Manufacturing

Quality AI for Auto Components

How a Tier 1 supplier achieved zero customer audit findings with explainable AI

Zero Audit Findings
-25% False Rejects
100% PLI Compliance

Customer Profile

A Tier 1 automotive component supplier with 4 manufacturing plants across Maharashtra and Gujarat, supplying to major OEMs including Tata Motors, Mahindra, and Maruti Suzuki. The company is a beneficiary of the PLI (Production Linked Incentive) scheme for automotive components.

Challenge

The company had deployed AI-powered visual inspection systems for quality control, but faced significant challenges that threatened both customer relationships and PLI compliance:

Inconsistent AI Performance

  • Quality inspection AI showing 15% variance across shifts and plants
  • Night shift rejection rates 20% higher than day shift
  • Newer plant in Gujarat underperforming versus established Maharashtra facilities
  • No explanation for why specific parts were rejected

Customer Audit Concerns

  • OEM customers conducting rigorous AI governance audits
  • Questions about AI decision-making they couldn’t answer
  • Risk of losing preferred supplier status
  • Increasing documentation requirements for AI-based inspection

False Reject Problem

  • 8% false positive rate causing material waste
  • Rejected parts requiring manual re-inspection
  • Production line disruptions for false alarms
  • Cost of quality higher than industry benchmarks

PLI Compliance Risk

  • Scheme requires consistent quality metrics
  • AI inconsistency creating documentation challenges
  • Risk of incentive clawback for quality issues
  • Need for auditable quality records

Solution

Rotavision deployed AI trust infrastructure across all 4 manufacturing plants:

Vishwas for Fairness & Consistency

  • Bias detection across shifts, plants, and production lines
  • Root cause analysis for performance variance
  • Continuous calibration to ensure consistent thresholds
  • Demographic parity across operating conditions

Guardian for Production Monitoring

  • Real-time accuracy tracking for each inspection station
  • Drift detection for model degradation
  • Automated alerts when performance deviates
  • Historical trend analysis for predictive maintenance

Explainability Layer

  • Visual heatmaps showing defect detection reasoning
  • Confidence scores for each inspection decision
  • Audit-ready documentation for every rejection
  • OEM-compatible reporting formats

Calibration Framework

  • Standardized lighting and camera setup validation
  • Cross-plant model synchronization
  • Shift-agnostic performance targets
  • Seasonal variation compensation

Implementation

Phase Scope Duration
Assessment Baseline variance analysis across plants Week 1-2
Root Cause Identification of bias sources Week 3-4
Calibration Model and environment standardization Week 5-8
Deployment Guardian monitoring across all stations Week 9-12
Audit Prep OEM documentation and training Week 13-14

Root Causes Identified

The variance analysis revealed several bias sources:

Issue Impact Resolution
Lighting variation 8% accuracy difference Standardized lighting rigs
Camera calibration drift 5% shift variance Weekly auto-calibration
Training data imbalance Plant-specific bias Cross-plant data augmentation
Threshold inconsistency False reject variance Unified threshold framework

Results

Metric Before After Change
Cross-Plant Variance 15% 2% -87%
Shift Variance 20% 3% -85%
False Reject Rate 8% 6% -25%
OEM Audit Findings 3 major 0 -100%
Inspection Documentation Manual Automated 100%

Customer Audit Success

The next OEM audit resulted in:

  • Zero findings related to AI-based inspection
  • Commendation for explainability and documentation
  • Preferred supplier status retained
  • Benchmark sharing request from OEM for their other suppliers

PLI Compliance

The improved quality consistency supported PLI scheme requirements:

  • Consistent quality metrics across reporting periods
  • Auditable inspection records for government review
  • Reduced rejections supporting production targets
  • Documentation ready for incentive claims

Financial Impact

Category Annual Savings
Reduced False Rejects ₹1.2 Cr
Manual Re-inspection ₹45 L
Audit Preparation ₹20 L
Customer Penalty Avoidance ₹80 L
Total ₹2.65 Cr

Plant Manager Feedback

“Our OEM customers were asking questions about our AI that we couldn’t answer. Now we have complete visibility into why the AI makes each decision. The last audit was the smoothest we’ve ever had.” — Plant Head, Pune Facility

What’s Next

The company is expanding AI trust infrastructure to:

  • Predictive maintenance for CNC machines
  • Supply chain demand forecasting
  • Energy consumption optimization
  • Worker safety monitoring

Rotavision is powered by Rotascale’s globally-proven AI trust platform.

Ready to Achieve Similar Results?

Let's discuss how Rotavision can help your organization.

Schedule a Consultation