. It involves letting assets run to failure rather than maintaining and repairing them, and so should only be used for non-critical relatively cheap assets that have little or no impact on safety and whose failure does not affect business operations. Preventive Maintenance Preventive maintenance involves carrying out regular maintenance on the assets as specified by the manufacturers. The aim is to ensure that the assets will not fail. Condition-Based Maintenance Condition-based maintenance focuses on the physical conditions of assets and how the assets are working. It is most suited to cases when a change in the value of an asset parameter to a new value outside its tolerance limits is a good indicator of impending problems with the asset. The decision rule for determining if a parameter is out of range must be very stable so that it can be applied to all conditions under which the asset operates. Predictive Maintenance If an asset is critical so that its failure has severe consequences, for example damage to the environment, service interruption and safety issues, the risk of the asset failing must be minimised. One way of doing this is to use predictive maintenance. Predictive maintenance goes further than condition-based maintenance. Rather than using a decision rule to establish if a parameter is within its limits, it uses predictive analytics to calculate the likelihood of asset failure as the state of the asset changes. Copyright PAM Analytics