Asset Criticality and Maintenance Strategy Selection

Introduction

Maintenance decisions should begin with the asset’s business and risk significance, not with a preference for a particular maintenance technology. A sophisticated predictive-monitoring system is wasteful when applied to a non-critical, easily replaceable component; conversely, run-to-failure is unacceptable for an asset whose failure can injure people, violate environmental limits, or stop a bottleneck process.

Asset criticality is the structured evaluation of the consequences and likelihood of asset failure. Maintenance strategy selection is the subsequent process of choosing how failure will be prevented, detected, tolerated, or eliminated. Together, these activities connect reliability engineering with production management, safety, quality, finance, and regulatory governance.

The central principle is that maintenance should be risk-based and economically defensible. The selected strategy must match the failure mechanism, the detectability of deterioration, the consequences of failure, and the organization’s ability to execute the required work.

Learning Objectives

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

1. Classify production assets according to function, system position, failure consequence, and maintainability.
2. Distinguish asset criticality from asset condition and maintenance priority.
3. Conduct a qualitative and semi-quantitative criticality and risk assessment.
4. Select run-to-failure, preventive, condition-based, predictive, redesign, or mixed strategies.
5. Evaluate strategy choices using reliability, risk, life-cycle cost, and operational criteria.
6. Recognize the limitations, governance requirements, and ethical implications of maintenance decisions.

1. Asset Classification as the Foundation of Maintenance Planning

An asset is an item, system, or network that provides value to an organization. In a factory, assets include machines, utilities, control systems, software, buildings, tooling, material-handling equipment, and safety systems. A useful asset hierarchy is:

Classification should capture more than equipment type. Each asset record should normally include its function, required performance, operating context, redundancy, duty cycle, energy use, safety role, failure modes, spare-parts requirements, and ownership.

Functional and physical classification

A physical classification groups assets by what they are: pumps, motors, compressors, robots, conveyors, or programmable logic controllers. A functional classification groups them by what they do: transfer material, control temperature, protect personnel, maintain product quality, or provide emergency service. Functional classification is more useful for criticality analysis because two identical pumps may have very different consequences if one serves a non-bottleneck wash station and the other supplies cooling water to a reactor.

Production role

Assets may be classified as:

• Bottleneck assets, whose loss restricts total system output.
• Non-bottleneck assets, where alternative capacity exists.
• Quality-critical assets, whose performance directly affects specification compliance.
• Safety-critical assets, whose failure may cause injury or loss of protective function.
• Environmental-critical assets, whose failure may cause emissions, discharge, or contamination.
• Support or utility assets, such as power, steam, cooling, compressed air, and information systems.
• Redundant assets, where standby capacity can perform the required function.

Redundancy reduces production consequence but does not automatically eliminate safety or environmental criticality. A standby safety system may itself be subject to statutory testing and hidden failure.

2. Criticality, Risk, Priority, and Condition

These terms are related but not interchangeable.

Criticality is the relative importance of an asset’s functional failure to organizational objectives. It is generally assessed before a particular failure occurs and changes when the process, product, regulation, or redundancy changes.

Risk combines the consequence of an unwanted event with the likelihood of that event. A common conceptual expression is:

The expression is a decision model, not a universal physical law. Consequence and likelihood must be defined on compatible scales, and the result should not be interpreted as precise probability unless the scales are quantitatively calibrated.

Maintenance priority is the urgency with which a work order should be executed. A low-criticality asset with an active leak may require immediate attention, while a high-criticality asset operating normally may require planned surveillance rather than emergency work.

Condition describes the present state of an asset: healthy, degraded, or failed. A critical asset may currently be in good condition; a low-criticality asset may be degraded. Thus, criticality determines the importance of controlling failure, while condition helps determine when action is required.

3. Failure Consequences and Failure Modes

A strategy cannot be selected intelligently from the asset name alone. Analysis must identify how the asset performs its function and how that function can be lost.

A functional failure is the inability to perform a required function to the specified standard. A failure mode is the specific physical or operational cause, such as bearing fatigue, seal leakage, software error, blocked filter, corrosion, loss of calibration, or operator-interface error. A failure effect describes what is observed locally and at the process level. A failure consequence describes why the effect matters.

The major consequence categories are:

1. Safety: injury, fatality, exposure, or loss of a protective barrier.
2. Environmental: release, emission, contamination, or breach of permit conditions.
3. Quality: nonconforming product, contamination, traceability loss, or customer risk.
4. Production: lost throughput, bottleneck interruption, or schedule disruption.
5. Financial: repair cost, overtime, scrap, energy loss, contractual penalty, and revenue loss.
6. Reputation and compliance: legal action, license conditions, or loss of stakeholder confidence.

A failure mode and effects analysis (FMEA) provides a structured way to connect function, failure mode, effect, cause, controls, and recommended action. A fault-tree analysis works backward from an undesired top event to combinations of causes using logical AND and OR relationships. FMEA is often bottom-up; fault-tree analysis is top-down. They are complementary rather than competing methods.

4. Criticality and Risk Assessment Methods

4.1 Qualitative screening

A first-pass screening classifies assets as high, medium, or low criticality using expert judgment. This is efficient for large asset populations but should use documented definitions. For example, a high-criticality asset may be one whose failure creates unacceptable safety consequences, stops a bottleneck for more than one shift, or causes a reportable environmental event.

Qualitative screening is appropriate when data are sparse, but it should be periodically challenged by operations, maintenance, engineering, quality, safety, and finance personnel. Sole reliance on the maintenance department can bias the analysis toward repair effort rather than business consequence.

4.2 Semi-quantitative risk scoring

A common method assigns scores to consequence and likelihood. For example:

where R is the risk score, C is a consequence score, and L is a likelihood score. If each is rated from 1 to 5, the score ranges from 1 to 25. A possible interpretation is 1–4 low, 5–9 moderate, 10–16 high, and 17–25 very high. These boundaries are organizational conventions and must be approved rather than assumed to be universally valid.

Likelihood may be estimated from historical failure frequency, reliability models, expert assessment, or a combination. If a failure rate is approximately constant and failures follow an exponential model, the probability of at least one failure during time interval t is:

Here, T is time to failure, λ is a constant failure rate with units of inverse time, and t is the exposure interval in the same time unit. The model assumes a constant hazard rate and statistical independence appropriate to the modeled population. It is unsuitable when infant mortality, wear-out, changing load, or strong preventive-maintenance effects dominate.

For example, if λ = 0.002 failures per operating hour and the interval is 100 operating hours:

The estimated probability of failure is approximately 18.1 percent under the stated assumptions. This is not a guarantee for one individual machine; it is a model-based estimate.

4.3 Expected loss and life-cycle reasoning

When monetary estimates are credible, expected loss can support prioritization:

If failure probability is 0.10 over a planning period and the consequence cost is $40,000, expected loss is $4,000 for that period. This calculation excludes the cost of risk-reduction actions and should not be used to monetize irreversible harm in a way that permits unsafe operation. Safety and legal constraints remain mandatory constraints, not merely financial trade-offs.

Life-cycle cost includes acquisition, installation, energy, operation, maintenance, downtime, modification, and disposal costs. A simple discounted cash-flow representation is:

where C₀ is initial cost, Cₜ is cost in period t, i is the discount rate per period, and the summation covers the asset life. The terms must use the same currency basis and time period. This model is useful for comparing strategies but is sensitive to discount rate, inflation treatment, residual value, and uncertain downtime costs.

5. Selecting a Maintenance Strategy

Maintenance strategy should be selected by failure mode, not simply by equipment class. One pump may need vibration monitoring, another may be run to failure, and a third may require redesign because repeated seal failures are caused by poor pipe alignment.

5.1 Run-to-failure or corrective maintenance

Run-to-failure permits an item to operate until functional failure, followed by repair or replacement. It is appropriate when:

• failure has negligible safety, environmental, quality, and production consequence;
• the item is inexpensive and readily replaceable;
• failure is readily detectable;
• repair time and spare availability are acceptable; and
• preventive or predictive tasks cost more than the risk they reduce.

Run-to-failure is not equivalent to neglect. It requires a defined spare, replacement method, safe isolation procedure, and response plan. It is unsuitable for hidden failures in protective systems.

5.2 Time- or usage-based preventive maintenance

Preventive maintenance is performed at predetermined intervals or usage levels. Examples include lubrication, filter replacement, calibration, inspection, overhaul, and scheduled replacement.

Time-based replacement is justified when failure probability increases predictably with age or when regulations prescribe an interval. It is less effective for random failures, where age does not meaningfully predict failure. Excessive interval-based work can introduce infant mortality through maintenance-induced defects, consume scarce technicians, and replace healthy components unnecessarily.

5.3 Condition-based maintenance

Condition-based maintenance initiates action when a measurable condition indicates deterioration beyond a defined threshold. Examples include temperature, pressure, leakage, oil contamination, electrical resistance, alignment, and visual damage.

A condition indicator should have a credible relationship to the failure mode and provide sufficient warning time. A threshold should be linked to a decision: continue operation, reduce load, schedule repair, or stop immediately. Data collection without a decision rule is monitoring, not effective condition-based maintenance.

5.4 Predictive maintenance

Predictive maintenance uses condition data, historical behavior, statistical models, or machine-learning methods to estimate future degradation or failure risk. It is a form of condition-based maintenance when its output supports action before functional failure.

Useful methods include vibration analysis for rotating equipment, thermography for electrical and mechanical anomalies, oil analysis, motor-current signature analysis, acoustic monitoring, process-performance analysis, and model-based prognostics. Predictive maintenance is valuable when failure is detectable in advance, the consequence justifies monitoring cost, and the organization can respond within the available warning interval.

A model’s accuracy alone does not establish business value. False alarms create unnecessary work; missed detections create false confidence. Performance should be evaluated using operational measures such as avoided downtime, precision of alerts, lead time, maintenance-induced failures, and total cost.

5.5 Functional testing and failure-finding

Some failures are hidden because the item is normally dormant, such as an emergency generator, relief device, alarm, or standby pump. A condition sensor may not reveal loss of function. Periodic functional testing is therefore required to expose hidden failures. The test interval should consider required availability, test effectiveness, degradation during testing, and applicable regulations.

5.6 Redesign and modification

Redesign is appropriate when a failure is unacceptable and existing maintenance tasks cannot control it effectively. Options include changing materials, improving lubrication, modifying operating conditions, adding redundancy, improving access, eliminating contamination, correcting alignment, upgrading controls, or changing the process itself.

Redesign should address the physical cause rather than repeatedly treating symptoms. However, modifications can introduce new hazards, cybersecurity exposure, training requirements, and spare-parts complexity. Management of change should therefore be applied.

6. A Structured Strategy-Selection Decision Process

A practical decision sequence is:

1. Define the required function and performance standard.
2. Identify functional failures and credible failure modes.
3. Determine safety, environmental, quality, production, and financial consequences.
4. Identify whether the failure is evident or hidden.
5. Assess whether an age or usage relationship exists.
6. Determine whether deterioration is detectable and whether warning time exceeds planning and repair time.
7. Compare task cost, risk-reduction value, and life-cycle effects.
8. Select a task, redesign, redundancy, or justified run-to-failure approach.
9. Specify the decision rule, interval, owner, data quality, and escalation path.
10. Review results and revise the strategy using evidence.

A simplified decision logic is shown below.

Figure: A simplified logic for matching failure characteristics to maintenance actions.

The flow is deliberately simplified. Real decisions may combine strategies: a critical motor may receive vibration monitoring, lubrication control, thermography, operator rounds, and a planned overhaul of a particular subsystem.

7. Worked Application: Packaging-Line Cooling Pump

The following is an illustrative case, not a report of an actual company.

Background

A packaging plant operates a cooling-water system for a high-speed sealing line. Two pumps are installed, but the standby pump is started manually and its availability is uncertain. The duty pump has experienced three mechanical-seal failures in twelve months.

Problem

Each failure causes approximately six hours of reduced production and creates a leakage hazard. The maintenance team proposes replacing seals every three months. Operations argues for installing vibration sensors. Engineering suspects pipe strain and poor alignment.

Analysis

The first step is to distinguish the failure mode from the asset label. The suspected failure modes are seal wear, shaft misalignment, pipe strain, bearing deterioration, and inadequate standby readiness. Production consequence is high because the line is a bottleneck. Leakage consequence may also be environmental or safety-significant depending on fluid temperature and treatment chemicals.

The three failures do not prove that a three-month age interval is optimal. They suggest a recurring cause, but the data are too limited to establish a reliable age-based replacement relationship. A seal replacement task may reduce symptoms while leaving the initiating cause unchanged.

A cross-functional review should verify:

• pump operating point and cavitation conditions;
• suction conditions, vibration, alignment, and soft-foot condition;
• pipe support and nozzle loading;
• seal material compatibility and installation quality;
• standby-pump automatic-start function;
• leak detection and isolation arrangements; and
• actual production and safety consequences.

Solution

The selected package of controls is:

1. Correct alignment and pipe strain; verify the pump operates near its intended duty point.
2. Add post-maintenance alignment and installation checks.
3. Install vibration and seal-leak monitoring only after confirming that the signals provide useful warning.
4. Establish operator rounds for leakage, pressure, temperature, and abnormal noise.
5. Function-test automatic standby start at a defined interval and record results.
6. Hold a critical seal and bearing spare because response time is consequential.
7. Do not adopt routine three-month seal replacement unless evidence demonstrates an age-related benefit.
8. Review failure data after implementation and revise the risk assessment.

Outcome measures

Success should be measured through seal-failure frequency, unplanned downtime, leak events, standby availability, false alarms, maintenance hours, and total cost. A reduction in failures without an increase in maintenance-induced defects is stronger evidence than sensor installation alone.

Learning points

The case demonstrates that criticality is not a justification for indiscriminate preventive replacement. A high-consequence asset requires disciplined failure analysis, verification of protective functions, and elimination of physical causes. Predictive technology is an enabling control, not a substitute for engineering fundamentals.

8. Implementation, Governance, and Data Quality

A criticality program becomes effective only when connected to the computerized maintenance management system (CMMS), work processes, materials planning, and performance review. Each selected task should specify the failure mode addressed, task instructions, acceptance criteria, interval or trigger, responsible role, required competence, safety controls, and escalation condition.

Data quality is a major limitation. Failure codes may be vague, operating hours may be missing, and corrective work may be recorded as preventive work. Before applying advanced analytics, organizations should establish consistent asset hierarchies, failure-mode taxonomies, timestamp conventions, and feedback from completed work.

Governance should include:

• documented scoring definitions and approval authority;
• periodic reassessment after process or product changes;
• management of change for equipment modifications;
• independent review of safety-critical assumptions;
• cybersecurity controls for connected monitoring devices;
• protection of personnel and contractor data; and
• auditability of decisions and overrides.

AI-based maintenance recommendations require particular caution. Models may reproduce biased historical decisions, confuse correlation with causation, or fail when operating conditions change. Human review remains necessary for safety-critical decisions. The organization should be able to explain what data supported an alert, what uncertainty exists, and who authorized the resulting action.

9. Performance Measures and Continuous Improvement

Useful measures should balance reliability, work quality, cost, and risk. Examples include:

Mean time between failures (MTBF) is commonly computed for repairable items as operating time divided by the number of observed failures. It is a descriptive measure, not automatically a physical constant and not necessarily the reciprocal of a meaningful failure rate under changing conditions. Comparisons should use consistent operating context and failure definitions.

The maintenance program should use a closed learning loop: plan, execute, inspect results, analyze deviations, update failure knowledge, and revise the strategy. Reliability-centered maintenance formalizes this logic by asking what functions matter, how they fail, what consequences arise, and which task is technically feasible and worth performing.

Challenges and Limitations

Criticality scoring can create false precision when ordinal scores are multiplied as though they were measured probabilities. Expert judgment can be affected by recency bias, departmental interests, and incomplete failure histories. Production pressure may encourage deferral of important tasks, while excessive risk aversion may generate unnecessary work.

Condition monitoring also has limitations: sensors can fail, thresholds can drift, data may be noisy, and a detectable signal may not provide enough warning for intervention. Predictive models may degrade after product mix, speed, materials, or environmental conditions change. Redundancy can reduce production risk but increase common-cause vulnerability and maintenance burden.

The correct response is not to abandon structured analysis but to state assumptions, preserve uncertainty, validate decisions in operation, and revisit them when evidence changes.

Contemporary Developments

Digital twins, edge computing, wireless sensors, and machine learning are expanding the scale and speed of maintenance analysis. Their strongest applications combine physics-based understanding with data-driven detection. Digital work instructions and augmented-reality support can improve execution, but they do not replace competence or isolation procedures.

Sustainability is also changing strategy selection. Extending component life, reducing unnecessary replacement, lowering energy consumption, and designing for repair can reduce material and carbon impacts. Conversely, additional sensors, batteries, network equipment, and computing have their own environmental footprints. Sustainable maintenance therefore requires life-cycle evaluation rather than assuming that more monitoring is always better.

Best-Practice Checklist

A mature organization should be able to answer yes to the following questions:

• Is every significant asset linked to a defined function?
• Are safety, environmental, quality, and production consequences assessed separately?
• Are failure modes and hidden failures documented?
• Is the selected task technically capable of managing the failure mode?
• Are warning time, response time, and spare availability compatible?
• Are task intervals and thresholds supported by evidence?
• Are safety-critical and regulatory tasks protected from informal deferral?
• Are maintenance-induced failures monitored?
• Are decisions reviewed after changes and actual failures?
• Can an auditor trace the reasoning from asset function to strategy?

Chapter Summary

Asset criticality expresses how important an asset’s functional failure is to organizational objectives; it is distinct from current condition and work priority. Effective assessment begins with asset hierarchy, function, operating context, failure modes, and consequences across safety, environment, quality, production, finance, and compliance.

Risk can be screened qualitatively or estimated using calibrated semi-quantitative and quantitative methods. Risk scores are decision aids, not precise truths. Reliability models and expected-loss calculations are useful only when their assumptions, data quality, and units are appropriate.

Run-to-failure is justified for low-consequence, replaceable items with acceptable response arrangements. Preventive maintenance is most defensible when age or usage relates to failure or when a standard is mandatory. Condition-based and predictive maintenance require detectable deterioration, useful warning time, reliable data, and an executable response. Hidden failures require functional testing, while recurring or unacceptable failures may require redesign.

The best strategy is often a combination of engineering correction, operator care, monitoring, proof testing, spares, and planned intervention. Governance, human judgment, management of change, cybersecurity, and continuous feedback are essential, especially when digital and AI technologies are used.

Key Terms

Asset: An item or system that provides value through a required function.

Asset hierarchy: Structured levels from site and area to system, maintainable item, and component.

Criticality: Relative importance of an asset’s functional failure.

Risk: Combination of consequence and likelihood for an undesired event.

Functional failure: Inability to perform a required function to its specified standard.

Failure mode: Specific physical or operational cause of functional failure.

Failure effect: Local or process-level result of a failure mode.

Failure consequence: Safety, environmental, quality, production, financial, or compliance significance of a failure.

FMEA: Structured analysis linking functions, failure modes, effects, causes, controls, and actions.

Fault-tree analysis: Top-down logical analysis of causes leading to an undesired top event.

Run-to-failure: Planned operation until functional failure followed by corrective action.

Preventive maintenance: Maintenance performed at predetermined time or usage intervals.

Condition-based maintenance: Maintenance initiated by evidence of a changing asset condition.

Predictive maintenance: Use of data and models to anticipate deterioration or failure.

Hidden failure: Failure not evident during normal operation of a protective or standby function.

Redundancy: Additional capacity or functionality available to perform a required function.

Life-cycle cost: Total cost of acquisition, operation, maintenance, downtime, and disposal.

Proof test: Deliberate test intended to reveal hidden failures.

Maintenance-induced failure: Failure caused or accelerated by maintenance activity.

Management of change: Formal control of risks arising from changes to assets, processes, or procedures.

Review Questions

Short Answer Questions

1. Distinguish asset criticality from asset condition and maintenance priority.
2. Why can two identical pumps require different maintenance strategies?
3. What characteristics make run-to-failure acceptable for an asset?
4. Why are hidden failures commonly managed through functional testing?
5. State two limitations of semi-quantitative risk scoring.

Long Answer Questions

1. Explain how asset hierarchy, functional analysis, and failure-mode analysis support maintenance strategy selection.
2. Compare preventive, condition-based, predictive, and redesign strategies, including their assumptions and limitations.
3. Discuss how safety, environmental, quality, production, and financial consequences should be integrated without reducing safety to a monetary trade-off.
4. Explain the role of data quality, governance, and management of change in a criticality-based maintenance program.
5. Describe how reliability-centered maintenance and life-cycle cost reasoning can complement one another.

Analytical / Application Questions

1. A packaging-line motor has a high production consequence but a readily available standby motor. Develop a criticality assessment and recommend a maintenance strategy, stating the assumptions that would change your recommendation.
2. An emergency generator passes monthly visual inspections but fails during an annual start test. Analyze the hidden-failure problem and propose a testing, maintenance, and governance approach.
3. A plant proposes installing predictive sensors on every rotating asset. Construct a decision framework that determines where monitoring is technically justified and where run-to-failure, preventive maintenance, or redesign is superior.

Multiple Choice Questions with Answers

Question 1

Which statement best distinguishes criticality from condition?

A. Criticality describes present physical health, while condition describes business consequence.

B. Criticality describes the importance of failure, while condition describes the present state of the asset.

C. Criticality is always measured in money, while condition is always measured in hours.

D. Criticality applies only to safety assets, while condition applies only to production assets.

Correct Answer: B. Criticality describes the importance of failure, while condition describes the present state of the asset.

Question 2

Under the constant-failure-rate exponential model, what is the probability of at least one failure in 100 hours when λ = 0.002 failures per hour?

A. Approximately 2.0%

B. Approximately 10.0%

C. Approximately 18.1%

D. Approximately 20.0%

Correct Answer: C. Approximately 18.1%

Question 3

Which situation most strongly supports run-to-failure?

A. Failure disables an emergency protection system.

B. Failure causes a reportable environmental release.

C. Failure affects a bottleneck with no standby capacity.

D. Failure is inexpensive, readily detectable, and has negligible consequence.

Correct Answer: D. Failure is inexpensive, readily detectable, and has negligible consequence.

Question 4

What is the principal purpose of a proof test?

A. To reveal hidden failures in standby or protective equipment.

B. To calculate the purchase price of an asset.

C. To replace all condition-monitoring activities.

D. To increase the nominal speed of a production line.

Correct Answer: A. To reveal hidden failures in standby or protective equipment.

Question 5

Which condition is essential for effective condition-based maintenance?

A. The asset must be replaced at a fixed age.

B. The deterioration must be detectable with sufficient warning time for action.

C. The equipment must be safety-critical.

D. The monitoring system must use artificial intelligence.

Correct Answer: B. The deterioration must be detectable with sufficient warning time for action.

Question 6

A failure has probability 0.10 during a planning period and a consequence cost of $40,000. What is the expected loss under the stated model?

A. $400

B. $4,000

C. $40,000

D. $400,000

Correct Answer: B. $4,000

Question 7

Which method is primarily bottom-up?

A. Fault-tree analysis

B. FMEA

C. Discounted cash-flow analysis

D. Life-cycle costing

Correct Answer: B. FMEA

Question 8

Why may routine replacement every three months be inappropriate for a repeatedly failing pump seal?

A. All preventive maintenance is prohibited for pumps.

B. The failure may be caused by alignment, pipe strain, or operating conditions rather than age.

C. Seal failures can never be detected.

D. A pump cannot have a standby unit.

Correct Answer: B. The failure may be caused by alignment, pipe strain, or operating conditions rather than age.

Question 9

Which statement about predictive-maintenance model accuracy is most appropriate?

A. High classification accuracy alone proves financial benefit.

B. A model eliminates the need for competent maintenance personnel.

C. False alarms, missed detections, warning time, and response capability must also be evaluated.

D. Predictive models are valid regardless of changes in operating conditions.

Correct Answer: C. False alarms, missed detections, warning time, and response capability must also be evaluated.

Question 10

What is the most appropriate response when an unacceptable failure cannot be controlled by an effective maintenance task?

A. Ignore the failure if its probability is low.

B. Extend the inspection interval indefinitely.

C. Consider redesign, redundancy, or another engineered control.

D. Classify the asset as non-critical.

Correct Answer: C. Consider redesign, redundancy, or another engineered control.