Project Summary
- Customer and industry: Forecast Technology Corp. (FATC) / semiconductor packaging and testing (OSAT).
- Equipment: DISCO DGP8761 wafer thinning grinder and polisher.
- Solution: Level 4 online life-prediction solution using the Aurora® 313MM2-R1 smart multivariate sensor and the RULe.TI004 edge AI inference gateway.
- Communication: Sensor to gateway over wired Modbus RS485; gateway to the FDC system over Modbus TCP industrial Ethernet.
One ultra-thin sensor is mounted near the grinding side at the end of each of the Z1 coarse-grinding, Z2 fine-grinding and Z3 dry-polishing spindles. The on-site edge AI gateway estimates the health index (HI) and remaining useful life (RUL) in real time and guides the maintenance team to plan servicing in non-production windows.
Pain Points
Wafer thinning (backgrinding and polishing) determines the final thickness and mechanical strength of the chip. The DISCO DGP8761 grinds to micron precision and polishes to relieve stress at tens of thousands of rpm. The spindles easily become unbalanced or eccentric, and the precision bearings, grinding wheels and polishing pads keep degrading. The financial risk of one sudden failure:
- Wafer scrap: A spindle burn-out or severe eccentricity causing breakage affects one lot of 25 wafers at USD 5,000 each, a direct material loss of USD 125,000.
- Downtime and qualification: The grinder is the only bottleneck of the wafer-thinning stage, so a stop starves the downstream dicing and die-bonding lines. Spindle replacement, chamber recovery tests and quality qualification (QUAL) take about 36 hours, and at USD 5,000 per hour the indirect loss reaches USD 180,000.
- Repair cost: Emergency spare OEM air spindle modules plus urgent repair labor and overtime are about USD 35,000.
- Total: Up to an estimated USD 340,000 (about NT$ 10.8 million) per incident.
The customer relied on the OEM spindle speed feedback, which has very little correlation with bearing wear or slight spindle eccentricity. By the time speed deviates visibly, only hours or even minutes of life may remain. The current mix of scheduled disassembly maintenance (PM) and run-to-failure repair tends toward over-maintenance (wasting good parts and labor) or under-maintenance, and run-to-failure repair carries the cost of spindle burn-out and wafer breakage.
Solution
- Ultra-thin, suited to grinding environments: The Aurora® 313MM2-R1 measures only 21 × 21 × 11.5 mm and fits the tightly spaced spindles inside the DGP8761. Its IP65 aluminum case resists water, dust and salt mist, so it withstands coolant mist, grinding slurry and silicon dust.
- Three sensors, three process steps: The Z1 coarse-grinding spindle monitors heavy-load vibration, the Z2 fine-grinding spindle monitors the high-frequency features of micron-level grinding, and the Z3 dry-polishing spindle monitors the spindle and consumables during polishing and stress relief.
- Zero-interference installation: Installation and wiring need only a short stop during scheduled maintenance (PM), and afterwards the mechanics, speed control and process of the tool are unaffected.
- Layered standard communication: Sensors are wired in parallel over Modbus RS485 to the RULe.TI004 edge inference gateway, which connects through Modbus TCP directly to FDC / SCADA.
- Physical AI and lean data: Signals are cleaned and features computed at the endpoint, reducing 307,200 Bytes/s of raw data to 200 Bytes/s (extraction ratio 651 ppm), a 99.9% smaller data footprint.
Alarm and Maintenance Workflow
- Anomaly sensing and root cause: When high-speed running starts to degrade slightly, the sensor notices abnormal feature energy. The RULe.TI004 fuses three-axis features to diagnose spindle unbalance / eccentricity, precision bearing wear, abnormal wear of the grinding wheel or polishing pad, or breakage risk.
- RUL estimation and proactive alarm: The on-site platform shows HI and RUL trends. When RUL falls below the safety threshold, an alarm goes to the FDC / SCADA control room and a work order is dispatched automatically.
- Planned precision maintenance: After receiving a work order with a clear diagnosis (for example "Z1 coarse-grinding spindle bearing wear, RUL low"), the maintenance team can schedule a stop in an idle or non-production window within the next few days, prepare the spindle, bearings and grinding consumables in advance, and service the tool before failure.
Benefits
- Avoid costly scrap and downtime (ROI): Preventing a single incident saves up to USD 340,000 (about NT$ 10.8 million), so the payback period is very short.
- Higher OEE: "Planned precision maintenance" replaces run-to-failure repair and blind scheduled PM, bringing unplanned downtime close to zero.
- Lower TCO: Ultra-thin miniature sensors shorten installation time, and edge feature extraction saves on-premise IT bandwidth, network and disk storage cost.
(Note: Financial figures are estimated from public industry reports and typical advanced-node production lines.)
Customer Track Record
- Customer: Forecast Technology Corp. (FATC)
- Equipment: DISCO DGP8761 wafer grinder and polisher / OSAT wafer thinning
- Solution level: Level 4 life prediction, with 3 sensors per tool on the Z1, Z2 and Z3 spindles
- Products: Aurora® 313MM2-R1 + RULe.TI004
- Result: Before real damage or breakage of a grinding spindle, wheel or pad, the system can detect high-frequency feature changes on the Z1 coarse and Z2 fine grinding spindles and issue a warning, helping the site team carry out planned maintenance and precise replacement. (Note: No field failure has occurred since deployment, so the warning lead time is not yet proven and depends on the risk threshold FATC can accept.)
Solution
- Aurora Equipment Health Monitoring Platform
- Consulting Service
If you are interested in this solution, please contact your local distributor for more information and a quotation. The full case document (in Chinese) is available below.