Shiping Zhang

Improved LMS Adaptive Algorithm and its Application of Time Delay Estimation in Power Plant

Time delay estimation plays an important role in accuracy of leakage location in industrial boiler furnace. Considering about interference, like strong background noise, reverberation and so on, in boiler furnace, the LMS (least mean square) adaptive filter algorithm should be improved to meet certain environment above .In order to makes the LMS algorithm have the characteristic of fast convergence and small steady state error, an improved, power-normalized and variable step-size discrete cosine transform LMS algorithm is proposed, which combines the power-normalized discrete cosine transform LMS algorithm with the variable step size LMS algorithm that uses sliding forgetting-weighted window . The time delay estimation simulation in the strong-noise environment verifies the improved DCT-MVSS LMS algorithm can achieve good performance.

Time Delay Estimation in Reverberant Environment Based on Acoustic Pyrometry

The acoustic thermometry has many advantages, compared with conventional methods of temperature measurements. For this technology, the sound field in normal temperature state of the boiler was simulated; acoustic source signal obtained the pseudo random sequence signal and time delay estimation selected the weighted cross-correlation method. Experiments show that when the boiler is not running, the sound field is enclosure sound field in the furnace. The weighted cross-correlation method can restrain the reverberation and obtain the accurate time delay estimation.

Application of Improved LMS Adaptive Algorithm in Acoustics Pyrometry

Acoustic pyrometry is a comparatively advanced method of temperature measurement developed in recent years, which possesses the essential characteristics of traditional temperature measurement approach. Considering the interferences, like strong background noise, reverberation and so on, in boiler furnace, the LMS (least mean square) adaptive filter algorithm should be improved to meet certain environment above. In order to make the LMS algorithm have the characteristic of fast convergence and small steady state error, an improved, power-normalized and variable step-size discrete cosine transform LMS algorithm is proposed, which combines the power-normalized discrete cosine transform LMS algorithm with the variable step size LMS algorithm that uses the sliding forgetting-weighted window. The time delay estimation simulation in the strong-noise environment verifies the improved DCT-MVSS LMS algorithm can achieve good performance.