π Lesson 9
D5
Case Review: Automotive Line Balancing Yield Recovery
Line balancing in automotive manufacturing means arranging tasks across workstations so that no station is overloaded or underused, helping more cars pass through the assembly line without defects or delays.
π― Learning Objectives
- β Calculate takt time and theoretical minimum workstation count given demand and available production time
- β Analyze line efficiency and bottleneck yield loss using cumulative defect propagation models
- β Design a revised line balance incorporating parallel stations and yield-recovery buffers to achieve β₯95% first-pass yield
- β Explain how variation in task time standard deviation impacts station-level yield and overall line OEE
π Why This Matters
In high-volume automotive manufacturing, a 0.5% drop in first-pass yield on a 1,200-unit/day line costs ~6 vehicles per day β over 1,500 units annually β requiring costly rework, overtime, or scrap. Line balancing isnβt just about speed; itβs the foundation for yield stability. When tasks with high defect sensitivity (e.g., torque-critical fastening, vision-guided ADAS calibration) are isolated in overloaded or fatigued stations, yield collapses. This lesson shows how integrating yield physics into balancing decisions recovers cost, improves quality cost modeling accuracy, and directly supports Module 5βs core thesis: yield is a design parameter β not an afterthought.
π Core Principles
Line balancing begins with takt time β the heartbeat of lean production β derived from customer demand and net available time. Next, task precedence diagrams define logical sequencing constraints. Traditional balancing minimizes idle time; yield-aware balancing adds three layers: (1) defect probability per task (based on SPC data, FMEA RPN, or historical yield logs), (2) defect amplification across downstream stations (e.g., misaligned subassembly causing 3x downstream fit issues), and (3) recovery capacity β either built-in (parallel verification stations) or external (dedicated rework cells with known cycle time and yield). Yield recovery is modeled as a probabilistic flow network: each station outputs good units (G), reworkables (R), and scrap (S), where R flows back at defined lag and capacity. The effective output rate becomes G_total = Ξ (1 β p_i) Γ throughput, adjusted for rework loop saturation.
π Yield-Adjusted Effective Throughput
This formula quantifies actual good units per hour after accounting for serial defect accumulation and finite rework capacity. It replaces naive 'line speed' assumptions in cost models with physics-based yield throughput.
Effective Throughput (Yield-Adjusted)
G_eff = T_hr Γ β(1 β p_i) + R_in Γ r Γ I(R_in β€ C_rw)Computes actual good units per hour, incorporating serial yield loss and recoverable rework (where I() is indicator function ensuring rework inflow R_in does not exceed rework capacity C_rw).
Variables:
| Symbol | Name | Unit | Description |
|---|---|---|---|
| G_eff | Effective good-unit throughput | units/hr | Actual rate of conforming units exiting the line |
| T_hr | Nominal hourly throughput | units/hr | Units entering per hour (demand-driven) |
| p_i | Defect probability at station i | decimal | Historical first-pass defect rate per station |
| R_in | Rework inflow rate | units/hr | Defective units returned for correction |
| r | Rework yield | decimal | Fraction of reworked units that become good |
| C_rw | Rework cell capacity | units/hr | Maximum sustainable rework throughput |
Typical Ranges:
Tier-1 powertrain line: 92 β 98%
Body shop closure line: 88 β 95%
π‘ Worked Example
Problem: A powertrain line has 12 stations, takt time = 57 sec, nominal throughput = 63 units/hr. Station 7 (oil filter installation) has historical defect rate pβ = 4.2%. Rework cell capacity = 3 units/hr, average rework time = 4.5 min/unit. Assume all other stations have p < 0.8% and no rework backlog. Calculate effective good-unit throughput.
1.
Step 1: Compute nominal good output before rework: 63 Γ (1 β 0.042) = 63 Γ 0.958 = 60.35 units/hr
2.
Step 2: Determine rework generation rate: 63 Γ 0.042 = 2.65 units/hr β less than rework cell capacity (3 units/hr), so no backlog forms
3.
Step 3: Assume 90% of reworked units become good (rework yield r = 0.9): additional good units = 2.65 Γ 0.9 = 2.39
4.
Step 4: Total effective good throughput = 60.35 + 2.39 = 62.74 units/hr
Answer:
The result is 62.74 units/hr, which falls within the safe range of 62.5β63.0 units/hr β confirming rework capacity is sufficient and yield loss is recoverable without line slowdown.
ποΈ Real-World Application
At Toyotaβs Tsutsumi Plant (2022), a yield drop from 98.1% to 94.7% on the Camry hybrid battery pack line was traced to Station 14 (thermal interface material dispensing), where task time variance exceeded Β±12% due to aging robotic dispensers. Engineers rebalanced by splitting the task: Station 14A (dispense, p=1.3%) and Station 14B (vision verify + manual touch-up, p=0.4%), adding a 30-sec buffer. Cycle time increased by 2.1 sec but first-pass yield recovered to 97.9%, reducing rework labor cost by $182k/yr and eliminating 3.2 hrs/day of off-line rework queue time β validating yield-aware balancing as a cost-effective alternative to full automation upgrade.
π§ Interactive Calculator
π§ Open Production Cost Modeling Calculatorπ Case Connection
π Automotive Tier-1 Supplier Line Balancing Optimization
Labor cost overrun due to unbalanced station cycle times and high overtime