Maintenance, Reliability and Asset Management

Introduction and context

Operations systems depend on assets: machines, buildings, vehicles, information systems, production tooling, utilities, software, and the infrastructure connecting them. An asset creates value only when it performs its required function at an acceptable level of safety, quality, cost, and environmental impact. Maintenance is therefore not merely a repair activity. It is a managerial and engineering discipline concerned with preserving and restoring function throughout an asset’s life cycle.

Contemporary operations managers must balance competing objectives. Excessive maintenance consumes labor, spares, and planned downtime; insufficient maintenance increases failures, safety exposure, quality losses, and customer disruption. The central challenge is to determine what should be maintained, when it should be maintained, how maintenance should be organized, and whether an asset should be modified, replaced, or retired.

Reliability, availability, and maintainability provide the analytical foundation. Asset management extends that foundation by integrating physical and digital assets with organizational strategy, risk, finance, compliance, and sustainability.

Learning objectives

By the end of this chapter, the reader should be able to:

1. Distinguish maintenance, reliability, maintainability, and availability.
2. Calculate and interpret basic reliability, failure-rate, maintainability, and availability measures.
3. Select among corrective, preventive, predictive, and prescriptive maintenance strategies.
4. Apply failure analysis tools such as FMEA, FTA, and root-cause analysis.
5. Explain total productive maintenance and the role of operators.
6. Evaluate replacement and life-cycle decisions using cost, risk, and performance evidence.
7. Manage physical and digital assets across their life cycles.
8. Design governance arrangements for data, safety, cybersecurity, and responsible use of AI in maintenance.

1. Foundational concepts

Maintenance and asset management

Maintenance comprises the technical, administrative, and managerial actions intended to retain an asset in, or restore it to, a state in which it can perform its required function. A required function includes both the primary output and specified performance limits. For example, a pump may be required to deliver 100 L/min at a defined pressure, not merely to rotate.

Reliability is the probability that an item performs its required function, under stated conditions, for a specified period. Reliability is therefore probabilistic and time-dependent.

Maintainability is the probability that a failed item can be restored to a specified condition within a stated time, using prescribed procedures and resources.

Availability is the ability of an item to be in a state to perform its required function when required. Availability combines failure behavior and restoration performance.

Asset management is the coordinated activity of an organization to realize value from assets. Value may include financial return, service, safety, resilience, compliance, social benefit, and environmental performance. It is broader than maintenance: maintenance preserves capability, while asset management decides how capability should be created, used, optimized, renewed, and disposed of.

A useful distinction is:

Failure and failure modes

A failure is the termination of an item’s ability to perform a required function. A failure mode describes how failure occurs, such as bearing seizure, seal leakage, software corruption, or loss of network connectivity. A failure cause is the underlying circumstance producing the mode, such as contamination, overload, poor installation, or an inadequate algorithm. A failure effect is the consequence for the system, users, safety, quality, or environment.

Failures may be:

• Functional failures, in which required performance is lost or falls outside limits.
• Potential failures, in which evidence indicates that functional failure is approaching.
• Evident failures, visible during normal operation or routine inspection.
• Hidden failures, not apparent until a protective or standby function is demanded.
• Random failures, whose probability is not strongly related to age in the observed period.
• Age-related failures, whose probability increases because of wear, fatigue, corrosion, or degradation.

The familiar “bathtub curve” describes three broad regions: early failures caused by design or installation defects, a relatively stable useful-life period, and a wear-out region. It is a model, not a universal law. Many electronic and software components do not exhibit a simple age-related pattern, while some mechanical components do.

2. Reliability and maintainability mathematics

Reliability function

Let T be the random time to failure. The reliability function is:

where R(t) is reliability at time t, P(T > t) is the probability of surviving beyond t, and F(t) is the cumulative distribution function of failure time. The expression applies to a defined population, operating environment, mission, and failure criterion.

If the failure rate is constant at λ failures per unit time, the exponential model gives:

The mean time to failure for this model is:

MTTF is normally used for non-repairable items. For repairable assets, mean time between failures (MTBF) is often used, but it must be defined carefully. In practice, MTBF may be calculated as operating time divided by the number of relevant failures. It is not automatically the same as the physical mean time to failure, especially when maintenance policies, operating conditions, and failure definitions differ.

The constant-rate model assumes statistically independent failures and a stable hazard rate. It is inappropriate when failure risk changes substantially with age or when common-cause failures dominate.

Hazard or failure rate

For a continuous failure-time distribution, the hazard rate is:

where f(t) is the failure-time probability density and R(t) is reliability. The hazard rate is a conditional rate, not necessarily a probability. Its units are inverse time.

A flexible model for age-related behavior is the Weibull distribution:

where β is the shape parameter and η is the characteristic life, the time at which approximately 63.2% of the population has failed. The associated hazard behavior is commonly interpreted as follows:

• β < 1: decreasing failure rate, often associated with early failures;
• β = 1: constant failure rate, equivalent to the exponential model;
• β > 1: increasing failure rate, often associated with wear-out.

The Weibull model is useful for fitting failure data, but parameter estimates depend on data quality, censoring treatment, sample size, and whether the selected failure population is homogeneous.

Maintainability and restoration time

Let Rₘ be the random repair time. Maintainability is:

For exponentially distributed repair times with repair rate μ per unit time:

The mean time to repair is:

This model assumes a stable repair process and does not automatically include waiting for parts, permits, access, specialist labor, or production release. For managerial decisions, elapsed restoration time is often more useful than hands-on wrench time.

Availability

For a repairable asset under steady-state conditions, with failure rate λ and restoration rate μ:

Using MTBF and MTTR, the same relationship is commonly written:

This is inherent or steady-state availability under assumptions such as statistically stable failure and repair processes and a single repairable unit. It excludes or treats differently such factors as logistics, preventive maintenance, administrative delay, and operating schedules.

Operational availability may be expressed as:

Here downtime should be defined consistently. If planned preventive-maintenance time, waiting time, and supply delays are included, operational availability will generally be lower than inherent availability.

Worked example: availability improvement

A packaging machine operates, on average, for 1,200 hours between failures. Average elapsed restoration time is 8 hours.

Thus steady-state availability is approximately 99.34%, subject to the model assumptions. If better parts management reduces restoration time to 4 hours while MTBF remains unchanged:

Availability rises to approximately 99.67%. The example demonstrates that reducing restoration time can produce a material gain even when the failure frequency does not change.

Series and parallel systems

For independent components arranged in series, the system works only when every component works:

For two independent components in parallel, the system works when at least one works:

These formulas require correct system logic and independence assumptions. Redundancy may improve reliability, but common-cause failures, shared power, shared software, and poor maintenance can eliminate the expected benefit.

3. Maintenance strategies and policy selection

Corrective maintenance

Corrective maintenance restores an asset after a failure. It is reasonable when the consequence of failure is low, repair is rapid, spares are inexpensive, and preventive intervention provides little benefit. It is inappropriate where failure creates unacceptable safety, environmental, legal, quality, or service consequences.

Run-to-failure is not the same as neglect. A managed run-to-failure policy specifies failure consequences, safe isolation, spares, response time, repair procedures, and escalation rules.

Preventive maintenance

Preventive maintenance is performed at predetermined intervals or according to prescribed criteria. It includes time-based replacement, inspection, lubrication, calibration, cleaning, and functional testing.

Time-based replacement is most defensible when failure probability increases predictably with age and replacement can be performed before the risk becomes unacceptable. Replacing an item simply because a calendar interval exists may introduce infant mortality through installation errors and may discard useful life.

A basic preventive policy can be evaluated through total cost, including planned intervention, corrective repair, downtime, risk, and quality consequences. The lowest maintenance invoice is not necessarily the lowest life-cycle cost.

Predictive and condition-based maintenance

Predictive maintenance uses measurements of actual condition to estimate degradation and recommend action. Typical techniques include vibration analysis, thermography, oil analysis, ultrasonic inspection, motor-current analysis, acoustic emission, process-performance monitoring, and software-health telemetry.

A condition indicator is useful only when it is linked to a failure mechanism and an intervention decision. A sensor that generates alerts without a response process creates alarm fatigue and data cost rather than reliability.

The P–F interval is the period between the point at which a potential failure becomes detectable (P) and functional failure (F). An inspection interval should be sufficiently shorter than the P–F interval to provide a realistic opportunity for planning and intervention. The exact interval depends on detectability, measurement uncertainty, consequence, and required response time.

Prescriptive maintenance

Prescriptive maintenance combines condition information, failure models, operational constraints, and optimization to recommend an action. It may advise continued operation, inspection, derating, repair, or replacement. Prescriptive systems should expose assumptions and confidence, retain human accountability, and prevent automated actions from violating safety or regulatory controls.

Selecting a strategy

A practical policy-selection process is:

1. Define the required function and performance standard.
2. Identify failure modes and consequences.
3. Determine whether a credible age or condition relationship exists.
4. Assess detectability and the available P–F interval.
5. Compare intervention cost with failure cost and risk.
6. Check technical feasibility, skills, spares, and access.
7. Pilot the policy and evaluate actual outcomes.

Reliability-centered maintenance (RCM) formalizes this logic. It asks what the asset must do, how it can fail, what causes each failure, what happens when it fails, and which task is technically feasible and worth performing. RCM does not mean “maintain everything more frequently”; it means selecting tasks according to function, failure mechanism, and consequence.

4. Failure analysis and improvement

FMEA and FMECA

Failure Modes and Effects Analysis (FMEA) is a structured bottom-up method. For each component or process step, the team records the function, failure mode, effect, cause, existing controls, detection method, and recommended action. FMECA adds explicit criticality analysis.

Traditional risk priority numbers multiply ordinal scores for severity, occurrence, and detection. Such multiplication can be misleading because ordinal scales do not necessarily have equal intervals, and different score combinations can produce the same number while representing different risks. Modern practice should use the score as a discussion aid, supplementing it with consequence-based analysis, evidence, and explicit escalation criteria.

Fault-tree analysis

Fault-tree analysis (FTA) is a top-down deductive method. It starts with a defined undesired top event and uses logical AND and OR relationships to identify combinations of causes. FTA is particularly useful for safety and system-level analysis. It requires a precisely defined top event and careful treatment of common-cause failures.

Root-cause analysis

Root-cause analysis (RCA) investigates why an event occurred and why organizational controls failed to prevent recurrence. Tools include causal trees, the five-whys technique, barrier analysis, and fishbone diagrams. RCA should avoid stopping at “operator error.” Human action is often influenced by design, training, workload, procedures, incentives, interfaces, and management systems.

Effective corrective action changes the system: redesigning a component, improving contamination control, changing a procedure, adding an interlock, revising procurement specifications, or strengthening competency assurance. Re-training alone is rarely sufficient when the underlying design remains unchanged.

5. Total productive maintenance

Total Productive Maintenance (TPM) is an organization-wide approach to maximizing equipment effectiveness through employee involvement, operator care, planned maintenance, focused improvement, quality maintenance, early equipment management, training, safety, and administrative improvement.

A commonly used performance measure is Overall Equipment Effectiveness (OEE):

Availability is the proportion of planned production time during which the equipment is operating; performance compares actual speed with ideal speed; quality is the proportion of good output. All three factors must be defined consistently. OEE is diagnostic, not a universal measure of business productivity. Maximizing OEE on one bottleneck may be valuable, while maximizing it on a non-bottleneck may merely create excess inventory.

For example, if availability is 0.90, performance is 0.95, and quality is 0.98:

OEE is approximately 83.79%. The decomposition identifies improvement opportunities more clearly than a single uptime percentage.

Autonomous maintenance enables operators to perform basic cleaning, inspection, lubrication, tightening, and abnormality identification. It does not transfer specialist work without training or alter authorization requirements. TPM succeeds when production and maintenance share asset ownership rather than competing for short-term output.

6. Replacement and life-cycle decisions

Replacement decisions should consider performance, risk, technology, energy, maintainability, residual value, and strategic capability—not only the age of the existing asset.

For a uniform annual cost comparison, the equivalent annual cost (EAC) of an asset can be expressed as:

where P is purchase price at time zero, i is the annual discount rate expressed as a decimal, n is economic life in years, Cₐ is equivalent annual operating and maintenance cost, and S is estimated salvage value at the end of year n. The capital-recovery factor and salvage adjustment assume annual cash flows and a constant discount rate. Taxes, inflation, financing, and risk may require a more comprehensive discounted-cash-flow model.

A replacement analysis should also include expected failure cost. If failure probability in a period is p and consequence cost is C, expected failure cost is:

This is useful for comparing alternatives but does not make catastrophic risks acceptable merely because their expected monetary value is low. Safety and legal constraints may require elimination or reduction regardless of expected value.

Life-cycle cost includes acquisition, installation, energy, labor, maintenance, downtime, upgrades, cybersecurity, decommissioning, disposal, and residual value. A cheaper purchase can be more expensive over the asset’s service life.

7. Physical and digital asset life-cycle management

An asset life cycle normally includes need definition, concept and design, acquisition, commissioning, operation, maintenance, modification, renewal, and disposal. Decisions made during design strongly influence later reliability. Design-for-maintainability includes access, modularity, standard fasteners, isolation points, lifting provisions, diagnostics, documentation, and safe work procedures.

Digital assets—software, control systems, data platforms, models, sensors, and industrial networks—require additional controls. Their degradation may appear as obsolete dependencies, unsupported operating systems, corrupted data, configuration drift, cyber vulnerabilities, or loss of interoperability rather than mechanical wear.

Important digital asset practices include:

• maintaining an accurate configuration and asset register;
• assigning ownership and classification;
• controlling versions, patches, interfaces, and credentials;
• backing up and testing restoration;
• separating critical operational networks where appropriate;
• monitoring data quality and model drift;
• planning technology refresh before vendor support ends;
• preserving audit trails for safety-relevant decisions.

Artificial intelligence can assist anomaly detection, remaining-useful-life estimation, work-order prioritization, spare-parts forecasting, and inspection image analysis. However, predictive performance can deteriorate when operating regimes change or failure data are sparse. Models should be validated against realistic baseline policies, monitored for false alarms and missed failures, and governed by accountable professionals. Explainability is especially important where an automated recommendation affects safety, employment, or regulatory compliance.

8. Governance, safety, and performance management

A mature maintenance system integrates maintenance, operations, engineering, procurement, finance, safety, quality, information technology, and supply-chain functions. Governance should define decision rights, escalation thresholds, data ownership, competence requirements, and audit arrangements.

Safety-critical maintenance requires approved procedures, isolation and lockout controls, permit-to-work systems, competency assurance, independent verification where necessary, and management of change. Maintenance creates risk as well as reducing it: incorrect reassembly, bypassed safeguards, poor configuration control, and rushed testing can introduce new hazards.

Useful performance indicators include:

A metric should lead to a decision or improvement action. Excessive measurement can create administrative work and encourage gaming. Managers should examine trends, distributions, criticality, and causal evidence rather than relying on a single target.

9. Illustrative case: reliability improvement in a beverage plant

Background

The following fictional case is illustrative. A beverage plant operates a high-speed filling line whose conveyor drive causes repeated unplanned stops. The line is a bottleneck, and a stop interrupts filling, labeling, and downstream packing.

Problem

The maintenance department reports an average MTBF of 240 operating hours and an average elapsed restoration time of 6 hours. Operators also report frequent abnormal vibration before failure, but inspection results are not recorded consistently. The plant has been replacing drive bearings every six months, although several bearings fail earlier and others are removed in apparently good condition.

Analysis

A cross-functional team defines the functional failure as inability to maintain the required conveyor speed and alignment. FMEA identifies bearing contamination, misalignment, excessive belt tension, and inadequate installation as principal causes. Review of work orders shows that “bearing failure” is often recorded without distinguishing lubrication loss from misalignment.

The team estimates current steady-state availability as:

Availability is therefore approximately 97.56% under the stated assumptions. The business impact is greater than this number suggests because failures occur at the bottleneck during high-demand periods.

Approach and solution

The plant adopts a layered policy. Operators perform daily cleaning and visual checks, including a defined vibration and temperature abnormality check. Maintenance technicians conduct monthly vibration measurements and verify alignment during planned stops. Bearings are no longer replaced solely by calendar age; replacement is triggered by condition evidence or a defined inspection finding. Procurement changes the specification for seals and contamination protection. Work orders require the technician to record failure mode and causal evidence.

A small number of sensors are installed at critical drive locations. The alert process specifies who reviews the signal, how quickly the asset must be inspected, and what operating restrictions apply. The team also revises installation tooling and trains technicians in alignment and bearing fitting.

Outcome

After six months, the plant observes fewer repeat failures, longer bearing life, and a reduction in emergency work. The most important improvement is not the sensor itself but the connection between condition data, competent diagnosis, planned intervention, and failure-mode coding. The team continues to monitor false alarms, missed detections, production effects, and total cost before expanding the program.

Learning points

1. A maintenance interval should reflect a failure mechanism, not habit.
2. Data quality and failure-mode definitions are prerequisites for analytics.
3. Bottleneck criticality changes the economic importance of a failure.
4. Predictive maintenance requires a response process, not only instrumentation.
5. Design, procurement, installation, and operations can be more influential than repair speed alone.

Challenges and limitations

Maintenance analysis is limited by incomplete failure histories, inconsistent definitions, censored data, changing operating conditions, and survivorship bias. A high MTBF may reflect under-reporting or a failure definition that excludes degraded performance. A low MTTR may be achieved by excluding waiting, diagnosis, and restart time. Reliability models can create false confidence when assumptions are not tested.

Organizations also face skills shortages, aging infrastructure, fragmented ownership, spare-parts obsolescence, cyber risk, and pressure to defer maintenance. Outsourcing can provide expertise but may weaken internal knowledge if contracts specify tasks rather than outcomes. Sustainability introduces further trade-offs: extending asset life may avoid embodied carbon, while newer equipment may reduce energy use and hazardous-material exposure.

Good practice is consequently evidence-based and risk-informed. It combines engineering judgment, statistical analysis, operator knowledge, financial evaluation, safety management, and continuous learning.

Chapter Summary

Maintenance preserves or restores asset function, while asset management coordinates decisions to realize value over the entire life cycle. Reliability measures the probability of performing a required function; maintainability concerns restoration; availability combines operating and repair behavior. Exponential and Weibull models are useful only when their assumptions fit the failure process. Series and parallel reliability models require valid system logic and independence assumptions.

Corrective, preventive, predictive, and prescriptive strategies should be selected according to failure mechanisms, consequences, detectability, cost, and risk. RCM, FMEA, FTA, and RCA provide complementary methods for analyzing failure and designing interventions. TPM integrates operators, maintenance, engineering, and management, while OEE decomposes equipment effectiveness into availability, performance, and quality.

Replacement decisions should use life-cycle cost, risk, service requirements, and strategic considerations. Physical and digital assets require configuration control, obsolescence planning, cybersecurity, data governance, and responsible use of AI. Mature systems emphasize clear definitions, competent work, safe controls, quality data, cross-functional ownership, and measures that support decisions rather than encourage local optimization.

Key Terms

Reliability: Probability that an item performs its required function for a specified time under stated conditions.

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

Availability: Ability of an item to be ready to perform its required function when needed.

Failure: Termination of the ability to perform a required function.

Failure mode: The manner in which an item fails.

Failure cause: The underlying reason a failure mode occurs.

Failure effect: The consequence of a failure mode.

Corrective maintenance: Maintenance performed after a failure to restore function.

Preventive maintenance: Maintenance performed at predetermined intervals or criteria.

Predictive maintenance: Maintenance based on measured condition and degradation evidence.

Prescriptive maintenance: Data-informed maintenance that recommends an action.

MTBF: Mean time between relevant failures of a repairable asset.

MTTR: Mean time to repair or restore, according to the stated time boundary.

Hazard rate: Conditional failure rate at a given time.

P–F interval: Time between detectable potential failure and functional failure.

FMEA: Bottom-up analysis of failure modes, effects, causes, and controls.

Fault-tree analysis: Top-down logical analysis of combinations causing a top event.

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

OEE: Product of availability, performance, and quality.

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

Review Questions

5 Short Answer Questions

1. Define reliability, maintainability, and availability, and explain how they differ.
2. What assumptions underlie the exponential reliability model?
3. Explain the difference between a failure mode, failure cause, and failure effect.
4. What is the P–F interval, and why does it matter for inspection planning?
5. Why is OEE more informative when its three components are examined separately?

5 Long Answer Questions

1. Compare corrective, preventive, predictive, and prescriptive maintenance. Identify conditions under which each is appropriate.
2. Explain how RCM, FMEA, FTA, and RCA complement one another in a maintenance-improvement program.
3. Discuss the assumptions and limitations of MTBF, MTTR, and availability as managerial performance measures.
4. Develop a life-cycle asset-management policy for a critical production system, including acquisition, operation, maintenance, renewal, and disposal.
5. Evaluate the opportunities and governance risks associated with using AI for predictive maintenance.

3 Analytical / Application Questions

1. An asset has MTBF of 500 hours and MTTR of 10 hours. Calculate steady-state availability. Then explain two operational factors that could make actual availability lower.
2. A plant is considering calendar-based bearing replacement, condition-based replacement, and run-to-failure. Construct a decision analysis using failure consequence, age dependence, detectability, intervention cost, and data requirements.
3. A safety-critical pump has redundant units, but both share a power supply and control software. Analyze why a simple parallel reliability calculation may overstate system reliability and recommend additional analysis.

Multiple Choice Questions with Answers

Question 1

Which measure is most directly concerned with the probability that an asset can be restored within a specified time?

A. Reliability

B. Maintainability

C. Availability

D. Criticality

Correct Answer: B. Maintainability

Question 2

Under the constant failure-rate model, which expression represents reliability at time t?

A. R(t) = 1 − e^(−λt)

B. R(t) = e^(−λt)

C. R(t) = λt

D. R(t) = 1 ÷ (λ + t)

Correct Answer: B. R(t) = e^(−λt)

Question 3

An asset has MTBF of 900 hours and MTTR of 9 hours. Its steady-state availability is closest to:

A. 90.0%

B. 98.0%

C. 99.0%

D. 99.9%

Correct Answer: C. 99.0%

Question 4

A Weibull shape parameter β greater than 1 is commonly interpreted as indicating:

A. A decreasing failure rate

B. A constant failure rate

C. An increasing failure rate

D. No measurable failure process

Correct Answer: C. An increasing failure rate

Question 5

Which maintenance approach uses measured asset condition to determine when intervention is required?

A. Run-to-failure maintenance

B. Calendar-only replacement

C. Predictive maintenance

D. Emergency procurement

Correct Answer: C. Predictive maintenance

Question 6

In FMEA, a failure mode describes:

A. The financial value of an asset

B. The manner in which an item fails

C. The annual maintenance budget

D. The organizational reporting structure

Correct Answer: B. The manner in which an item fails

Question 7

Which statement best describes the P–F interval?

A. The time between purchasing and commissioning an asset

B. The time between detectable potential failure and functional failure

C. The time between two unrelated assets failing

D. The time required to calculate OEE

Correct Answer: B. The time between detectable potential failure and functional failure

Question 8

If availability is 0.92, performance is 0.95, and quality is 0.98, OEE is closest to:

A. 85.6%

B. 87.5%

C. 89.7%

D. 95.0%

Correct Answer: A. 85.6%

Question 9

Which is a central principle of total productive maintenance?

A. Maintenance responsibility belongs exclusively to specialists

B. Operators can contribute to routine care and abnormality detection

C. All assets should be replaced at the same age

D. Production output should always override safety controls

Correct Answer: B. Operators can contribute to routine care and abnormality detection

Question 10

Why can a simple parallel reliability calculation overstate the reliability of redundant assets?

A. Parallel systems cannot contain more than one asset

B. Redundant assets may share common-cause failures

C. Reliability is unrelated to system configuration

D. Repair time is always zero in parallel systems

Correct Answer: B. Redundant assets may share common-cause failures