Spare Parts, Maintenance Logistics and Resource Optimisation

Introduction

Maintenance performance depends not only on diagnosing failures and selecting suitable maintenance tasks, but also on having the correct parts, tools, people and information available when they are needed. A technically excellent maintenance programme can still fail if a bearing is unavailable, a repairable module is awaiting overhaul, or a specialist technician cannot be mobilised. Spare-parts management therefore connects reliability engineering with inventory control, procurement, logistics, finance and operations.

The central decision is not simply how to minimise inventory. It is to determine the economically and operationally appropriate level of availability for each item. Holding too few spares creates downtime, emergency freight, production loss and unsafe workarounds. Holding too many creates tied-up capital, storage expense, deterioration, obsolescence and disposal losses. The appropriate policy depends on failure consequences, demand characteristics, replenishment time, repair capacity and supply-chain risk.

This chapter develops methods for classifying critical spares, forecasting intermittent demand, setting reorder points and service levels, and balancing inventory cost against downtime risk and resilience requirements.

Learning objectives

After studying this chapter, a student should be able to:

1. Distinguish consumable, repairable, rotable, insurance and operational spares.
2. Classify spare criticality using consequence, probability and supply-risk criteria.
3. Explain why intermittent demand requires methods different from ordinary production inventory.
4. Calculate safety stock, reorder points, cycle stock and service measures under stated assumptions.
5. Incorporate repair turnaround, capacity, obsolescence and supplier risk into stocking decisions.
6. Evaluate a maintenance-logistics policy using cost, reliability and operational-performance measures.

1. The role and classification of spare parts

A spare part is an item held to restore or support an asset after wear, damage or failure. Spare parts differ from ordinary production materials in several important ways:

• Demand is often generated by random failures rather than a smooth production schedule.
• A part may have a very high value but extremely low demand.
• The cost of not having it can greatly exceed its purchase price.
• Some parts can be repaired and returned to stock.
• Demand may be intermittent, lumpy, dependent on an installed population, or affected by maintenance policy.

1.1 Functional categories

Consumables are used during maintenance and are not normally recovered, such as lubricants, seals, fasteners and cleaning materials. Their demand may be predictable from planned work.

Repairables can be restored through inspection, repair, refurbishment or remanufacture. Examples include pumps, electronic cards and hydraulic actuators. Their inventory system includes both serviceable stock and items in the repair pipeline.

Rotables are repairable items designed to be exchanged as complete units. A failed unit is removed and replaced by a serviceable one, after which the failed unit enters the repair cycle.

Insurance spares are held because their absence could cause unacceptable consequences, even though demand is rare. A turbine rotor or specialised control module may fall into this category.

Operational spares support routine maintenance and usually have repeat demand. Their stocking policy can often use conventional reorder-point methods.

Capital spares are expensive, long-lead items whose availability is important over the life of an asset. They require lifecycle planning rather than only a short-term inventory calculation.

1.2 Criticality is multidimensional

Unit price is not an adequate measure of importance. A low-cost gasket may stop a process if it is unavailable, whereas an expensive component may have a built-in redundant standby. A useful criticality assessment considers:

1. Safety and environmental consequence: possibility of injury, hazardous release or environmental harm.
2. Production consequence: expected lost output and restart time.
3. Failure likelihood: failure rate, age, duty cycle and condition information.
4. Restoration time: replacement and testing duration.
5. Supply risk: lead time, supplier concentration, transport difficulty and geopolitical exposure.
6. Substitutability: approved alternatives, interchangeability and local fabrication.
7. Repairability: repair success rate and repair turnaround time.
8. Obsolescence exposure: discontinued designs, software dependence and changing standards.

A practical risk measure is the expected loss associated with a stockout:

where P(F) is the probability of a failure during the planning period, P(SO | F) is the probability of a stockout given that failure, and C(SO) is the consequence cost of the stockout. This expression is a screening model, not a complete inventory optimisation formula. It assumes that the probabilities and consequence estimate are defined for the same period and that the consequence cost can be meaningfully estimated.

A risk matrix is useful for initial classification, but high-consequence items should also receive engineering review. A common policy is to classify items as critical, essential or routine, while separately recording supply-risk and obsolescence flags.

2. Inventory quantities and maintenance logistics

The main inventory quantities are:

• On-hand stock: physically available items.
• Allocated stock: reserved for a known work order or customer.
• On-order stock: ordered but not yet received.
• In-repair stock: failed repairable items being restored.
• Pipeline stock: items in transit or in external processing.
• Available-to-promise stock: stock expected to be available after considering allocations and reliable receipts.

For a repairable item, the required population is influenced by demand during the repair cycle. A simple balance is:

where each term is a number of units. This identity is a stock-accounting relationship; it does not determine the correct total population by itself.

Logistics includes purchasing, receiving inspection, storage, preservation, kitting, issue to maintenance, return of removed units, repair management, transport and data control. Poor logistics can create a “false shortage”: the part exists somewhere in the organisation but cannot be identified, released, transported or verified in time.

2.1 The repair loop

For repairables, the effective replenishment time is not merely supplier lead time. It may include removal, diagnosis, transport, queueing, repair, testing, certification and return. If repair capacity is limited, failures can accumulate in the repair queue, increasing the turnaround time and the required rotable pool.

Repairable inventory should therefore be managed with:

• clear serviceable and unserviceable status codes;
• maximum repair turnaround targets;
• approved repair procedures and quality records;
• visibility of units by serial number where traceability is required;
• controls against cannibalisation that create hidden shortages;
• periodic review of repair yield and repeat failures.

3. Forecasting intermittent and lumpy demand

3.1 Why ordinary averages can mislead

For a frequently used consumable, mean demand may provide a reasonable starting point. For a critical spare, demand may be zero for many periods and then several units after a major failure. The mean alone hides the frequency, variance and timing of demand. Forecasting must distinguish:

• Intermittency: many zero-demand periods.
• Lumpiness: high variation in non-zero demand size.
• Obsolescence or phase-out: demand declines because the installed base is being removed.
• Dependent demand: demand follows planned overhauls or known equipment population.
• Common-cause events: one event creates simultaneous demand for many identical units.

Data should be cleaned before forecasting. Transfers, emergency substitutions, duplicate issues, returns and stockout-censored demand must not be treated as ordinary observations. A zero issue may mean no requirement—or that the item was unavailable.

3.2 Basic forecasting measures

For a series of demand observations d₁, d₂, ..., d_n, the sample mean is:

and the sample standard deviation is:

The standard deviation formula assumes observations are sufficiently comparable and is more useful when demand is approximately stable. It becomes unreliable when the series contains many zeros, structural changes or censored demand.

For intermittent demand, Croston-type methods separately estimate the size of a non-zero demand and the interval between non-zero demands. If z_t is the latest non-zero demand size and p_t is the interval since the previous non-zero demand, smoothed estimates can be represented as:

The forecast rate is then approximately:

where 0 < α ≤ 1. These methods assume that non-zero demand size and inter-demand interval can be modelled separately and that the process is reasonably stationary. Variants correct bias and can perform better when demand changes, but no intermittent-demand method can compensate for poor failure data or a changing installed base.

For very rare, high-consequence items, engineering scenarios and failure-rate models may be more informative than a purely statistical forecast. If failures follow a Poisson process with constant rate λ per unit time, the probability of k failures in time T is:

This model assumes independent events, a constant failure rate and no common-cause failures. It should not be applied automatically to ageing equipment, dependent failures or small samples.

4. Reorder points, safety stock and service levels

4.1 Continuous-review reorder point

In a continuous-review system, an order is placed when inventory position reaches the reorder point. Inventory position is normally:

For constant demand rate d and deterministic lead time L, the basic reorder point is:

where dL is expected lead-time demand and SS is safety stock. Units must be consistent; for example, demand in units per week and lead time in weeks produce units.

If demand during lead time is approximately normal with mean μ_LT and standard deviation σ_LT, a cycle-service-level formulation is:

where z is the standard-normal quantile associated with the target probability of no stockout during one replenishment cycle. The safety stock is SS = z × σ_LT. This formula assumes a suitable normal approximation, reliable estimates, independent or appropriately modelled variability, and that a stockout means demand cannot be met from available stock.

If demand per period has mean μ_d and standard deviation σ_d, and lead time is constant at L periods, then:

If both demand and lead time vary, with independent demand and lead-time variation, an often-used approximation is:

where L̄ and σ_L are the mean and standard deviation of lead time. The formula excludes the interaction term under the stated approximation and independence assumption. Correlation, non-normal demand and intermittent demand require simulation or an appropriate discrete model.

4.2 Service-level definitions

Two service measures must not be confused:

• Cycle service level: probability that all demand is met during a replenishment cycle.
• Fill rate: proportion of demand units supplied immediately from stock.

A high cycle service level does not necessarily imply a high fill rate when demand quantities are lumpy. Management should specify which measure is required and over what time horizon.

4.3 Worked example: reorder point

A commonly used seal has weekly demand with mean 8 units and standard deviation 3 units. Supplier lead time is constant at 4 weeks. The maintenance store targets a 95% cycle service level, for which z ≈ 1.645.

Lead-time demand mean:

μ_LT = 4 × 8 = 32 units.

Lead-time standard deviation:

σ_LT = √4 × 3 = 6 units.

Safety stock:

SS = 1.645 × 6 = 9.87 ≈ 10 units.

Thus:

ROP = 32 + 10 = 42 units.

The result is appropriate only if the demand distribution is sufficiently stable and approximately normal. If the item is actually intermittent, the calculation should be tested against historical stockout and demand-event data.

5. Order quantities and total cost

For a regularly consumed item, the economic order quantity model balances ordering and holding costs:

where D is annual demand in units per year, S is ordering or setup cost per order, and H is annual holding cost per unit. The model assumes constant known demand, instantaneous replenishment, no shortages, constant costs and divisibility of order quantities. These assumptions are often unrealistic for critical spares, so Q* should be treated as a benchmark rather than an automatic decision.

A simplified annual relevant cost is:

The first term is annual ordering cost and the second is average cycle-stock holding cost. Safety stock and shortage consequences should be added when comparing alternative policies. For repairables, the relevant “order quantity” may be the number of rotables in the pool, not a supplier purchase batch.

A broader decision criterion is expected total lifecycle cost:

The terms should be estimated over the same time horizon and discounted if the analysis spans many years. Shortage cost may include lost contribution margin, restart expense, expedited freight, contractual penalties and safety consequences. It should not be double-counted with production-loss estimates.

6. Capacity, obsolescence and resilience

Inventory cannot be separated from repair capacity. If repair demand exceeds repair throughput, work-in-process grows and serviceable stock falls. Capacity planning should examine technicians, test benches, specialist vendors, calibration facilities and spare repair kits. A small increase in repair turnaround can require additional rotable units.

Obsolescence is especially important for electronic, software-dependent and discontinued items. Controls include:

• lifecycle alerts and last-time-buy reviews;
• approved redesigns and form-fit-function alternatives;
• preservation and periodic testing;
• component harvesting only under controlled quality procedures;
• documentation of firmware, compatibility and configuration;
• strategic purchase of a limited number of high-risk items.

Supply-chain resilience means the ability to continue maintenance under disruption. Resilience measures include qualified second sources, local repair capability, supplier capacity audits, alternate transport routes, strategic stock, mutual-aid agreements and accurate bills of material. Dual sourcing is not automatically superior: a second supplier may introduce quality, interchangeability or regulatory risks.

A risk-based policy may use different strategies:

7. Digital systems and artificial intelligence

An enterprise asset-management system should connect equipment bills of material, failure codes, work orders, inventory transactions, supplier records and repair status. Data governance is essential: an incorrect unit of measure or duplicate part number can create more risk than an unsophisticated forecasting method.

Analytics can identify abnormal consumption, estimate lead-time distributions, detect stockout-censored demand and prioritise supplier risks. Machine-learning models may use condition indicators to forecast impending component replacement, but they require representative failure data and must be validated against maintenance practice. A prediction that cannot be explained, inspected or acted upon may not improve reliability.

AI should support—not replace—engineering judgement. Human approval remains necessary for safety-critical substitutions, changes to approved parts, release of repaired equipment and decisions involving regulatory compliance. Models should be monitored for drift as equipment populations, suppliers and maintenance policies change.

8. Illustrative case: packaging-line drive modules

Background

A packaging plant operates 18 identical conveyor drives. Each drive contains one removable electronic drive module. A failed module causes a conveyor stoppage, and the line cannot operate safely in bypass mode. Modules are repairable. External repair has a nominal turnaround of 10 working days, but historical turnaround ranges from 8 to 20 days. The installed modules are ageing, and the original manufacturer has announced a possible product withdrawal.

Problem

The plant holds one serviceable spare and frequently sends failed modules to the repair vendor without tracking their expected return. During two previous failures, the line was stopped for several days because the spare was already allocated or awaiting repair. Purchasing proposed four additional new modules, while maintenance proposed improving repair turnaround and preserving tested units.

Analysis

The item is high criticality because its stockout consequence is high, demand is intermittent and the supply risk is increasing. Historical demand alone is insufficient because two failures occurred during a period of elevated temperature and poor ventilation. The team therefore separates failure prevention from stocking analysis.

The logistics review finds:

• one module is serviceable and unallocated;
• two modules are at the repair vendor;
• one repaired module failed its functional test and requires rework;
• repair records do not identify repeat failure modes;
• a compatible successor exists but requires software validation.

The apparent stock position of four units is therefore misleading. Only one is immediately deployable.

Approach and solution

The plant adopts a controlled rotable pool. Every module receives a status code: serviceable, allocated, installed, failed, in repair, awaiting test or rejected. A maximum repair turnaround of 12 working days is agreed, with escalation at day 10. The repair vendor must provide failure-mode and test records.

A temporary policy holds two immediately serviceable modules because the consequence of a stockout is severe and the repair pipeline is uncertain. The policy is reviewed after six months using actual failure frequency, repair yield and turnaround. The plant also validates the successor module, improves cabinet cooling and introduces a condition check on power quality. A last-time-buy decision is made only after comparing the cost of original modules, successor validation, repair and obsolescence risk.

Outcome

The principal improvement does not come from simply increasing the stock quantity. Visibility of the repair pipeline reveals that the plant had nominal inventory but insufficient usable inventory. Better turnaround, environmental correction and configuration control reduce both demand and required stock. The remaining buffer is justified as an insurance spare and is reviewed against lifecycle information.

Learning points

1. Serviceable inventory is more important than nominal inventory.
2. Repair capacity is part of spare-parts capacity.
3. A criticality decision must include consequence and supply risk, not price alone.
4. Forecasts should be corrected for stockouts and changing failure conditions.
5. Obsolescence decisions require technical validation and lifecycle analysis.

9. Implementation framework and limitations

A practical implementation sequence is:

1. Build an accurate equipment-to-part bill of material.
2. Clean part numbers, units, descriptions and interchangeability data.
3. Classify criticality and identify safety- or regulation-controlled items.
4. Segment demand into regular, intermittent, repairable and insurance categories.
5. Estimate demand, lead time, repair turnaround and stockout consequences.
6. Set service objectives appropriate to consequence and cost.
7. Select reorder, pool, insurance-stock or last-time-buy policies.
8. Validate policies through historical replay or simulation.
9. Monitor stockouts, emergency purchases, fill rate, cycle service, repair turnaround, obsolescence and inventory value.
10. Review policies after asset changes, supplier changes or major failures.

Important limitations include sparse data, changing installed populations, common-cause failures, inaccurate lead times, hidden stock, uncertain shortage costs and human workarounds. Statistical confidence is weak for very rare failures. In such cases, scenario analysis and engineering judgement should be made explicit rather than disguised as precise forecasting.

Governance is also important. Safety-critical parts require approved specifications, traceability, inspection and documented release authority. Counterfeit or unauthorised parts can compromise safety and legal compliance. Ethical practice requires that production pressure never justify bypassing engineering approval or concealing a stockout-related risk.

Chapter Summary

Spare-parts management is a reliability and risk-management activity, not merely a purchasing function. Parts should be classified by function, consequence, demand behaviour, repairability, supply risk and obsolescence. Intermittent demand requires careful data cleaning and may require Croston-type methods, failure-rate models or scenario analysis rather than simple averages. Reorder points combine expected lead-time demand with safety stock, but the normal approximation is valid only under appropriate assumptions. Cycle service level and fill rate measure different outcomes.

For repairable and rotable items, serviceable stock, repair turnaround and repair capacity must be managed as one system. Total cost includes holding, ordering, repair, obsolescence, emergency and shortage costs. Resilient policies may use qualified alternate sources, strategic inventory, local repair and lifecycle planning. Digital systems and AI can improve visibility and forecasting, but data quality, validation, human approval and configuration control remain essential.

Key Terms

Spare part: An item held to restore or support an asset.

Consumable: A part or material normally used up during maintenance.

Repairable: An item that can be restored to service after failure.

Rotable: A repairable unit exchanged as a complete assembly.

Insurance spare: A rare-use spare held against a high-consequence failure.

Criticality: The importance of an item based on consequence, likelihood and recoverability.

Intermittent demand: Demand containing many periods with zero demand.

Lumpy demand: Demand with substantial variation in the size of non-zero events.

Lead time: Time between placing an order and receiving usable stock.

Inventory position: On-hand plus on-order minus allocated inventory.

Reorder point: Inventory position at which replenishment is initiated.

Safety stock: Buffer held against demand or supply uncertainty.

Cycle service level: Probability of meeting all demand during a replenishment cycle.

Fill rate: Proportion of demand units supplied immediately from stock.

Repair turnaround time: Time from removal or receipt of a failed item to return of a tested serviceable item.

Obsolescence: Loss of usefulness or supportability caused by technological or lifecycle change.

Stockout: A condition in which required demand cannot be met from available stock.

Supply-chain resilience: Ability to maintain required service during disruption.

Critical spare: A spare whose unavailability creates unacceptable operational, safety or environmental risk.

Review Questions

Short Answer Questions

1. Why is purchase price alone an inadequate measure of spare-part criticality?
2. Distinguish cycle service level from fill rate.
3. What is meant by intermittent demand?
4. Identify four elements that may contribute to repair turnaround time.
5. Why must stockout-censored demand be treated carefully in forecasting?

Long Answer Questions

1. Explain a multidimensional framework for classifying spare parts and recommend policies for critical, essential and routine items.
2. Describe how a repairable-spares system differs from a consumable-spares system.
3. Derive and explain the continuous-review reorder-point model, including its assumptions and limitations.
4. Discuss how obsolescence and supply-chain resilience influence spare-parts strategy.
5. Explain how an enterprise asset-management system can improve maintenance logistics while introducing new governance risks.

Analytical / Application Questions

1. Weekly demand for a part has mean 12 units and standard deviation 4 units. Lead time is constant at 3 weeks and the target cycle service level is 97.5%, with z = 1.96. Calculate lead-time demand, safety stock and reorder point, stating assumptions.
2. A repairable module has four units in the organisation: one serviceable, two at a repair vendor and one failed but awaiting diagnosis. Evaluate whether the organisation has four usable spares and propose controls for the repair loop.
3. A high-value component has almost no recorded demand but a 16-month supplier lead time and severe production consequences. Compare a purely historical forecast with an insurance-spare and lifecycle-risk approach. Which additional data would you collect?

Multiple Choice Questions with Answers

Question 1

Which factor most directly distinguishes an insurance spare from an ordinary operational spare?

A. It is always the least expensive item.

B. It is held against a rare but high-consequence event.

C. It is consumed during every planned maintenance task.

D. It has no need for preservation or inspection.

Correct Answer: B. It is held against a rare but high-consequence event.

Question 2

If weekly mean demand is 8 units and constant lead time is 4 weeks, what is the expected lead-time demand?

A. 2 units

B. 12 units

C. 32 units

D. 64 units

Correct Answer: C. 32 units

Question 3

In a continuous-review system, inventory position is normally calculated as:

A. On-hand plus on-order minus allocated stock

B. On-hand minus all stock in transit

C. Annual demand divided by order quantity

D. Safety stock plus annual demand

Correct Answer: A. On-hand plus on-order minus allocated stock

Question 4

Which statement correctly describes fill rate?

A. The probability that no demand occurs during lead time

B. The proportion of demand units supplied immediately from stock

C. The percentage of suppliers that deliver on time

D. The probability that a repair is successful

Correct Answer: B. The proportion of demand units supplied immediately from stock

Question 5

The basic reorder-point formula for normally approximated lead-time demand is:

A. ROP = ordering cost ÷ holding cost

B. ROP = annual demand × order quantity

C. ROP = mean lead-time demand + z × lead-time standard deviation

D. ROP = repair turnaround − supplier lead time

Correct Answer: C. ROP = mean lead-time demand + z × lead-time standard deviation

Question 6

Why can a simple average be misleading for intermittent demand?

A. It always overestimates demand by 100%.

B. It hides zero-demand periods and variation in non-zero demand events.

C. It cannot be calculated for any data set.

D. It measures supplier quality rather than demand.

Correct Answer: B. It hides zero-demand periods and variation in non-zero demand events.

Question 7

For a repairable rotable, which stock is immediately deployable?

A. A failed unit awaiting diagnosis

B. A unit at a repair vendor

C. A tested and released serviceable unit

D. A rejected unit awaiting disposal

Correct Answer: C. A tested and released serviceable unit

Question 8

The economic order quantity model primarily balances:

A. Ordering cost and holding cost

B. Safety risk and software complexity

C. Failure rate and vibration amplitude

D. Repair skill and machine speed

Correct Answer: A. Ordering cost and holding cost

Question 9

Which assumption is required by the simple Poisson failure-count model used in the chapter?

A. Failure events are independent with a constant rate.

B. Every failure requires exactly two spare parts.

C. Lead time is always zero.

D. Demand must be normally distributed.

Correct Answer: A. Failure events are independent with a constant rate.

Question 10

Why is repair capacity relevant to the required rotable pool?

A. Lower repair capacity can lengthen turnaround and increase units tied up in repair.

B. Repair capacity affects only purchase price, never availability.

C. Repair capacity eliminates the need for serviceable stock.

D. Repair capacity makes obsolescence impossible.

Correct Answer: A. Lower repair capacity can lengthen turnaround and increase units tied up in repair.