Planck constant from CODATA data
CODATA periodically adjusts the values of fundamental constants using results of laboratories worldwide. The results have different precision and do not always agree within their stated uncertainties. Since 2019 the Planck constant is exactly 6.626 070 15 × 10−34 J·s — the kilogram is defined through it, so there is something to compare with.
What the paper showed
| Method | Estimate of h, 10−34 J·s | Rel. uncertainty, 10−9 |
|---|---|---|
| Birge method (weighted mean corrected for inconsistency) | 6.626 070 150 (69) | 10 |
| Self-refining IF&PA (IF&PA-A) | 6.626 070 136 (06) | 1.3 |
On synthetic sets (100 problems each with normal and uniform distributions) the deviations and uncertainties of IF&PA-A were smaller than those of the Birge method. The demo above uses the light version IF&PA-L, so its uncertainty is larger. Open the data in the calculator →
Muravyov S.V., Khudonogova L.I., Ho M.D. Adjustment of fundamental physical constant values using the interval fusion with preference aggregation. Measurement, 2020, 163, 108037.
Interlaboratory comparisons
A comparison has to establish a reference value and check whether each laboratory agrees with it. For this one computes En = (xk − X) / √(Uk2 + UX2), where U = 2u are expanded uncertainties: |En| ≤ 1 means a satisfactory result. If one laboratory has shifted the reference value X, honest laboratories come under suspicion.
The preference aggregation method was applied to comparisons of electrical quantities, including the search for the largest consistent subset of results.
Muravyov S.V. Ordinal measurement, preference aggregation and interlaboratory comparisons. Measurement, 2013, 46(8), 2927–2935. · Muravyov S.V., Marinushkina I.A. Processing data from interlaboratory comparisons by the method of preference aggregation. Measurement Techniques, 2016, 58(12), 1285–1291.
Wireless sensor networks
Every transmission drains the node's battery, and the farther the node is from the base station, the costlier it is. The goal is to keep the result accurate while transmitting less. Robust fusion means you need not fear that a faulty sensor is among the few readings sent.
Muravyov S.V., Tao Sh., Chan M.Ch., Tarakanov E.V. Consensus rankings in prioritized converge-cast scheme for wireless sensor network. Ad Hoc Networks, 2015, 24(A), 160–171. · Khudonogova L.I., Muravyov S.V. Interval data fusion with preference aggregation for balancing measurement accuracy and energy consumption in WSN. Wireless Personal Communications, 2021, 118, 2399–2421.
Weld defect recognition
Automatic defect recognition needs brightness thresholds that separate a defect from the background. Illumination and background change along the weld, so a single threshold for the whole image often fails. IF&PA fuses the intervals obtained from the strips into thresholds for region growing (RG) and for the double threshold of an edge detector (ED).
Muravyov S.V., Nguyen D.C. Automatic segmentation by the method of interval fusion with preference aggregation when recognizing weld defects. Russian Journal of Nondestructive Testing, 2023, 59(12), 1280–1290. · Muravyov S.V., Nguyen D.C. Method of interval fusion with preference aggregation in brightness thresholds selection for automatic weld surface defects recognition. Measurement, 2024, 236, 114969.
Robust regression and solar panel degradation
Least squares is sensitive to outliers: one distant point turns the line (the “leverage effect”), and with observations of different precision the estimate loses efficiency. IF&PA estimates the slope from the intervals each point gives — and a distant point remains just one vote.
On real performance ratio data of solar power plants the method was applied to a robust estimate of the panel degradation rate.
Muravyov S., Khudonogova L., Pak A. Robust determination of performance loss rate for photovoltaic systems. IEEE Sensors Letters, 2024, 8(9), 7004504. · Muravyov S.V., Khudonogova L.I., Ho M.D. Analysis of heteroscedastic measurement data by the self-refining method of interval fusion with preference aggregation — IF&PA. Measurement, 2021, 183, 109851.
Multi-criteria assessment
Properties of different physical nature cannot be added, but objects can be ordered by each of them. Aggregating the rankings gives the final assessment. Here the “opinions” are ready-made rankings rather than intervals: this is the original problem of the theory IF&PA grew out of. Edit the rankings of the microclimate zones and watch the result change.
The full demo with the pairwise comparison graph and Kemeny distances is on the Theory page.
Labour standards and scheduling in construction
Estimating seasonal labour standards from production records contaminated with errors and carrying them over to the scheduling of linear construction. The paper is in preparation. The experiment can already be tried — on a teaching project or on your own data:
Current directions
- Anomaly detection in data by interval fusion.
- Improving the accuracy and robustness of linear regression analysis.
- Building the calibration characteristic of a primary measuring transducer.
- Data mining of degradation monitoring of solar energy components.
- Network scheduling of production work based on preference aggregation.