Reducing the false alarm rate requires strict control at four stages: hardware selection and installation, signal processing, algorithm logic, and post-maintenance. Following this approach can stably control the false alarm rate below 3%:
1. Source Control: Eliminating Basic Deviations in Hardware and Installation (Solving 60% of False Alarms)
Strict Pairing Selection: Sensors from the same manufacturer, model, and batch must be selected to ensure consistent initial accuracy and aging characteristics. Initial calibration must meet the following requirements: zero-point deviation ≤ 0.03%FS, full-scale deviation ≤ 0.05%FS. Replace any sensors that do not meet these standards.
Standardized Installation to Control Accuracy: Ensure the coaxiality of the two sensors is ≤ 0.05mm and the height difference of the sensing surfaces is ≤ 0.02mm. Tighten with the same torque. After installation, allow the system to heat to operating temperature and stabilize for 2 hours to release stress before recalibrating. Install the sensors side-by-side (with a spacing ≤ 10mm) to avoid differences in operating conditions.
2. Signal Preprocessing: Filtering Interference to Reduce False Alarms (Further Reduce False Alarms by 20%)
Dual Filtering Combination: First, median filtering is applied to remove instantaneous spikes caused by injection molding shock and electrical interference; then, moving average filtering is applied to smooth the signal and eliminate random fluctuations.
Outlier Removal: Individual sampling points exceeding the 3σ range of historical fluctuations are directly removed and not included in the comparison calculation.
Synchronous Temperature Compensation: Temperature compensation is applied to both sensors to offset output drift caused by temperature differences.
3. Algorithm Logic Optimization: Adapting to Operating Conditions to Avoid Unreasonable Judgments (Further Reducing False Alarms by 15%)
Avoiding fixed thresholds, dynamically switching thresholds according to process stages: For example, in injection molding, the threshold is relaxed to 0.2-0.3%FS during the injection stage, tightened to 0.05-0.1%FS during the cooling stage, and amplified to 1.5 times the threshold during mold opening and closing to filter vibration interference;
Adding delayed confirmation: An alarm is only triggered if the deviation exceeds the limit for ≥10 seconds; a momentary exceedance followed by recovery is directly judged as interference;
Replacing single amplitude judgment with trend judgment: Alarms are only triggered for continuously increasing deviations; occasional fluctuations are not triggered, adapting to the slow aging characteristics of sensors;
Regularly and automatically updating thresholds: A new threshold is calculated weekly using 3σ of historical deviations, automatically adapting to drift caused by sensor aging, eliminating the need for manual adjustment.
4. Regular Maintenance: Preventing Accumulated Aging Deviations (Resolving the Remaining 5% of False Alarms)
Automatic Zero-Point Alignment Upon Each Power-On: Automatically corrects the zero-point deviation of the two sensors under zero-pressure steady-state conditions, eliminating minor drifts from the previous day;
Manual Full-Scale Calibration Every 3 Months: Realigns sensitivity, offsetting deviations caused by different aging rates of the two sensors;
If the deviation approaches the threshold, the FS can be slightly relaxed by 0.03-0.05% to reduce false alarms without affecting detection accuracy.
