Project Summary
- Customer and industry: Forecast Technology Corp. (FATC) / semiconductor packaging and testing (OSAT).
- Equipment: DISCO DFD6362 wafer dicing saw (Yunlin fab).
- Solution: Level 4 online life-prediction solution using the Aurora® 313MM2-R1 smart multivariate sensor and the RULe.TI002 edge AI inference gateway.
- Communication: Sensor to gateway over wired Modbus RS485; gateway to the FDC system over Modbus TCP industrial Ethernet.
Ultra-thin sensors are mounted at the ends of the two high-speed cutting spindles next to the blade holder. They estimate the machine health index (HI) and remaining useful life (RUL) in real time and warn before failure, so the site can carry out planned maintenance in non-production windows.
Pain Points
Wafer dicing is the key step that separates a whole wafer into individual dies. The DISCO DFD6362 cuts to micron precision at tens of thousands of rpm. The high-speed spindle easily develops eccentricity or unbalance, and its precision bearings and cutting blades suffer high-frequency wear. The financial risk of one sudden failure:
- Wafer scrap: A spindle burn-out or wafer breakage affects one lot of 25 12-inch advanced-node wafers. At USD 5,000 per wafer after packaging and test, the direct material loss reaches USD 125,000.
- Downtime and qualification: Replacing the spindle, running chamber recovery tests and completing quality qualification (QUAL) takes about 24 hours on average. At USD 5,000 of output value per hour, the indirect loss reaches USD 120,000.
- Repair cost: Emergency replacement of the OEM air spindle module plus urgent repair labor and overtime is about USD 25,000.
- Total: Up to an estimated USD 270,000 (about NT$ 8.6 million) per incident.
The customer relied on the OEM spindle speed feedback. Speed variation has very little correlation with early bearing wear or slight spindle eccentricity, so early anomalies go unnoticed. By the time speed shows an obvious deviation, the structure is often close to burn-out or seizure, with only hours or even minutes of life left, too late to prepare parts and schedule maintenance.
Solution
- Ultra-thin, resistant to mist and dust: The Aurora® 313MM2-R1 measures only 21 × 21 × 11.5 mm and fits the very tight spindle housing. Its IP65 aluminum case withstands coolant mist and dust from cutting. Installation and cabling need only a short stop during routine maintenance, and afterwards the sensor has zero interference with the OEM tool and the process.
- Layered standard communication: Sensors are wired in parallel over Modbus RS485 to the RULe.TI002, which connects through Modbus TCP directly to the customer's FDC / SCADA with no complex software conversion.
- Physical AI diagnosis: Signals are cleaned and features computed at the endpoint, then HI and RUL are estimated in real time to identify spindle unbalance / eccentricity, bearing wear, blade wear or breakage risk.
- 99.9% less data: High-frequency raw data of 307,200 Bytes/s (76.8 kHz sampling) is reduced at the sensor to 200 Bytes/s of lightweight features (extraction ratio 651 ppm), saving cleanroom bandwidth and on-premise storage.
Alarm and Maintenance Workflow
- Anomaly sensing and multivariate diagnosis: When the spindle develops slight eccentricity, bearing wear or blade degradation, the sensor captures the feature energy and the RULe.TI002 AI engine dynamically selects features to identify the failure mode.
- RUL estimation and proactive alarm: The on-site platform shows HI and RUL trends. When RUL falls below the safety threshold, an alarm is sent over Modbus TCP to the FDC control room and a maintenance work order can be dispatched at the same time.
- Planned on-site maintenance: The maintenance supervisor receives a work order with a specific diagnosis (for example "Spindle 2 bearing wear, RUL low"). No emergency stop is needed. The team schedules a non-production or changeover window within the coming days, prepares the spindle and bearings in advance and replaces them before failure.
Benefits
- Avoid costly scrap and downtime (high ROI): Preventing a single incident saves up to USD 270,000 (about NT$ 8.6 million), so the payback period is very short.
- Higher OEE: Moves from "repair after breakdown" or "blind scheduled maintenance" to planned precision maintenance, bringing unplanned downtime close to zero.
- Lower TCO: Small sensors and parallel RS485 wiring make installation fast, and edge feature extraction saves on-premise IT network, bandwidth 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 DFD6362 wafer dicing saw / Yunlin fab
- Solution level: Level 4 life prediction
- Products: Aurora® 313MM2-R1 + RULe.TI002
- Result: A warning was issued before the spindle failed. High-speed spindle eccentricity and bearing wear were diagnosed precisely, the maintenance team scheduled the replacement, and a loss of up to NT$ 8.6 million (USD 270,000) in scrap and downtime was avoided.
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.