What data metrics are needed for a dual-sensor redundancy comparison experiment?

Algorithm performance and engineering feasibility are the two main data metrics for a dual-sensor redundancy comparison experiment. To guaranty that no important information is missed, they are separated into three categories: engineering practical metrics, auxiliary verification metrics, and core mandatory metrics.

I. Core Mandatory Metrics (which directly determine performance quality and must be calculated)

In order of priority, these four measures form the basis for evaluating the quality of algorithms:

Name of Metric

Method of Definition and Calculation

Physical Significance

Pass/Fail Criteria

Rate of False Alarms

The quantity of false alarms Total alarms × 100%

evaluates the algorithm's capacity to prevent needless downtime and filter interference, which is especially important in industrial settings.

≤5%, Outstanding ≤2%

Rate of False Alarms

The total number of real defects × the number of missed real faults × 100%

evaluates how well the algorithm detects actual errors.

≤1%, Excellent = 0%

Response Latency Average

Average time for all actual problems from occurrence to alarm

determines how quickly the algorithm responds.

≤15 seconds, Outstanding ≤5 seconds

Change Rate of False Alarm Rate

After three months of operation, the false alarm rate

The long-term stability of the algorithm's adaptive aging is measured by the absolute magnitude of the first false alert rate.

≤2%, outstanding ≤1%

II. Auxiliary Verification Metrics (for a more thorough analysis of algorithm properties)

The following optional metrics are used for detailed examination of variations in algorithm performance:

The standard deviation of all deviations that occurred during the steady-state phase is known as the standard deviation of deviation. The algorithm's output is more stable and has a higher anti-interference capability when the deviation is smaller.

Maximum Deviation: The maximum deviation value under typical operating conditions, which is used to assess the reasonableness of the threshold setting.

Detection Rate: 1 - False Negative Rate = Count of actual fault detections × Total number of actual faults × 100%. Calculations can be made using either the false negative rate or the false negative rate.

The algorithm's overall accuracy is measured as follows: (True alarms + True normal detections) ÷ Total number of detections × 100%.

III. Useful Engineering Indicators (Must Be Considered When Determining Viability) Standards for performance must be met, and usability is just as crucial because these metrics affect future maintenance expenses:

Computing Power Utilization: The PLC controller's CPU utilization rate following algorithm execution. This shouldn't go above 30% in industrial settings to prevent interfering with other operations.

Debugging Time: The overall amount of time spent debugging between installation and production. Easier implementation is indicated by shorter times.

Maintenance Frequency: The frequency at which parameters must be manually modified or calibrated each month. Lower maintenance costs result from fewer times.

problem Location Accuracy: The manual verification accuracy rate following the algorithm's detection of a sensor problem. Increased precision makes troubleshooting easier.

Just four key indications are required for basic verification: false alarm rate, missed alarm rate, average reaction delay, and false alarm rate change rate. Additional auxiliary and engineering indicators can be incorporated for algorithm development or solution selection.