How to Optimize the False Alarm Rate of Dual-Sensor Redundancy Comparison

The key to maximizing dual-sensor redundancy comparison's false alarm rate is "distinguishing between real faults and operational interference." The false alarm rate can be lowered to less than 5% by making fine modifications in the following four dimensions:

 

1. Set dynamic thresholds for the most crucial step of the process. A set threshold will unavoidably result in false alarms because the pressure changes significantly during injection molding. The injection, holding pressure, cooling, and mold opening status signals from the injection molding machine must serve as the foundation for the automatic switching logic:

Pressure variations are significant during the High-Pressure Dynamic Period (Injection/Holding Pressure). A 3–5 second delay confirmation is enabled, and the deviation threshold is expanded to 0.2%–0.3%FS. Only when the deviation keeps going beyond the limit-filtering out immediate effects-is an alarm set off.

Pressure is steady during the low-pressure steady-state period (cooling, mold opening, and zero pressure). To precisely capture zero-point drift and small anomalies, the threshold is tightened to 0.05%–0.1%FS.

changeover Transition Period: To prevent false triggering owing to rapid deviations brought on by abrupt pressure changes, alarm judgment is momentarily deactivated during the first two seconds of process stage changeover.

 

2. Improve the Data Filtering Algorithm to Remove Noise

The primary cause of false alarms is high-frequency noise in the original signal, which must be preprocessed before comparison:

Combined Filtering: Use moving average filtering (window size 10–20 points) to smooth out random fluctuations after using median filtering to eliminate abrupt spikes (like electrical interference);

Differential Suppression: Determine how quickly the two sensors' differences are changing. The difference is considered transitory interference and is not logged as a defect if it jumps but then swiftly returns to normal.

 

3. Create a "Trend + Amplitude" dual criterion.

To prevent sporadic anomalies setting off alarms, the persistence of the deviation should be taken into account in addition to its magnitude:

Duration Determination: Before sounding an alert, the deviation must be sustained for N seconds (for example, 10 seconds) or N sampling periods once it over the threshold;

Trend Slope Determination: Keep an eye on the deviation change's slope. If it is a strong oscillation (an interference feature), ignore it or just log it; if it is a moderate linear increase (an aging characteristic), issue an early warning but do not instantly shut down.

 

4. Frequent Dynamic Weight Adjustment and Synchronous Calibration

The two sensors' varying rates of aging over time may result in higher reference deviation and false warnings.

Automatic Zero-Point Synchronization: The two sensors' zero-point references are automatically realigned under zero-pressure circumstances following each power-on and temperature stabilization.

Dynamic Weight Update: The stability of the two sensors is assessed every a certain number of cycles (e.g., 5000 cycles), and the more stable sensor is assigned a larger weight to minimize overall deviation changes brought on by a single sensor's modest drift.

By eliminating more than 90% of false alarms caused by interference from operational conditions, our integrated approach guaranties that alarms only target actual sensor problems.