BRN Predictive Maintenance System - PHM

Platform core functional architecture:

1. Multi-source full-stack data perception system
   • Inventory and new integration: Upwards, it supports seamless access to the enterprise's existing DCS/PLC system's existing process volume (such as speed, temperature, flow, pressure, current, power, etc.) through protocols such as OPC, Modbus, and TCP/IP; the self-developed hardware such as edge intelligent acquisition terminals and wireless temperature sensors is deployed downwards to collect high-frequency vibration and temperature signals.
   • Data quality cleaning: The system has a built-in intelligent data quality determination and anti-brushing mechanism, which can automatically identify and block invalid range data such as abnormal bias voltage, abnormal dead point, abnormal fluctuation, and excessive range to ensure the true and reliable analysis data from the source.
2. Adaptive AI big data warning engine - Intelligent operating condition modeling: Provides a visual rule editor, and the system can automatically identify that the equipment is currently in different operating states such as “high speed”, “low speed” or “shutdown” based on the logical calculation of multiple process parameters. By configuring the equipment operating condition model to dynamically adjust the warning threshold, false alarms and omissions caused by changes in operating conditions are effectively prevented.
   • AI residual detection: integrates deep learning frameworks such as TensorFlow, PyTorch, Keras, etc., with more than 200 built-in machine learning algorithms. Through data cleaning, feature extraction, and dimensional reduction (fusion of multi-domain features such as FFT, FRFT, WT, etc.) and regression training, intelligent detection of abnormal device trends is realized.
3. Mechanism rule library and expert precision diagnosis map - Motor and mechanical fault rule library: built-in complete expert mechanism rules, covering dozens of fault modes such as rotor unbalanced/misaligned, structural looseness, bearing pitting damage, gear wear, and fixed rotor eccentricity, rotor strip failure, and fluid-like water pump cavitation and fan surge.
   • Expert analysis maps: The platform provides powerful professional in-depth analysis tools, supports time-domain waveforms, spectrum maps, envelope demodulation spectra (extracting early fine pulses), and inverted spectrum; provides step spectrum to specifically solve frequency blurring caused by changes in the speed of frequency conversion equipment; and supports axial trajectories, bird diagrams, and cascaded waterfall maps for start-stop transient analysis.
   • Stress wave ultra-early damage analysis: Stress wave analysis techniques are introduced to extract the three characteristic indicators of SWE, SWPE, and SWPA through stress wave energy maps, time domain maps, vibration histograms, and spectral maps to capture damage characteristics at a very early stage of friction or impact.
4. Full-link intelligent operation and maintenance and visual management - Closed loop elimination management: supports the mobile terminal (inspection assistant) and computer terminal to connect to the closed-loop process of “status monitoring - intelligent diagnosis - alarm enhancement (multiple alarms uniformed/alarm suppression) - generation of work orders - acceptance and assignment - spare parts maintenance - completion confirmation”.
   • Multi-level health portrait: Provide group-level/factory-level large screens and large screens for single device monitoring. The “equipment health score” is automatically calculated using dynamic weight and dynamic score methods, fault patterns are intuitively located through radar maps and trend maps, asset risk is assessed, and regular and immediate equipment diagnosis reports are automatically generated.
   • Enterprise equipment knowledge base: built-in fault case library, mechanism model library, bearing library for various manufacturer models and knowledge base of various documents, supports users to independently create folders to upload maintenance data to achieve the accumulation and inheritance of enterprise industrial knowledge.