The study of high efficiency on-line thermal unknown input estimation algorithm

• Main Authors: HOU WEN-TIEN, 侯文田
• Other Authors: TUAN PAN-CHIO
• Format: Others
• Language: zh-TW
• Published: 1998
• Online Access: http://ndltd.ncl.edu.tw/handle/95925932876711726091

博士 === 中正理工學院 === 國防科學研究所 === 86 === An on-line thermal unknown''s estimation method using Input Estimation algorithm has been successfully employed in 1-D and 2-D Inverse Heat Conduction Problems (IHCP). However, implementing this method involves handling various system parameters such as measurement noise, modeling error, external disturbance and changes due to time variation. In this thesis, two kinds of adaptive method were proposed to solve this problem: First, we use Root Mean Square Error (RMSE) as performance index, versus forgetting factor under input numerical experiment to identify not only the performance surface, but also the optimal or robust forgetting factor. To directly synthesize Kalman filter with input estimator, this work also presents an efficient robust forgetting zone, capable of providing a reasonable tracking time-lag and noise filtered estimation results. Next, an adaptive weighting input estimation algorithm which efficiently and robustly on-line estimate time-varied thermal unknowns is proposed. The adaptivity is accomplished by the Kalman filter which provides a regression equation between the bias innovation and thermal unknown. Then the adaptive weighting recursive least squares estimator is proposed based on this regression model to extract the unknowns which defined as the input. The maximum likelihood type estimator (M-estimator) combining Huber psi-function is used to construct the adaptive weighting forgetting factor as the function of biased innovation at each time step, so that the unknown is estimated under the system involved measurement noise, modeling error, and unpredictable time-varying changes of the unknowns. This proposed algorithm simply upgrades the conventional input estimation approach, and facilitates practical implementation. Finally, the superior results are verified through one and two dimensional inverse heat conduction problems in several time varying estimate cases and two bench mark performance test. Thus, the on-line real time IHCP hardware loop implementation is easily realized by this adaptive and robust input estimation technique.