Applications and Occasions
- Move from "repair after failure" to maintenance scheduled by prediction: avoid unplanned downtime and cut spare-parts inventory and emergency repair costs.
- Steady rotating equipment such as motors, pumps, fans, gearboxes, reducers and vacuum pumps, and non-steady equipment such as robot arms and handling mechanisms.
- Lines and plants that already use, or plan to use, Level 3 sensors and need to answer "how long can the machine keep running?"
- Integrates with in-house SCADA / FDC / PHM systems as the data source for equipment health management.
Level 4 answers one question: how long can the machine keep running (remaining useful life, RUL).
Setup and Usage
- Install Level 3 sensors: fix 213MM2-R1 / 313MM2-R1 (RS485) or 213MM1-B1 (Bluetooth) on the measurement points and enter machine parameters such as speed range, bearing model, gear teeth and blade count so the sensor can compute damage-mode features.
- Check the connection: for cable runs over 50 m add a 125 ohm terminating resistor at the end; power the sensors with 12 - 24 V; when several sensors share one COM port their station numbers must differ; if you use a COM port on the industrial PC, make sure it is set to RS485. In the software always use the sensor ID as both equipment serial number and sensor serial number.
- Collect data and filter operating conditions: the data-collection module receives features every second, and an idle-detection threshold removes stopped and standby data so only real running data remains. Look at the data before setting the threshold; covering 3 - 5 running cycles is enough.
- Train the health (HI / EHI) model: build a baseline from normal running data; the deviation of new data from that baseline is the health index. At least 14 days of running data is recommended, and the model must be retrained after maintenance.
- Train the remaining useful life (RUL) model: train on the health-index history, ideally about 30 days of data after health inference starts; the prediction horizon is about half of the history length (for example 30 days of data predicts the next 15 days).
- Online inference and viewing: view health, damage-mode feature contribution (which feature causes the anomaly) and remaining life on the dashboard; results can be sent to in-house SCADA / FDC / PHM systems to schedule maintenance.
- Retrain after maintenance: after repair or part replacement, rebuild the health and life models; disable data collection during maintenance so abnormal data is not learned.
Life estimation software: RULe
Level 4 uses RULe life-estimation software, running on an edge AI gateway (industrial PC).
| Item | RULe |
|---|---|
| Position | Next-generation life estimation: device management, data collection, training, inference and visualization |
| Modeling | Build an EHI health baseline first, then train the RUL model on the EHI trend |
| Data needed | EHI: at least 14 days of running data; RUL: about 30 days after EHI inference starts |
| Preprocessing | Operating-condition filter (idle detection) removes stopped data |
| Output | Health index HI, damage-mode feature contribution, RUL |
| Features | Multi-device, multi-sensor management; single-unit or separate training / inference architecture |
| Interface | Web-based (data-collection system and RULe dashboard) |
Industrial PC specifications
RULe runs on an industrial PC (IPC) in six configurations by sensor count: 2 / 4 / 8 channels (single unit) and 16 / 32 / 64 points (separated training / inference).
| Item | 2-ch | 4-ch | 8-ch | 16 points | 32 points | 64 points |
|---|---|---|---|---|---|---|
| Model | RULe.TI002 | RULe.TI004 | RULe.TI008 | RULe.XI016 | RULe.XI032 | RULe.XI064 |
| CPU | Core i5 | Core i5 | Intel Core i7-14700T | Intel Core i3-N305 | Intel Core i3-N305 | Intel Core i3-N305 |
| Memory | DDR5 16 GB | DDR5 32 GB | DDR5 64 GB | 8 GB | 8 GB | 8 GB |
| Storage | SSD 512 GB | SSD 1 TB | SSD 2 TB (2 years of data retention) | 64 GB | 64 GB | 64 GB |
| RS232 / 485 / 422 serial ports | 2 | 2 | 2 | - | - | - |
| Ethernet | 2 | 2 | 2 | 2.5GbE x 2 | 2.5GbE x 2 | 2.5GbE x 2 |
| Power | 9 - 36 V | 9 - 36 V | 10 - 35 V | 3-pin DC-in 9 - 36 V | 3-pin DC-in 9 - 36 V | 3-pin DC-in 9 - 36 V |
| IPC size | 134 x 105 x 65 mm | 134 x 105 x 65 mm | 180 x 77.2 x 220 mm | 214 x 115.5 x 65.8 mm | 214 x 115.5 x 65.8 mm | 214 x 115.5 x 65.8 mm |
| Display output | HDMI | HDMI | HDMI | - | - | - |
| USB | 6 (2 x USB 3.0 + 4 x USB 2.0) | 6 (2 x USB 3.0 + 4 x USB 2.0) | 6 (4 x USB 3.0 + 2 x USB 2.0) | 6 | 6 | 6 |
| Operating system | Linux Ubuntu 24.04 | Linux Ubuntu 24.04 | Linux Ubuntu 24.04 | - | - | - |
"-" means no data yet; please contact us for details.
For more points (16 / 32 / 64) use the separated training / inference architecture: the RULe Life Prediction Inference Gateway with an AI training server. For 2 / 4 / 8 points a single unit, the RULe Life Prediction (Training and Inference) Gateway, is available.
Sensors and software
Sensors (Level 3, support HI / RUL)
- 213MM2-R1, 313MM2-R1: RS485, built-in damage-mode diagnosis and custom band features
- 213MM1-B1: Bluetooth wireless, needs a Bluetooth collector
- For sensor selection see Level 3 solution
Software and accessories
- Life estimation software: RULe
- Data collection / middleware and device management: RULe, IOTech
- Edge AI gateways: RULe Life Prediction (Training and Inference) Gateway, RULe Life Prediction Inference Gateway
- For specifications and quotations please contact us