Uncertainty Analysis For Pump Test Based On Monte Carlo Method And Maximum Entropy

Since promulgation of "Guide to the Expression of Uncertainty in Measurement", the theory of uncertainty have been used widely on the various branches. In recent years, many people committed to find the the best evaluation methods, and obtained a lot of tremendous achievements, but there are still few researchers in area of the uncertainty on the branche of pump test, what is more, the overall level of China's pump industry relatively backward, it is imperative to seek an effective mean of evaluation.In this dissertation,, we put forward a new method that Uncertainty in Measurement which based on prevailing problems exist on the pump test uncertainty theory, we assess uncertainty of the single-stage centrifugal pump performance test parameters, basement of Monte Carlo and Maximum Entropy method, the method combines the advantages of the two theories to carry out uncertainty evaluation.The research mainly engaged in the following areas:Analysis feasibility of uncertainty evaluation, and the advantages and disadvantages of each method, compare the applicability;Analysis of pump performance test system to find the sources of uncertainty, and the methodsof eliminating, Simulated data collection about Single-stage water pump, we carry out uncertainty evaluation by the way of Bessel formula on the dates;We distributed simulation on data using the maximum entropy, to achieve the most objective data distribution types, obtained with the distribution of a certain number of random numbers according to the type of distribution and corresponding rejection method, then to evaluat using the Monte Carlo method Evaluation of the uncertainty, compare the results with the traditional way to assessment results, and draw conclusions;We learn about algorithms aiming at problems of initial value is difficult to choose during the maximum entropy solution, we given a kind of particle swarm algorithm based on the scheme, no longer need artificial given initial value, only need a given range, by itself, the space program can search optimal solution, assessment procedures are simplified, avoid some random factors especially selective initial value problems caused the improper convergence, and test data generation going tobe calculate, and compared conventional algorithm results to verify the efficiency of the algorithm.