Maintenance, Reliability and Asset Management

Introduction and context

Production systems create value only when their equipment, facilities, information systems and human processes perform as required. Maintenance is therefore not merely a repair function. It is a managerial system for preserving or restoring an asset’s ability to deliver safe, capable and economically valuable service. Reliability concerns the probability that an asset performs its required function for a specified period; availability concerns whether it is ready for use when required; and asset management integrates technical, financial, operational and risk decisions across the asset life cycle.

For an operations manager, the central question is not “How can we minimize maintenance expenditure?” but “What combination of maintenance, inspection, renewal and operational controls produces the required performance at acceptable life-cycle cost and risk?” Excessive preventive work can waste resources and introduce maintenance-induced failures, whereas insufficient maintenance creates breakdowns, unsafe conditions, lost production and environmental exposure.

Learning objectives

After studying this chapter, you should be able to:

1. Distinguish reliability, maintainability, availability and asset-management performance.
2. Explain common failure patterns and select appropriate maintenance policies.
3. Calculate basic reliability, failure-rate, maintainability and availability measures.
4. Compare corrective, preventive, predictive, condition-based and total productive maintenance.
5. Use risk, life-cycle cost and criticality to prioritize maintenance decisions.
6. Evaluate the managerial, ethical, regulatory and technological implications of maintenance strategy.

1. The maintenance system and the asset life cycle

An asset is an item, system or network that provides a service or produces value. Examples include a CNC machine, a power transformer, a warehouse conveyor, a vehicle fleet, a hospital MRI scanner and enterprise software. Asset management begins before acquisition and continues through design, commissioning, operation, maintenance, modification and disposal.

Maintenance activities are usually classified as follows:

• Corrective maintenance: action taken after a failure has occurred to restore function.
• Preventive maintenance: planned action performed at predetermined intervals or according to prescribed criteria to reduce the probability or consequence of failure.
• Predictive maintenance: use of measurements, models and trends to estimate degradation and determine when intervention is needed.
• Condition-based maintenance: maintenance initiated by evidence that an asset’s condition has crossed, or is approaching, a decision threshold. Predictive maintenance is a major means of implementing condition-based maintenance.
• Proactive maintenance: action directed at eliminating root causes of failure, such as poor lubrication, misalignment, contamination or unsuitable operating conditions.
• Improvement maintenance: modification of equipment or procedures to improve capability, safety, maintainability or reliability.

A useful asset-management decision considers five interacting dimensions: performance, risk, cost, compliance and stakeholder value. A low-cost maintenance plan is not necessarily economical if it causes unreliable delivery or unacceptable risk.

Maintenance strategy and asset criticality

Criticality expresses the importance of an asset to the organization. It is commonly assessed using consequences of failure—such as safety, environmental, customer, quality, production and financial consequences—combined with the likelihood of failure. A simple qualitative matrix may classify assets as low, medium or high criticality. A high-criticality asset normally requires stronger preventive controls, spare-parts planning, redundancy, condition monitoring and management review.

Criticality should not be confused with failure frequency. A rarely failing emergency generator may be more critical than a frequently failing non-bottleneck pump. The correct policy depends on both failure behavior and consequences.

2. Reliability concepts and quantitative measures

Reliability function

Let T be the random variable representing time to failure. The reliability function is the probability that an item survives beyond time t:

Here, F(t) is the cumulative distribution function of failure time. R(t) is dimensionless and lies between 0 and 1. The expression assumes a clearly defined failure criterion, a specified operating environment and a population or probability model relevant to the decision.

The probability density of failure time, when it exists, is f(t). The failure-rate or hazard function is:

λ(t) has units of failures per unit time and represents the instantaneous conditional rate of failure among items that have survived to time t. It is not, in general, the same as the proportion of items that fail per unit time in the original population.

The relationship between reliability and hazard is:

This requires a non-negative, suitably defined hazard function. For a constant hazard λ, the exponential model follows:

The exponential model is useful for a random-failure period with approximately constant hazard. It should not be applied automatically to wear-out failures.

Mean time measures

For a non-repairable item, mean time to failure is:

For a repairable item, mean time between failures (MTBF) commonly denotes the average operating time between successive failures. Its precise definition must be stated because organizations sometimes use MTBF to mean total cycle time, including repair and downtime. Mean time to repair (MTTR) is the average active repair time; logistics and waiting time may be reported separately or included in mean downtime.

Reliability data are frequently modeled with the Weibull distribution:

where β is the shape parameter and η is the characteristic life, measured in the same time unit as t. When β < 1, the hazard generally decreases, indicating early-life failures. When β = 1, the model becomes exponential with constant hazard. When β > 1, the hazard generally increases, consistent with wear-out. The Weibull model is a statistical approximation and should be validated against data and failure mechanisms.

Availability and maintainability

Availability is the probability that an asset is in a state to perform its required function when called upon. Several definitions are used:

For a stable repairable system under a simple renewal approximation:

where MTBF and MTTR use compatible units and downtime is represented by repair time. This is often called inherent or basic availability when logistics, administrative delays and preventive-maintenance time are excluded. Operational availability should include realistic support and delay effects. The formula must not be used when the stated definitions are inconsistent.

Maintainability is the probability that a failed item can be restored within time t under stated conditions. If repair time follows an exponential distribution with repair rate μ:

where μ has units of repairs per unit time and the average active repair time is approximately 1/μ. Real repair processes may not be exponential because diagnosis, access, parts and skill requirements vary.

Worked example: availability improvement

A packaging machine has an average operating time between failures of 240 hours and an average active repair time of 6 hours. Its approximate inherent availability is:

Thus, inherent availability is approximately 97.56%. If parts and technician waiting add 4 hours per event, operational availability using total average downtime of 10 hours becomes:

The result shows why improving logistics can produce a larger availability gain than reducing hands-on repair time alone.

3. Failure patterns and maintenance-policy selection

The traditional “bathtub curve” describes three broad hazard regions: decreasing early-life hazard, approximately constant random-failure hazard and increasing wear-out hazard. It is a conceptual model, not a universal law. Modern complex equipment often exhibits multiple failure modes, and empirical studies have shown that age-based replacement is effective for some mechanisms but not for all.

A maintenance policy should be selected by failure consequence, detectability, degradation behavior, intervention cost and data quality. A scheduled replacement is justified when failure probability rises with age and replacement is cheaper or safer than run-to-failure. It is usually inappropriate when failures are random and replacement does not reduce the risk.

Corrective maintenance

Run-to-failure can be rational for low-criticality, inexpensive and easily replaceable items where failure creates little disruption. It is not equivalent to neglect: spares, access, safe isolation and response procedures must be prepared. Corrective maintenance is unacceptable when failure threatens life, regulatory compliance, major environmental harm or severe production loss.

Preventive maintenance

Time- or usage-based tasks include lubrication, calibration, inspection, adjustment, overhaul and replacement. Their effectiveness depends on a real relationship between age or usage and failure probability. Excessively frequent tasks increase labor, spare-part consumption and the risk of human error.

Predictive and condition-based maintenance

Condition indicators may include vibration, ultrasound, thermography, oil debris, motor current, pressure, acoustic emission, energy consumption and process-quality signals. A sound program requires:

1. A failure mode and its measurable precursor.
2. A reliable sensor or inspection method.
3. A threshold or trend rule linked to an action.
4. Sufficient warning time for planning and safe intervention.
5. Competent interpretation and data governance.

Monitoring without a decision rule is data collection, not maintenance strategy.

Reliability-centered maintenance

Reliability-centered maintenance (RCM) asks what functions an asset must perform, how it can fail, what causes each failure, what consequences result, and which task is technically and economically appropriate. Its logic commonly includes functional analysis, functional-failure definition, failure-mode analysis, consequence classification and task selection. RCM may recommend condition-based tasks, scheduled restoration, scheduled discard, failure-finding tasks, redesign or deliberate run-to-failure.

Total productive maintenance

Total productive maintenance (TPM) treats equipment effectiveness as a shared responsibility. Operators perform routine inspection, cleaning, lubrication and abnormality identification, while maintenance specialists address technical tasks and improvement. TPM commonly emphasizes planned maintenance, autonomous maintenance, focused improvement, quality maintenance, training, early equipment management and safety.

Overall equipment effectiveness (OEE) is a diagnostic measure:

Availability in OEE is normally planned-production-time availability, performance compares actual speed with ideal speed, and quality is good units divided by total units. OEE percentages should not be added across machines; aggregation requires careful weighting by time, output or economic significance. OEE is not a complete measure of asset value and can be gamed by changing definitions or schedules.

4. Life-cycle cost and economic decisions

Life-cycle cost includes acquisition, installation, energy, operation, maintenance, downtime, risk, modification and disposal costs. A simplified discounted-cost model is:

C₀ is initial cost, Cₜ is cost in period t, i is the discount rate per period, and n is the analysis horizon. Costs must be expressed in consistent currency and time units. Residual value, inflation treatment, taxes and uncertainty should be handled explicitly.

A maintenance decision should compare expected total cost, not only the visible maintenance invoice. Expected annual failure cost can be represented conceptually as:

For multiple mutually exclusive failure modes, expected costs may be summed only when probabilities and consequences are defined consistently and interactions are understood. Safety and environmental decisions cannot be reduced to financial expectation alone; legal duties and intolerable risks impose constraints.

Spare-parts and resource decisions

Critical spares should be prioritized according to failure consequence, lead time, commonality, obsolescence risk, storage cost and repairability. A low-cost part with a long supply lead time may deserve a higher stocking priority than an expensive part available overnight. Maintenance capacity planning must consider labor skills, contractor dependence, shutdown windows, tools and permits.

5. Performance measurement and governance

Useful measures include:

Metrics should be defined in a data dictionary. For example, “availability” must specify whether planned shutdowns, waiting for parts and minor stops are included. Leading indicators—inspection completion, defect elimination and training—should complement lagging indicators such as breakdowns and incidents.

A computerized maintenance-management system (CMMS) supports work orders, asset registers, failure codes, schedules, labor, parts and history. Enterprise asset-management systems extend this view to procurement, finance, projects and disposal. Poor master data, inconsistent failure coding and incentives that suppress reported failures can make sophisticated systems misleading.

6. Digital technologies and AI

Sensors, industrial Internet of Things platforms, edge computing and digital twins can improve visibility of asset condition. Machine-learning models may detect anomalies or estimate remaining useful life. However, a prediction is not automatically a maintenance decision. Models require representative failure data, stable operating conditions, validated labels and controls against false alarms and missed detections.

A responsible implementation should include:

• human review for safety-critical decisions;
• cybersecurity, access control and secure software updates;
• traceability of data, model versions and decisions;
• monitoring for drift and bias;
• a fallback procedure when sensors or communications fail;
• protection of employee and supplier information.

AI should augment engineering judgment rather than obscure accountability. Explainability is particularly important when a model recommends deferring or accelerating intervention on a critical asset.

7. Applied case: illustrative pharmaceutical packaging line

Background

A fictional pharmaceutical manufacturer operates a high-speed blister-packaging line. The line contains a forming station, printer, sealing unit, vision inspection system and cartoner. A stoppage interrupts batch schedules and may cause material loss. Product quality and traceability are regulated.

Problem

The maintenance department reports high emergency-work hours and an OEE of 68%. Management initially proposes replacing all drive motors every 12 months. Historical records, however, are incomplete and show that many stoppages arise from sensor contamination, misalignment and changeover errors rather than motor wear.

Analysis

A cross-functional team creates an asset hierarchy and ranks the sealing unit and vision inspection system as high criticality. It conducts failure-mode analysis and separates failure mechanisms. Motor replacement has no demonstrated age-related pattern, while vibration and thermal trends identify bearing and alignment deterioration. Vision-system failures are often detectable through image-quality drift before a complete stoppage. Product-contact and safety controls require documented inspection and proof testing.

The team also clarifies metrics. OEE availability excludes planned sanitation but includes minor stops; emergency work is measured as a proportion of labor hours; repeat failures are linked to a common failure-mode code.

Approach and solution

The company adopts a mixed policy:

• condition monitoring for critical rotating equipment;
• weekly operator cleaning and abnormality checks for sensors and cameras;
• calibration and proof testing for inspection and safety functions;
• scheduled replacement only for components with evidence of wear-out;
• run-to-failure for selected low-cost, non-critical components with stocked spares;
• root-cause analysis for repeat failures;
• CMMS controls for work orders, permits, parts and failure coding;
• a pilot AI anomaly detector used only as a decision-support tool.

Maintenance planners calculate warning time from condition trends and reserve planned windows. Operators are trained to report abnormalities without penalty, reducing concealment of early defects.

Outcome

After six months, emergency labor falls, repeat sensor failures decline and OEE improves. The most important gain is not a single percentage increase but better causal knowledge: the organization stops applying uniform age-based replacement to assets with different failure mechanisms. Regulatory records also become more complete. The company continues to review model false alarms and does not use the pilot system for autonomous safety decisions.

Learning points

1. Criticality and failure mechanism should drive policy.
2. Better data definitions can reveal improvement opportunities before new technology is purchased.
3. Operator involvement is an operational-control mechanism, not simply a cultural initiative.
4. OEE must be interpreted with safety, quality, cost and compliance measures.
5. Predictive analytics creates value only when it leads to timely, technically valid action.

8. Challenges, limitations and ethical issues

Maintenance data are often censored because an asset has not yet failed, truncated because observation began after installation, or distorted by undocumented repairs. Aggregating unlike assets can conceal failure modes. Statistical confidence intervals and engineering judgment are therefore important when making high-cost decisions.

A replacement policy can also create waste if functioning components are discarded prematurely. Conversely, deferring maintenance may transfer risk to operators, contractors, customers or communities. Ethical maintenance requires truthful reporting, competent authorization, safe isolation, adequate staffing and refusal to bypass safeguards merely to meet production targets.

Regulations may require pressure-vessel inspection, electrical testing, lifting-equipment certification, environmental controls, functional safety validation or documented pharmaceutical calibration. Compliance is a minimum condition, not proof that an asset strategy is optimal. Managers remain responsible for ensuring that contractors are qualified and that accountability is not lost through outsourcing.

9. Best-practice implementation roadmap

A practical implementation sequence is:

1. Define required asset functions and performance standards.
2. Build a verified asset register and criticality ranking.
3. Improve failure coding, work-order quality and data definitions.
4. Analyze failure modes and consequences, beginning with bottlenecks and high-risk assets.
5. Select policies using technical feasibility, risk and life-cycle cost.
6. Establish job plans, permits, spares, skills and response arrangements.
7. Pilot condition monitoring or TPM practices on a representative area.
8. Measure outcomes using balanced leading and lagging indicators.
9. Review results, update intervals and eliminate root causes.
10. Govern changes through operations, maintenance, engineering, finance, safety and quality stakeholders.

The mature organization does not seek zero maintenance or zero failures in every context. It seeks controlled risk, reliable service, efficient intervention and continuous learning over the asset life cycle.

Chapter Summary

Maintenance and asset management integrate technical performance with operational, financial, safety and compliance objectives. Reliability is the probability of surviving a specified period; failure rate describes conditional failure intensity; maintainability concerns restoration; and availability describes readiness for service. Constant-hazard, exponential models are useful in limited contexts, while Weibull analysis can represent decreasing, constant or increasing hazard.

Corrective, preventive, predictive, condition-based, proactive and improvement maintenance are not competing labels to be applied universally. Their suitability depends on failure mechanism, criticality, detectability, warning time, intervention cost and consequence. RCM provides a structured policy-selection logic, while TPM develops shared equipment ownership and OEE provides a diagnostic view of availability, performance and quality.

Life-cycle costing, criticality analysis, quality data, governance and human judgment are essential to avoid both over-maintenance and under-maintenance. Digital monitoring and AI can improve prediction, but they require validated data, cybersecurity, explainability and human accountability. Effective asset management is ultimately a disciplined process of converting evidence into safe, timely and economically sound decisions.

Key Terms

Reliability: Probability that an asset performs its required function for a specified time and conditions.

Availability: Probability or proportion of time an asset is ready to perform its required function.

Maintainability: Probability that an asset can be restored within a specified time.

Failure rate: Conditional rate at which surviving items fail.

Hazard function: Another term for the instantaneous conditional failure rate.

MTTF: Mean time to failure for a non-repairable item.

MTBF: Mean operating time between failures for a repairable asset, subject to a stated definition.

MTTR: Mean active time required to repair an item.

Corrective maintenance: Maintenance performed after failure to restore function.

Preventive maintenance: Planned maintenance intended to reduce failure probability or consequences.

Predictive maintenance: Maintenance supported by measured condition or degradation forecasts.

Condition-based maintenance: Maintenance triggered by evidence of asset condition.

Reliability-centered maintenance: Functional and consequence-based method for selecting maintenance tasks.

Total productive maintenance: Organization-wide approach to improving equipment effectiveness and operator involvement.

Overall equipment effectiveness: Product of availability, performance and quality ratios.

Criticality: Relative importance of an asset based on failure likelihood and consequences.

Life-cycle cost: Total discounted cost of acquiring, operating, maintaining and disposing of an asset.

Failure mode: Specific way in which an asset can fail to perform a function.

Remaining useful life: Estimated time before an asset can no longer perform acceptably or requires intervention.

Review Questions

Short Answer Questions

1. Distinguish reliability from availability.
2. What are the three broad regions of the conceptual bathtub curve?
3. Why is MTBF an ambiguous measure unless its definition is specified?
4. State the three components of OEE.
5. Give two conditions under which run-to-failure may be appropriate.

Long Answer Questions

1. Explain how criticality, failure mechanism and detectability influence maintenance-policy selection.
2. Compare preventive, predictive and condition-based maintenance, including their limitations.
3. Explain the role of reliability-centered maintenance in an asset-management program.
4. Discuss how life-cycle cost analysis can change an apparently low-cost maintenance decision.
5. Evaluate the benefits and governance risks of using AI for predictive maintenance.

Analytical / Application Questions

1. An asset has MTBF of 500 hours and MTTR of 10 hours. Calculate its approximate inherent availability and explain one reason operational availability could be lower.
2. A plant proposes annual replacement of a component whose failures appear random and whose replacement does not reduce risk. Critically evaluate the proposal and recommend a data-based alternative.
3. Design a maintenance-policy mix for a bottleneck production machine containing a safety interlock, a low-cost cooling fan and a bearing showing measurable vibration degradation. Justify each policy using risk, failure behavior and life-cycle cost.

Multiple Choice Questions with Answers

Question 1

Which expression defines reliability at time t?

A. P(T ≤ t)

B. P(T > t)

C. MTTR ÷ MTBF

D. Downtime ÷ uptime

Correct Answer: B. P(T > t)

Question 2

An asset has MTBF of 240 hours and MTTR of 6 hours. Its approximate inherent availability is:

A. 90.00%

B. 94.00%

C. 97.56%

D. 102.50%

Correct Answer: C. 97.56%

Question 3

A Weibull shape parameter β greater than 1 generally indicates:

A. Decreasing hazard

B. Constant hazard

C. Increasing hazard

D. Zero failure probability

Correct Answer: C. Increasing hazard

Question 4

Which maintenance policy is most directly triggered by measured vibration crossing a decision threshold?

A. Condition-based maintenance

B. Unplanned corrective maintenance

C. Design-for-disposal maintenance

D. Administrative maintenance

Correct Answer: A. Condition-based maintenance

Question 5

The three factors multiplied in OEE are:

A. Reliability, maintainability and risk

B. Availability, performance and quality

C. Cost, speed and utilization

D. Safety, inventory and throughput

Correct Answer: B. Availability, performance and quality

Question 6

Which situation most strongly supports scheduled replacement?

A. Failure probability increases with age and replacement is economical

B. Failures are entirely random and replacement has no effect

C. The item is non-critical and instantly available

D. No failure criterion has been defined

Correct Answer: A. Failure probability increases with age and replacement is economical

Question 7

In the simple repairable-system model, operational availability is lower than inherent availability mainly because it includes:

A. Only product quality losses

B. Support and administrative delays

C. The discount rate

D. The Weibull shape parameter

Correct Answer: B. Support and administrative delays

Question 8

The principal purpose of reliability-centered maintenance is to:

A. Apply the same maintenance interval to every asset

B. Eliminate all corrective maintenance

C. Select tasks according to functions, failure modes and consequences

D. Maximize the number of preventive work orders

Correct Answer: C. Select tasks according to functions, failure modes and consequences

Question 9

Which is a leading maintenance indicator?

A. Last year’s production loss

B. Completed inspection compliance

C. Historical number of breakdowns

D. Past environmental incident count

Correct Answer: B. Completed inspection compliance

Question 10

A predictive-maintenance AI model should not be used without attention to:

A. Data quality and false-alarm consequences

B. The color of the equipment

C. The age of the organization’s logo

D. Whether all assets have identical failure modes

Correct Answer: A. Data quality and false-alarm consequences