Comparative Evaluation of the Accuracy and Robustness of Various Information Processing Algorithms from Tripled Measuring Channels

Using the example of a three-channel system for measuring the level of filling of tanks of launch vehicles, we perform a comparative assessment of the accuracy and robustness of various algorithms for processing measurement information received from three measuring channels of the same type. We consider various processing algorithms of three equal-precision measurements: finding the arithmetic mean of three equal-precision measurements and a majority choice of three equal-precision measurements. For these algorithms, the probability density function of the measurement errors and confidence limits of the measurement error are found using the assumption of uniform or normal laws of the error distribution of three equally accurate measurements. A generalized algorithm for processing redundant information is proposed. The essence of the generalized algorithm is the addition of three equally accurate measurements taking into account individual weighting factors. Weight coefficients are selected based on the values of three equally accurate measurements: the minimum and maximum values are assigned the same weighting coefficient, and the median is assigned a coefficient which normalizes the sum of the three weighting factors to one. For a generalized algorithm by means of simulation modeling, the confidence limits of the measurement error are determined depending on the selected weighting factors. The robustness of the considered processing algorithms is evaluated for three equally accurate measurements performed on three measuring channels identical in design, with a random perturbation applied to one of the channels. Recommendations on the selection of weights for three equal measurements are given. For the fuel level measurement system under consideration, it is justified to use an algorithm for processing measurement information based on a majority choice of three equally accurate measurements.