Aiming at the characteristics of nonlinearity

infinite dimension

and spatiotemporal coupling of the temperature field in horseshoe flame glass kiln

a spatiotemporal modeling method based on kernel and linear local tangent space alignment(KLLTSA)algorithm is proposed.Firstly

the temperature data of the kiln is dimensionally reduced using the KLLTSA algorithm nonlinear method to construct spatial basis functions(BFs)that characterize the nonlinear spatial information.Then

the low-order time coefficients are obtained by spatiotemporal separation

which are fitted to construct the temporal model using the back propagation neural network(BPNN).Finally

spatiotemporal reconstruction is used to obtain the global spatiotemporal distribution of the kiln temperature.The proposed method is validated using real data

and the results show that the relative error of the model for temperature prediction is kept within 1%.Compared with the traditional prediction model

the root mean square error(RMSE)

temporal normalized absolute error(TNAE)

and spatial normalized absolute error(SNAE)of the model are reduced by 18%

50.9% and 33.2%

respectively

indicating higher prediction precision and better robustness.