Eliminating In-Line Rework to Cut Cycle Time by 6% in EV Motor Assembly

How a global tier-1 electric vehicle powertrain supplier used Linecraft AI’s multi-modal cycle clustering to unmask silent in-line rework cycles on automated welding cells—driving a 6% cycle time reduction and elevating First-Time Quality to 99.4%.

Customer Background & Manufacturing Context

Electric vehicle traction motors require extraordinary precision. In modern hairpin stator production, hundreds of copper wire segments are inserted into laminated stator slots, twisted, and welded at high speed. The resistance welding station forms the structural and electrical backbone of the entire motor. Because this station operates in a rigid serial line, any slowdown directly throttles downstream varnish impregnation, rotor insertion, and final testing.

The customer—a major supplier to North American and European electric passenger vehicle brands—was ramping production on a high-volume stator line. However, line throughput persistently lagged behind planned targets. Machine availability was high and major alarms were rare, but the welding station consistently failed its hourly quota.

The Operational Challenge: Silent In-Line Rework

The customer's internal engineering team attempted standard time-motion analyses and reviewed SCADA alarm reports, but the data presented an enigma:

• No Downtime Registered: The machine rarely faulted out or halted. It was continuously running, which led plant management to conclude that the robot's physical motion path was simply too slow.
• Hidden Secondary Cycles: What the SCADA system failed to report was that the machine controller was programmed with an automated "retry" routine. Whenever an optical sensor or resistance threshold failed on the first weld pulse, the robot automatically repositioned the clamp and attempted a secondary weld sequence.
• The Hidden Capacity Drain: Because these retries happened inside the cycle, operators never touched the machine and maintenance never received an alarm. Yet each retry inflated cycle duration by 15 to 22 seconds, silently bleeding 18% of the station's theoretical output.

"The machine wasn't breaking down—it was quietly doing double the work on every third stator without telling anyone."

How Linecraft AI Solved the Mystery

The customer connected Linecraft Ikshana directly to the welding cell's programmable logic controller and multi-channel weld monitor. Ikshana ingested every individual weld step: pneumatic clamp extension, pre-squeeze time, primary current firing, resistance feedback curve, post-squeeze cooling, and index rotation.

Using advanced distribution clustering algorithms, Ikshana segmented thousands of production cycles into statistical archetypes, revealing the exact anatomy of the performance loss.

Data Discovery: Unmasking the 30% Rework Cluster

The cycle time histogram generated by Linecraft provided unequivocal mathematical proof of the problem:

Figure 1: Linecraft Ikshana cycle histogram analysis. The sharp left cluster represents nominal single-pass welds (70% of cycles). The broad anomalous cluster on the right highlights in-cycle re-welding routines responsible for the throughput deficit.

Linecraft’s deep-dive diagnostics established:

• 70% Nominal Execution: Stators with pristine hairpin alignment completed the full 24-point weld sequence in 38 seconds, well within takt time.
• 30% Anomalous Rework: In 3 out of every 10 stators, slight dimensional spring-back in incoming copper hairpins caused initial contact resistance to breach tolerance. This automatically triggered an in-cycle re-clamp and secondary weld sequence, pushing total cycle time past 55 seconds.
• Cumulative Throughput Penalty: These silent re-welds accounted for 100% of the customer's line capacity deficit.

Customer Success Interventions: DFM & Tooling Upgrades

With precise metallurgical and mechanical root causes identified, the customer engaged cross-functional teams to implement permanent fixes:

• Design for Manufacturing (DFM) Spec Tightening: Collaborated with the copper hairpin forming supplier to tighten bend radius tolerances and eliminate spring-back variation.
• Adaptive Electrode Pressure Control: Modified the PLC pneumatic valve firing profile to apply progressive clamping force, ensuring stable surface contact on the initial squeeze.
• Automated Tip Dressing Integration: Programmed periodic automated electrode tip conditioning after every 250 cycles to prevent copper oxide buildup from skewing resistance readings.

Quantified Customer ROI: 6% Cycle Time Reduction

Within two weeks of implementing the DFM and PLC adjustments, the secondary rework cluster completely vanished from Linecraft's real-time dashboards:

Key Takeaways for Manufacturing Leaders

• Watch Out for Silent In-Cycle Retries: Modern smart automation frequently hides internal retry routines. If a machine never faults but never hits rate, in-line rework is almost always occurring.
• Multi-Modal Histograms Expose Hidden Modes: A normal distribution means clean operation. A bimodal or multimodal distribution is incontrovertible evidence of distinct, unmanaged operational states.
• Bridge Quality and Performance: Cycle time and product quality are two sides of the same coin. Solving dimensional and quality defects is often the fastest route to unlocking machine speed.