Author: Jerry Larkin, BSc (Hons), MSc, MIEI, AMIMechE, plant reliability engineer, GE Healthcare, Cork The upcoming Meeta National Maintenance Conference on 'Risk and confidence in asset management' will take place in Engineers Ireland, 22 Clyde Road, Dublin 4 on 21 and 22 November 2013. Jerry Larkin will address six fundamental issues concerning failure that will enable delegates to eliminate, reduce or mitigate failures in equipment, processes or systems. This article presents one of these – the influence of time in the selection of equipment maintenance strategies. One of the primary functions of any maintenance department is to maximise the plant and equipment performance, and thus contribute to the manufacturing/business effort – the equipment now becomes an asset. Preventing or reducing the occurrence of failure demands an understanding of the nature, sources and characteristics of failure. Fortunately, many of these characteristics are universal and enable us to deal with failures in a more methodical way. For instance, the distribution or occurrence of failures over time is a key consideration for choosing the optimum maintenance strategy. [login type="readmore"] [caption id="attachment_8848" align="alignright" width="384"] Figure 1: Common maintenance strategies[/caption] Fig.1 (right) presents the equipment maintenance strategies in a hierarchy to reflect their relative value and effectiveness. In many organisations, it may be only an aspiration to ‘design out’ – or to significantly modify the plant – in order to increase reliability and reduce failures. We will thus consider the two most common maintenance strategies: preventive (time-based) maintenance, and predictive (condition-based) maintenance. PREVENTIVE VS PREDICTIVE - CHOOSING A STRATEGY It is important to make a distinction between time-dependent and random failures. Time-dependent failure includes wear, corrosion or fatigue. A population of several identical plant items with a time-dependent failure mode (e.g. powder mills subject to abrasive wear) might exhibit the failure curve shown in red below in Fig.2 (below). There is a tight distribution around a mean life of 18 months with no failure occurring before 12 months. It is easy to see that setting up a preventive maintenance routine to service these on a 12-month frequency should prevent failures, i.e. give another period of failure-free running. Time-dependent failure modes such as wear or corrosion exhibit a narrow distribution and lend themselves to time-based preventive maintenance. [caption id="attachment_8854" align="alignright" width="663"] Figure 2: Time-dependent and random failure modes (click to enlarge)[/caption] Random failures include load variation, operation, environment, process variation and component strength. These failure distributions are exemplified by the green and blue curves in Fig.2 (right) and may have no initial failure-free period. For the group of items in blue, failures happen completely randomly with the same probability from one period to the next. One useful analogy is the human case where breaking a leg would be random failure, and hip replacement would be time-dependent. For complex/critical plant having random failure modes, the most effective policy is predictive maintenance. PREVENTIVE MAINTENANCE Preventive maintenance is actions carried out at predetermined intervals of time (or other criteria such as miles, cycles, etc), and is intended to reduce the probability of failure of a plant item. The maintenance action might be an adjustment, repair or replacement of a component, or  it could be an overhaul of the complete assembly. These preventive maintenance actions are called PM routines and are scheduled to occur at regular time (or cycle) intervals. The interval for a given plant item is derived from manufacturers’ recommendations, plant history, average life statistics or a combination of these sources. For example, a plant item might have the following associated routines:

  • Weekly – lubrication
  • Monthly – adjustment of drive chain
  • Three-monthly – check for component wear
  • Annually – complete overhaul
Preventive maintenance is best applied to plant items having time-dependent failure modes such as wear, corrosion or looseness. i.e., if a pipeline suffers abrasive wear and the pipe wall is worn through in 15 months, then the failure is prevented if the pipe is replaced (or repaired) every 12 months – this could be done conveniently during an annual plant shutdown period. The advantages of preventive maintenance include:
  • Reduced equipment failures;
  • Reduced unplanned production DT;
  • Minimised product losses;
  • Reduced labour and spares costs; and
  • Facilitation of better work planning.
The main disadvantage is that time-based preventive maintenance can result in performing maintenance tasks too early or too late. Components are replaced that may have considerable residual life left. Conversely, failures may occur before the scheduled maintenance action. Also, because the maintenance action is invasive (e.g. disassembly/component replacement/assembly), problems may be caused that did not exist before the maintenance intervention. This is the corollary of surgical (or iatrogenic) error for humans, and the incidence rates are remarkably similar i.e. circa. 20 per cent probability of causing a problem that was not there before the intervention. PREDICTIVE MAINTENANCE Preventive maintenance is initiated as a result of knowledge of the condition of an item derived from performance and/or parameter monitoring. Here, maintenance is only initiated when deterioration occurs in machine condition. This can be detected by continuous or periodic measurements of a symptom (such as temperature, flow rate, pressure or vibration), which can be taken as a parameter of the overall condition of the plant item. When a pre-determined limit is reached, the maintenance action is triggered as necessary. [caption id="attachment_8861" align="alignright" width="459"] Figure 3: Trend of deteriorating condition over time (click to enlarge)[/caption] For example, the condition of rotating equipment such as motors can be determined by measuring the vibration – perhaps every month. The measurements are trended and action taken when deterioration reaches a pre-determined level. The major difference from time-based preventive maintenance is that action is only taken when needed – if no deterioration occurs, then no action is needed or taken. In terms of advantages, predictive maintenance predicts and therefore avoids equipment failures. Early detection of deterioration prevents collateral damage. Predictive maintenance reduces unplanned production downtime, minimises product losses, reduces labour/spares costs and gives many weeks’ or even months’ notice of failure. Prediction also reduces planned production downtime and it is less invasive than preventive maintenance. However, predictive maintenance has grown and developed as a result of the limitations of time-based preventive maintenance and the main challenge is choosing the correct application. PREVENTIVE VS PREDICTIVE MAINTENANCE – AN EXAMPLE The life of an actual pump is modelled in three long-time segments as shown in Fig.4 below: [caption id="attachment_8864" align="alignright" width="289"] Figure 4 (i)[/caption] Fig.4 (i) - the pump runs in acceptable condition for an initial period of several months then suffers a catastrophic failure. The item is then brought back to acceptable condition by a repair or significant refurbishment. A PM action is scheduled to prevent further failure (i.e. time-based Preventive Maintenance policy). [caption id="attachment_8866" align="alignright" width="275"] Figure 4 (ii)[/caption] Fig.4 (ii) – a further catastrophic failure occurs at B, before PM action is due. In this case, the root cause was incorrectly assumed to be time-dependent – having taken18 months to first failure, the PM action was scheduled at a 12-monthly interval. However, because the failure mode was in fact random (could happen at any time), the PM action as scheduled could not prevent the failure at B. [caption id="attachment_8868" align="alignright" width="295"] Figure 4 (iii)[/caption] Fig.4 (iii) – the pump is again restored to original optimum condition, but now a predictive strategy is adopted. The condition of the pump is now monitored and trended, and actions taken only as necessary. Predictive maintenance enables better management of the deterioration process with condition levels now being regulated and controlled. Jerry Larkin has worked in various positions in the pharmaceutical industry including project engineer, maintenance engineer and engineering team leader. He is currently plant reliability engineer at GE Healthcare in Cork. Larkin’s academic qualifications include BSc (Hons) and a master’s degree in maintenance engineering. He is the Hon Secretary of MEETA and is a past winner of the Smith Testimonial awarded by Engineers Ireland.