The total experiment duration varies depending on the verification objective, typically ranging from 14 to 90 days. A specific schedule can be found below:
1. Basic performance verification (only verifying that the new algorithm is better than the old one): Total duration 14-21 days. Routine production anti-interference testing: Run continuously for 14 days, covering at least 5000 production cycles. The sample size is sufficient to eliminate random fluctuations and obtain stable false alarm rate data. Fault simulation testing: Can be interspersed within routine testing, requiring an additional 2-3 days to complete; it does not need to take up too much extra time. Suitable scenarios: Engineering transformation in small and medium-sized factories, upgrading and verifying existing solutions. This duration is sufficient to obtain reliable conclusions.
2. Algorithm Development/Module Decomposition and Verification: Total Duration 30-45 Days. Each experimental group requires at least 10-14 days of testing. With 2-3 experimental groups for module verification plus a baseline control group, the total duration exceeds 30 days. The total duration can be compressed through "sequential testing": run different groups sequentially on the same hardware, 10 days for each group, totaling 30 days for 3 groups. This saves hardware costs compared to parallel testing.
3. Long-Term Stability Verification (Verifying Adaptive Aging Capability): An Additional 90 Days. If it is necessary to verify the algorithm's adaptation to sensor aging, after basic testing, it needs to run continuously without adjustments for 90 days. The false alarm rate should be calculated monthly, and the rate of change in the false alarm rate should be measured to verify long-term stability.
If only feasibility verification is needed and long-term stability data is not required, this step can be omitted, and simulated aging tests can be used instead. The total duration should still be kept within one month.
Quick Selection Reference:
Table Verification Objective Recommended Total Duration
Engineering Modification, confirming the new algorithm is better than the old one 14 days
Algorithm Development, analyzing module effects 30 days
Formal Product Finalization, verifying long-term reliability 14 days basic testing + 90 days long-term testing = 104 days
The testing duration should not be less than 14 days. Insufficient sample size will lead to statistical errors and unreliable results. For slow-variable process industries, it is recommended to extend the testing duration for each group by 3-5 days to ensure sufficient coverage of operating cycles.
