Mixed-Model, Multi-Model, and Flexible Line Design

Introduction and Context

Traditional assembly-line balancing assumes that one product, with relatively stable task times and precedence relationships, flows through a sequence of workstations. Contemporary operations rarely satisfy these assumptions. Automotive, electronics, medical-device, appliance, and consumer-product operations increasingly produce several variants on the same line. Products may share common tasks but differ in components, processing times, tooling, quality requirements, and routing. Consequently, capacity planning must address not only how much capacity is needed, but also which product, in which sequence, using which resource and workforce capability.

A mixed-model line produces different product models intermixed in a planned sequence. A multi-model line produces models in campaigns or batches, often requiring setups between models. A flexible line can change its task assignments, tooling, equipment, or labor allocation in response to product mix and operating conditions. These designs improve responsiveness and asset utilization, but they create difficult interactions among precedence, takt time, zoning, parallel stations, setups, sequence-dependent changeovers, equipment compatibility, and worker skills.

This chapter develops a graduate-level framework for designing and managing such systems. The central principle is that line balance is not a single static calculation. It is a coordinated design problem involving product architecture, demand, resources, sequence, workforce capability, and uncertainty.

Learning Objectives

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

1. Distinguish single-model, mixed-model, multi-model, and flexible line designs.
2. Calculate takt time, workload, theoretical minimum stations, balance loss, and effective capacity.
3. Formulate line-balancing decisions with alternative task assignments, parallel stations, zoning, equipment, setup, and skill constraints.
4. Explain how sequencing affects workload smoothness and changeover performance.
5. Evaluate workforce flexibility and technology-enabled line designs.
6. Apply a structured method to a realistic mixed-model line-design problem.
7. Critically assess operational, human, ethical, and governance implications.

1. Fundamental Concepts

1.1 Takt time and cycle time

Takt time expresses the maximum time allowed per unit to satisfy demand during the available production time:

where T is takt time in time per unit, A is net available production time per planning period, and D is required demand in units per period. The units of A and D must use the same period. Net available time excludes planned breaks, maintenance, meetings, and other nonproduction intervals.

For a line to meet demand reliably, the effective cycle time must generally satisfy C_e ≤ T. However, variability, downtime, micro-stoppages, and quality losses mean that designing exactly to takt can be unsafe. If a line has availability a, performance efficiency p, and first-pass yield y, its effective output rate is approximately:

where C is the planned cycle time and Q_e is good output per planning period. This multiplicative expression assumes the losses are measured consistently and are sufficiently independent for planning purposes. In practice, correlated losses may require simulation or empirical adjustment.

1.2 Work content and theoretical station count

Let t_im denote the processing time of task i for model m, and let d_m be the proportion of total demand represented by model m, with ∑d_m = 1. The weighted average work content is:

For a mixed-model line with cycle time C, the theoretical minimum number of stations is:

This is a lower bound, not necessarily a feasible design. It ignores precedence restrictions, indivisible tasks, zoning, equipment requirements, setup losses, and worker availability. If tasks are model-specific, each model’s task must also be assigned to a station capable of performing it.

A common balance-efficiency measure is:

where W is total assigned workload per representative unit, N is the number of stations, and C is cycle time. Balance loss is 1 − E. For mixed-model lines, W should be based on weighted work content, and the result should not be interpreted as a complete measure of performance because it excludes variation and changeover losses.

1.3 Mixed-model, multi-model, and flexible lines

A mixed-model line is often preferred when customers require short lead times and models share a substantial process structure. A multi-model line may be superior when setup times are significant or products require incompatible equipment. Hybrid systems may use a common mixed-model final assembly line supplied by model-specific subassembly cells.

2. Product and Process Representation

2.1 Precedence networks

Each product is represented by a directed acyclic graph. A node represents a task, and an arc i → j means task i must precede task j. In a mixed-model system, some precedence relationships are common to all models, while others are model-specific.

A practical representation uses:

• a common task set for operations performed on every model;
• optional tasks activated by particular models;
• alternative task modes, such as manual or automated execution;
• resource eligibility sets identifying compatible stations or machines;
• zoning constraints that prohibit particular tasks from being placed together or require them to remain in a defined area.

The line designer must distinguish logical precedence from physical adjacency. Two tasks may be consecutive in the precedence graph but assigned to different stations, provided the cycle-time and material-transfer conditions are satisfied.

2.2 Alternative task assignments

A task may be assigned to one of several stations, or performed by one of several technologies. Alternatives arise because of:

• shared tooling or specialized equipment;
• ergonomic restrictions;
• model-specific fixtures;
• quality containment requirements;
• parallel stations;
• worker certification requirements;
• make-or-buy or manual-automation choices.

An alternative is not automatically beneficial. It may reduce local workload while increasing material handling, staffing complexity, maintenance requirements, or quality risk. Feasible alternatives should therefore be evaluated using total system cost and service performance rather than station utilization alone.

2.3 Parallel stations

Parallel stations perform the same or compatible work simultaneously. If a task group has workload w and cycle time C, the minimum number of identical parallel stations based only on workload is:

This is valid only when work can be divided across stations, demand is sufficiently continuous, and both stations possess the necessary tools and skills. If a task is indivisible and takes longer than C, parallelization does not solve the problem; the task must be redesigned, split, automated, or assigned a longer effective cycle through a different architecture.

Parallel stations can be dedicated by model or shared across models. Shared parallelism improves flexibility but requires dispatching rules, common tooling, and visibility of queue conditions.

2.4 Zoning and spatial constraints

Zoning assigns tasks to areas with special conditions. Examples include:

• paint or chemical zones;
• cleanrooms and controlled environments;
• high-voltage or high-temperature areas;
• noise and vibration zones;
• ergonomic reach or lifting zones;
• customer-specific security zones.

Zoning is a hard feasibility constraint, not merely a layout preference. A theoretically balanced line can be infeasible if it places incompatible operations together or creates unsafe travel paths. Design reviews should include safety engineering, industrial hygiene, quality, maintenance, and logistics representatives.

3. A General Optimization Formulation

A useful mixed-model formulation assigns tasks to stations while respecting precedence and resource conditions. Let:

• x_ism = 1 if task i for model m is assigned to station s, otherwise 0;
• C be the target cycle time;
• t_im be task time;
• y_s = 1 if station s is opened;
• r_is = 1 if station s is eligible for task i.

A basic assignment constraint is:

The station workload constraint is:

Precedence can be represented by requiring the station of predecessor i not to occur after the station of successor j:

for every precedence arc i → j and relevant model. This simple expression assumes station numbers increase in flow direction and that the tasks belong to the same model representation. More elaborate formulations are needed when tasks can be split, performed in parallel, or assigned to alternative routing paths.

Resource eligibility can be imposed by:

A practical objective may minimize a weighted combination of station count, labor cost, equipment investment, expected changeover cost, ergonomic exposure, and imbalance risk. For example:

where N is station count, K is capital or equipment cost, S is setup cost, R is a risk or variability measure, and the coefficients represent managerial priorities. Because these terms have different units, they must be normalized or converted to a common economic or utility scale before combining them. Exact optimization may become computationally difficult as product variety and constraints increase; mixed-integer programming, constraint programming, heuristics, and discrete-event simulation are complementary methods rather than substitutes.

4. Sequencing and Setup Management

4.1 Sequencing in mixed-model lines

The sequence determines when each model reaches each station. A good sequence smooths the consumption of components and distributes work content over time. If a high-content model is followed by another high-content model, stations may experience temporary overload even though average workload is feasible.

A common sequencing concept is workload smoothing. For a sequence of length L, let [[MATH;t_{im}]] be the time at station [[MATH;s]] for the model occupying position [[MATH;k]]. The cumulative workload through position [[MATH;k]] is:

where [[MATH;m(h)]] is the model scheduled in position [[MATH;h]]. A sequence is smoother when cumulative workload stays close to its target trajectory [[MATH;kC]]. The objective may minimize the sum of absolute deviations or squared deviations, depending on whether large errors should be penalized disproportionately.

4.2 Sequence-dependent setups

In multi-model lines, changing from model i to model j may require setup time s_ij. If a production campaign contains transitions, total setup time is:

where [[MATH;L]] is the number of positions in the sequence. This formula assumes setups occur between consecutive positions and that setup times are known and additive. If setup can be performed offline, only the portion that consumes line time should be included in the cycle-time calculation.

Sequence-dependent setup creates a trade-off: grouping similar models reduces setup time, while intermixing models reduces finished-goods inventory and improves responsiveness. Changeover reduction methods include externalizing setup work, standardizing fixtures, using quick-release connections, preparing kits, and applying single-minute exchange principles.

5. Workforce and Skill Constraints

A flexible line depends on more than cross-training claims. It requires verified skill matrices, training capacity, certification rules, fatigue controls, and real-time assignment discipline. A skill matrix may record whether worker [[MATH;w]] is qualified for task [[MATH;i]], but qualification does not necessarily imply equal speed or equal quality performance.

Let [[MATH;z_{ws} = 1]] indicate that worker [[MATH;w]] is assigned to station [[MATH;s]]. A basic staffing condition is:

for each operating station. If a worker can operate only one station in a period, then:

Skill eligibility can be represented with [[MATH;z_{ws} ≤ q_{wi}]] when the worker must perform task [[MATH;i]] and [[MATH;q_{wi}]] equals 1 if qualified. More realistic models include learning curves, fatigue, absenteeism, break coverage, and team-based work.

Multi-skilled workers provide resilience during absenteeism and mix changes, but excessive flexibility can create role ambiguity, training cost, and hidden dependence on a small number of experts. Job rotation should be designed with employee participation and ergonomic evidence; it should not be used to transfer instability to workers without adequate recovery time or compensation.

6. Equipment, Automation, and Flexible Technology

Equipment constraints include machine capacity, tooling compatibility, maintenance windows, calibration requirements, and common-resource contention. A station may be technically capable of performing a task but unavailable because its fixture is occupied by another model or because a safety interlock prevents simultaneous operations.

Automation is attractive where tasks are repetitive, hazardous, highly precise, or labor constrained. However, automation may reduce flexibility if fixtures are model-specific, programming changes are slow, or maintenance capability is weak. Flexible automation—robotic tooling, programmable fastening, machine vision, and automatic identification—can support variety, but only when data integrity and error recovery are robust.

Artificial intelligence can assist with demand-mix forecasting, sequence generation, anomaly detection, dynamic labor allocation, and predictive maintenance. It should support, not replace, engineering judgment. Governance requires traceable data, validation against real production outcomes, cybersecurity, protection of employee information, and human authority to override unsafe recommendations. A model that optimizes throughput while increasing ergonomic risk is not an acceptable line-design solution.

7. Worked Example: Mixed-Model Assembly Line

An illustrative line produces three models with the following demand mix and work content:

The weighted work content is:

Suppose net available time is 420 minutes per shift and demand is 14 units per shift. Takt time is:

The theoretical minimum number of stations is:

The lower bound of one station is not operationally feasible if some model-specific tasks cannot be combined, if precedence creates indivisible groups, or if equipment and ergonomic limits require separation. Suppose engineering analysis identifies two feasible stations with weighted workloads of 14.8 and 13.4 minutes. Their balance efficiency is:

The low efficiency does not automatically mean the design is poor. It may reflect unavoidable precedence, model-specific equipment, safety zoning, or the need for capacity protection. If the line operates at a planned cycle time of 30 minutes with availability 0.90, performance 0.95, and first-pass yield 0.98, expected good output is approximately:

Thus, two stations at takt are not sufficient to achieve 14 good units per shift under these losses. The design must reduce losses, add time or parallel capacity, lower the required output, or provide a deliberate capacity buffer. This illustrates why theoretical balancing based solely on average work content can produce an unsafe decision.

8. Applied Case: Illustrative Electric-Cargo-Van Line

Background

An illustrative electric-cargo-van plant assembles three vehicle variants on one final-assembly line: a short-range van, a long-range van, and a refrigerated-body variant. The variants share 70% of tasks, but the long-range model requires additional battery modules and the refrigerated model requires insulated-body installation. Battery work must occur in a high-voltage zone, while adhesive application must occur in a controlled-temperature zone.

Problem

Average demand appears to fit the existing line, but daily output fluctuates. High-range vehicles cause overload at battery installation, while refrigerated vehicles create adhesive and inspection delays. A small group of certified technicians performs several critical operations, creating vulnerability to absence.

Analysis

The engineering team maps common and optional tasks, then calculates weighted workload by station. It discovers that average workload is acceptable, but the model sequence produces clusters of high-content vehicles. In addition, the battery fixture is compatible with only one station, and the adhesive station cannot be moved into the high-voltage zone. The apparent problem is therefore a combination of sequencing, zoning, equipment, and skills—not simply insufficient labor.

Approach and solution

The plant adopts four measures:

1. It uses a level sequence that limits consecutive high-content models and smooths cumulative station workload.
2. It installs a second compatible battery fixture at a parallel station for peak periods.
3. It cross-trains operators through staged certification and creates break coverage for high-voltage work.
4. It redesigns the adhesive kit so preparation occurs offline, reducing line-exposed setup time.

A daily control board displays model sequence, station workload, certified staffing, equipment status, and planned recovery actions. The sequence is frozen within a defined time window except for safety or material emergencies, preventing frequent disruptive resequencing.

Outcome

The line achieves more stable hourly output and reduces dependence on a single technician. The improvement does not arise from balance efficiency alone; it comes from coordinating sequence, parallel capacity, zoning, setup design, and workforce capability. The plant also retains a contingency plan for a fixture failure and defines escalation rules when model mix exceeds the validated operating envelope.

Learning points

The case demonstrates that mixed-model design is a socio-technical system. Average capacity can be adequate while short-term capacity is inadequate. Robust design requires explicit treatment of model sequence, constrained resources, safety zones, skills, and recovery policies.

9. Implementation Framework and Performance Measures

A disciplined implementation sequence is:

1. Confirm demand, available time, service requirements, and product mix.
2. Map product variants, common tasks, optional tasks, precedence, and quality gates.
3. Measure task times by model and identify variability rather than relying solely on standard times.
4. Identify equipment, zoning, ergonomic, skill, and material-flow constraints.
5. Generate alternative station architectures, including parallel and hybrid options.
6. Optimize or heuristically construct assignments and sequences.
7. Test alternatives with discrete-event simulation or controlled pilot runs.
8. Validate safety, quality, maintainability, and worker capability.
9. Implement visual controls, standard work, escalation rules, and feedback loops.
10. Rebalance when demand mix, product design, equipment, or staffing changes.

Important measures include good units per hour, schedule attainment, station starvation and blocking, first-pass yield, changeover time, overtime, utilization, ergonomic risk, training coverage, and recovery time after disruption. Utilization should not be maximized blindly. Very high utilization may reduce resilience and increase queueing, fatigue, and quality problems.

10. Challenges, Limitations, and Governance

Mixed-model line design has several limitations. Standard times may be inaccurate for rare variants. Weighted averages conceal peak-load conditions. Optimization models may omit human adaptation, informal workarounds, and maintenance realities. Simulation results are only as credible as their assumptions and input distributions. A mathematically optimal design may be unacceptable if it creates excessive physical or cognitive burden.

Governance issues include safety compliance, traceability of quality decisions, fair allocation of training opportunities, protection of worker monitoring data, and clear accountability for algorithmic recommendations. Regulatory requirements may apply to electrical safety, machinery guarding, chemical exposure, product traceability, and working time. Line design should be reviewed through formal management-of-change procedures whenever product, process, software, equipment, or staffing assumptions change.

Best practice is to treat flexibility as a capability that must be engineered and maintained. It requires modular fixtures, clear standard work, cross-training, reliable data, preventive maintenance, and management systems that reward safe problem solving rather than concealed workarounds.

Chapter Summary

Mixed-model, multi-model, and flexible lines address the reality of product variety and uncertain demand. Takt time provides a demand-based design reference, while weighted work content provides only a lower bound on station count. Feasible design must also account for precedence, alternative assignments, parallel stations, zoning, equipment compatibility, setups, sequence dependence, and workforce skills.

Mixed-model sequencing smooths workload but may increase coordination requirements. Multi-model campaigns reduce setup frequency but can increase inventory and responsiveness losses. Flexible lines improve resilience when supported by modular technology, verified skills, sound data, and disciplined governance. Optimization models help structure decisions, but simulation, engineering review, worker participation, safety analysis, and implementation controls remain essential.

The most important managerial lesson is that line balance is a system property. A line with excellent average balance may still fail because of sequence peaks, constrained equipment, absent specialists, quality losses, or unsafe work. Effective design therefore combines quantitative analysis with socio-technical judgment.

Key Terms

Takt time: Available production time divided by required demand.

Cycle time: Planned elapsed time between successive completed units at a station or line.

Mixed-model line: A line producing different models in an interleaved sequence.

Multi-model line: A line producing models in campaigns or batches.

Flexible line: A line capable of changing assignments, resources, routing, or workforce allocation.

Work content: Total processing time required to produce a unit.

Balance efficiency: Assigned workload divided by available station capacity.

Precedence relationship: A requirement that one task occur before another.

Parallel station: One of multiple stations performing the same or compatible work.

Zoning: Restriction or grouping of work according to spatial, safety, environmental, or process conditions.

Alternative task assignment: More than one feasible station, worker, or technology for a task.

Sequence-dependent setup: Changeover time determined by the preceding and succeeding models.

Model mix: Proportions of different models in total demand.

Cross-training: Preparing workers to perform multiple tasks or operate multiple stations.

Skill matrix: A record of worker qualifications and capability levels.

Starvation: A station is idle because required input has not arrived.

Blocking: A station cannot release work because the downstream position is unavailable.

Effective capacity: Output capability after accounting for losses such as downtime, speed loss, and defects.

Workload smoothing: Sequencing models to reduce short-term variation in station workload.

Flexible automation: Programmable equipment capable of handling product or process variation.

Review Questions

Short Answer Questions

1. Distinguish a mixed-model line from a multi-model line.
2. Why is N_min = ⌈W̄/C⌉ only a lower bound on the number of stations?
3. Identify two examples of zoning constraints.
4. What is the operational purpose of workload-smoothing sequences?
5. Why can high utilization reduce line resilience?

Long Answer Questions

1. Explain how precedence, equipment eligibility, zoning, and worker skills interact in flexible line design.
2. Compare the advantages and disadvantages of mixed-model and multi-model production.
3. Develop and explain a line-design optimization framework for a product family with optional tasks.
4. Discuss how sequence-dependent setup times affect the choice between intermixing and batching models.
5. Evaluate the role of simulation, optimization, and worker participation in validating a flexible line.

Analytical / Application Questions

1. A line has 480 net minutes per shift and demand of 16 units. Calculate takt time. If weighted work content is 34 minutes, calculate the theoretical minimum station count and explain why the result may be infeasible.
2. Three models have shares 0.5, 0.3, and 0.2, with work contents 20, 30, and 40 minutes. Calculate weighted work content and assess whether two stations with a 30-minute cycle provide sufficient average capacity.
3. Design a response for a line in which a high-content model overloads one station, its fixture is unique, and only two workers are certified. Specify sequencing, equipment, and workforce actions, and identify risks in your proposal.

Multiple Choice Questions with Answers

Question 1

Takt time is calculated as:

A. Demand divided by available time

B. Available production time divided by demand

C. Work content divided by station count

D. Cycle time multiplied by utilization

Correct Answer: B. Available production time divided by demand

Question 2

If weighted work content is 28 minutes and cycle time is 30 minutes, the theoretical minimum station count is:

A. 1

B. 2

C. 28

D. 30

Correct Answer: A. 1

Question 3

Which condition most directly creates a zoning constraint?

A. A task has a low average time

B. A product has stable demand

C. A task requires a controlled-temperature environment

D. A station has excess labor

Correct Answer: C. A task requires a controlled-temperature environment

Question 4

The principal purpose of workload-smoothing sequencing is to:

A. Eliminate all precedence relationships

B. Reduce short-term workload variation across the sequence

C. Guarantee zero setup time

D. Increase the number of product variants

Correct Answer: B. Reduce short-term workload variation across the sequence

Question 5

A sequence-dependent setup time is determined primarily by:

A. Only the total daily demand

B. The preceding and succeeding models

C. The number of workers on leave

D. The theoretical minimum station count

Correct Answer: B. The preceding and succeeding models

Question 6

Which statement about the theoretical minimum station count is correct?

A. It always equals the feasible station count

B. It ignores all task precedence

C. It is a lower bound based on workload and cycle time

D. It includes all downtime and quality losses automatically

Correct Answer: C. It is a lower bound based on workload and cycle time

Question 7

Two parallel stations are most appropriate when:

A. The task is forbidden by safety rules

B. Workload exceeds one station’s capacity and the work is compatible with duplication

C. The task is indivisible and longer than the cycle time

D. No worker is qualified for the task

Correct Answer: B. Workload exceeds one station’s capacity and the work is compatible with duplication

Question 8

If a line produces 14 planned units and availability, performance, and yield are 0.90, 0.95, and 0.98, respectively, expected good output is approximately:

A. 9.80 units

B. 11.73 units

C. 14.00 units

D. 15.87 units

Correct Answer: B. 11.73 units

Question 9

Which is a significant governance requirement for AI-assisted line balancing?

A. Removing all human override authority

B. Using only historical data without validation

C. Ensuring traceability, cybersecurity, and safety review

D. Maximizing utilization regardless of ergonomic effects

Correct Answer: C. Ensuring traceability, cybersecurity, and safety review

Question 10

Cross-training primarily improves flexibility by:

A. Making all tasks identical

B. Allowing qualified workers to cover multiple tasks or stations

C. Eliminating equipment constraints

D. Removing the need for standard work

Correct Answer: B. Allowing qualified workers to cover multiple tasks or stations