| I&T Solution |
Development of Artificial Intelligence (AI) Model for Predictive Maintenance of Air Handling Units (AHUs)
(REF: S-1633) |
| Trial Project |
|
| Solution Feature |
- The solution comprises of installation of vibration sensors to monitor the vibration of AHU equipment. Based on ISO-10816, the vibration level reflects the equipment health.
- Aside from obtaining overall vibration level to indicate general equipment health, the time series domain can be changed to frequency domain (via FFT) to analyze for specific faults in various components.
- Using FFT to train AI model which can identify potential shaft misalignment, unbalance, loose foot, bearing defect and worn belt
- Provide HMI for maintenance personnel to monitor the equipment health as well as receive notifications and alarms which may indicate potential fault
- Can be designed to integrate with EMSD IBMS or onsite BMS
|
| Trial Application and Expected Outcome |
- The vibration sensors will be installed on an AHU, installation points and methods will be identified on-site.
- The sensors will link to a data logger which aggregates and in turn passes data to a central server which performs the machine learning and data storage.
- Supervised AI learning will be adopted by importing samples of different types of FFTs representing different faults. These faulty FFTs are either generated by a faulty equipment or tweaking the equipment’s normal FFT to reflect the fault.
- Once the AI model is trained, it will be deployed on either running AHU or the test bed AHU equipment to test for its accuracy. Ongoing fine tuning will be performed to ensure the accuracy of the model.
- Ultimately, the AI should be able to pinpoint (based on FFT spectrum), potential shaft misalignment, unbalance, loose foot, bearing defect and worn belt
|
| Additional Solution Information |
Nam Wah Proposal for AHU Predictive Maintenance W-0517.pdf
|
| Info on I&T Solution Provider |
|
