How to Reduce the False Alarm Rate of a Dual-Sensor Redundant Comparison System?

Strict control at four stages-hardware selection and installation, signal processing, algorithm logic, and post-maintenance-is necessary to lower the false alarm rate. By using this strategy, the false alarm rate can be consistently kept below 3%:

1. Source Control: Resolving 60% of False Alarms by Removing Fundamental Deviations in Hardware and Installation

Strict Pairing Selection: To guaranty constant starting accuracy and aging characteristics, sensors from the same manufacturer, model, and batch must be used. Zero-point deviation < 0.03%FS and full-scale deviation ≤ 0.05%FS are prerequisites for first calibration. Any sensors that don't adhere to these requirements should be replaced.

Standardized Installation to Control Accuracy: Make sure the height difference between the sensing surfaces is ≤ 0.02mm and the coaxiality of the two sensors is ≤ 0.05mm. Use the same torque to tighten. Prior to recalibrating, let the system reach operational temperature and stabilize for two hours. To prevent variations in operation circumstances, install the sensors side by side (with a distance of < 10mm).

2. Signal Preprocessing: Reducing False Alarms by 20% by Filtering Interference

Dual Filtering Combination: Moving average filtering is used to smooth the signal and get rid of random fluctuations after median filtering has eliminated immediate spikes brought on by injection molding shock and electrical interference.

Outlier Removal: Sampling points that surpass the 3σ range of past variations are immediately eliminated and excluded from the comparison computation.

Synchronous Temperature Compensation: To counteract output drift brought on by temperature variations, temperature compensation is provided to both sensors.

3. Algorithm Logic Optimization: Reducing False Alarms by 15% by Adjusting to Operating Conditions to Prevent Unreasonable Judgments

avoiding fixed thresholds and dynamically changing them in accordance with the stages of the process; for instance, in injection molding, the threshold is amplified to 1.5 times the threshold during mold opening and closing to filter vibration interference, tightened to 0.05-0.1%FS during the cooling stage, and relaxed to 0.2-0.3%FS during the injection stage;

Adding delayed confirmation: A brief exceedance followed by recovery is immediately considered interference; an alarm is only set off if the deviation surpasses the limit for ≥10 seconds;

Trend judgment is used in place of single amplitude judgment; alarms are only sent off for consistently rising deviations; sporadic fluctuations are not set off in order to accommodate the slow aging features of sensors;

Frequently and automatically updating thresholds: Every week, a new threshold is computed using 3σ of past deviations. This eliminates the need for manual modification by automatically adjusting to drift brought on by sensor aging.

4. Consistent Upkeep: Preventing Accumulated Aging Deviations (Taking Care of the Final 5% of False Alarms)

Automatic Zero-Point Alignment Upon Every Power-On: Removes small drifts from the previous day by automatically correcting the two sensors' zero-point deviation under zero-pressure steady-state settings;

Every three months, a manual full-scale calibration is performed to realign sensitivity and correct for variations brought on by the two sensors' varying rates of aging.

To lower false alarms without compromising detection accuracy, the FS can be slightly relaxed by 0.03–0.05% if the deviation is close to the threshold.