Blended coal combustion is a primary mean to improve economic efficiency of the thermal units.However

due to the poor grindability of low quality coal and their low calorific value

the coal feed quantity will become huger at the same load

which will easily lead to blockage of the coal mills.A transition process called "critical blocking" state between normal state and blocked state has been found in engineering applications.Accurately recognizing the critical blockage state of the medium speed mills is of great significance for units' safe and stable operation.A target pattern recognition was found by wavelet multiscale decomposition in combination with the D-S evidence theory

the original evidences were selected after mechanism analysis.Then

the evidences' eigenvalues were extracted and aligned in the time domain by wavelet multiscale decomposition.Furthermore

membership functions were constructed by typical samples

and the basic probability assignments of each evidences were gotten.Finally

recognition was realized by the D-S evidence theory.The actual operating data were computed to validate the effectiveness of the method.