The wavelet transform has partial features

and boasting stronger advantage in signal de-noising.However

the threshold value choise in de-noising process will directly affect the result of signal processing

and both of soft and hard threshold value de-noising methods have their deficiencies.For this

an adaptive learning algorithm of gradient has been introduced for seeking the optimal threshold value

and the said method being applied in signal processing of thermal Engineering process

providing signal more nearing to the actual one for fault diagnosis and system identification.