The core of optimizing the dual-sensor redundancy comparison algorithm is to upgrade from "simple threshold comparison" to "dynamic trend + weighted fusion," focusing on solving the problems of false alarms and accuracy fluctuations. Specifically, this optimization involves four steps:
1. Introduce sliding window filtering to filter instantaneous interference
Avoid comparing single-point instantaneous values. Instead, collect data from the most recent 5-10 minutes, calculate the moving average of the outputs from both sensors, and then compare them.
This filters out temporary deviations caused by pressure fluctuations and temperature transients during the injection molding and holding phase, avoiding false alarms triggered by instantaneous disturbances.
It is recommended to set the window size to 20-30 sampling points, which smooths out noise without delaying alarm response.
2. Dynamic Adaptive Thresholds Replace Fixed Thresholds
Fixed thresholds cannot adapt to different operating conditions. Instead, they are dynamically adjusted based on production status:
During steady-state production (holding pressure stage): The threshold is set more strictly (e.g., 0.1%FS) to ensure high accuracy.
During dynamic switching (injection/mold opening stage): The threshold is automatically relaxed to 0.2%-0.3%FS to avoid false alarms due to drastic pressure changes.
The system automatically switches the corresponding threshold based on the current process stage of the injection molding machine, significantly reducing the false alarm rate during non-production stages.
3. Add Trend Judgment Logic to Differentiate Between Aging and Disturbances
Simply looking at the magnitude of the deviation can easily lead to misjudgments. A deviation change rate judgment is added:
If the deviation suddenly jumps but then quickly recovers, it is judged as a temporary disturbance, and no alarm is triggered.
If the deviation continues to increase unidirectionally within 24 hours (e.g., an increase of 0.02mA per day), even if the current threshold is not exceeded, an early warning of "slow sensor aging" is issued, enabling predictive maintenance.
4. Employing Weighted Fusion Output to Improve Accuracy Under Normal Operating Conditions
When the deviations of two sensors are within acceptable limits, instead of simple averaging, weights are assigned based on historical stability:
Sensors with longer-term stability are given higher weights (e.g., 0.6:0.4), making the final output closer to the high-precision side;
Weights are recalculated periodically (e.g., weekly) based on the drift of both sensors to ensure the output is always optimal.
This optimization scheme can reduce the false alarm rate by more than 90%, while improving measurement accuracy during normal production by 10%-15%, making it ideal for high-end precision injection molding scenarios.
