What are some commonly used dual-sensor redundancy comparison algorithms?

Three types of common dual-sensor redundancy comparison algorithms may be distinguished: basic introduction type, optimized advanced type, and intelligent adaptive type, which can adapt to different levels of complexity. They have the following particular traits:

 

1. Fixed Threshold Direct Comparison Method: Basic Introductory Type

The most traditional initial algorithm, the easiest to use, and appropriate for straightforward retrofits in small and medium-sized factories:

Core Logic: Establish a defined deviation threshold; an alarm is immediately set off if the two sensors' outputs deviate by more than the threshold. There is no need for more reasoning.

Benefits: Requires little development work and can be written directly to pre-existing PLCs without complicated settings.

Cons: Very high false alarm rate; unable to adjust to process changes and sensor aging; limited to situations with minimal dependability needs; presently being phased out.

 

2. Staged Dynamic Threshold Comparison Method: Optimized Advanced Type

The widely used algorithm in industrial settings nowadays, appropriate for injection molding and other periodic production scenarios:

Core Logic: Set distinct thresholds at each stage of the process; loosen the threshold during dynamic phases like injection, start, and stop; tighten the threshold during steady-state production stages; and incorporate dual filtering and delayed confirmation to reduce interference.

Benefits: The false alarm rate is decreased by more than 70% when compared to a fixed threshold; it is easy to implement, doesn't require sophisticated processing power, and can be configured by regular engineers.

Applicable Scenarios: Currently the most popular mainstream solution for detecting redundancies in injection molding and process industries.

 

3. 3σ Adaptive Threshold Update Method for Advanced Optimization

reduces long-term false alarms even more by giving the dynamic threshold the capacity to react automatically:

Core Logic: Every week, historical deviation data is gathered, and the 3σ principle is used to determine the deviation's usual fluctuation range. The threshold is automatically adjusted to account for drift brought on by aged sensors.

Benefits include: 30% fewer false alarms than with a set dynamic threshold; strong long-term operational stability; and no need for human periodic threshold change.

Relevant Situations: Mass production lines that run continuously and require less manual maintenance.

 

4. Intelligent Adaptive Type: Exponentially Weighted Baseline Drift Compensation Method (EWMA)

An intelligent optimization technique that is very good at managing gradual sensor aging in high-end scenarios:

Core Logic: Automatically corrects for the output of the aging sensor by using Exponentially Weighted Moving Average (EWMA) to adjust the baseline deviation between two sensors in real time. Only when the deviation above the correction range does an alarm sound.

Benefits include automatic gradual aging, a very low long-term false alarm rate, an extended sensor replacement cycle, and lower maintenance costs.

Cons: Requires some processing, but it is easy to construct and can be fully supported by current PLCs.

 

5. Advanced Intelligent Type: Anomaly Detection Technique in Machine Learning

A high-end production line solution that is entirely automated:

Core Logic: Uses historical normal data to train models like Isolation Forest and LSTM, automatically learns the normal deviation range under various processes, and instantly detects abnormal deviations.

Benefits: Extremely flexible, little false alarm rate in intricate and dynamic operations;

Cons: Currently exclusively utilized in high-end medical and aeronautical scenarios; requires a lot of historical data for training; requires somewhat more processing power; is not cost-effective for small and medium-sized industries.