How to Optimize the Dual-Sensor Redundancy Comparison Algorithm

Upgrading from "simple threshold comparison" to "dynamic trend + weighted fusion," which focuses on resolving the issues of false alarms and accuracy variations, is the fundamental step in enhancing the dual-sensor redundancy comparison method. In particular, there are four processes involved in this optimization:

1. To filter instantaneous interference, implement sliding window filtering.

Do not compare instantaneous single-point values. Instead, gather information from the last five to ten minutes, compute the moving average of the two sensors' outputs, and compare them.

This prevents false alarms brought on by sudden disruptions by filtering out transient deviations brought on by pressure changes and temperature transients during the injection molding and holding phase.

To reduce noise without slowing down the alarm response, it is advised to set the window size to 20–30 sample points.

 

2. Fixed thresholds are replaced by dynamic adaptive thresholds

Different operating conditions cannot be accommodated by fixed thresholds. Rather, they are dynamically modified according to the state of production:

To guaranty excellent precision during steady-state manufacturing (holding pressure stage), the threshold is set more precisely (e.g., 0.1%FS).

In order to prevent false alarms caused by abrupt pressure changes, the threshold is automatically reduced to 0.2%–0.3%FS during dynamic switching (injection/mold opening stage).

The technology dramatically lowers the false alarm rate during non-production stages by automatically switching the matching threshold based on the injection molding machine's current process stage.

 

3. Use Trend Judgment Logic to Distinguish Between Disturbances and Aging

Just focusing on the deviation's size can quickly result in incorrect conclusions. There is an additional deviation change rate judgment:

The deviation is considered a transient disturbance and no alarm is set off if it abruptly jumps but then swiftly recovers.

Even if the current threshold is not surpassed, an early warning of "slow sensor aging" is sent, allowing predictive maintenance, if the deviation increases unidirectionally within 24 hours (e.g., an increase of 0.02mA per day).

 

4. Using Weighted Fusion Output to Increase Accuracy in Typical Operating Situations

Weights are assigned based on past stability rather than just averaging when two sensors' variances fall within acceptable bounds:

Higher weights (e.g., 0.6:0.4) are assigned to sensors with longer-term stability, bringing the final output closer to the high-precision side;

To guaranty that the output is always ideal, weights are updated on a regular basis (e.g., weekly) based on the drift of both sensors.

 

This optimization approach is perfect for high-end precision injection molding scenarios because it may lower the false alarm rate by over 90% while increasing measurement accuracy during regular production by 10%–15%.

1328-915