Taking the mill vibration in a million kilowatts coal-fired generating unit as the soft measurement object

the mill vibration model combining phase space reconstruction with extreme learning machine( ELM) was established. Moreover

the predicted results for mill vibration with and without information input were compared. By phase space reconstruction

this model

which concerns the multiple parameters and variable coal types comprehensively

reconstructs the vibration intensity into m-dimensional matrix with τ delay. The last dimension will be applied as the output of the ELM

while the others will be adopted as the input with other parameters. The test results show that

the average relative error between the measured value and predicted value is within 6. 27%

which is superior to the conventional ELM's and SVM's. The coal type information has a certain effect on the mill vibration.