Abstract: For there are some errors in the observation matrix and background value matrix of the non-equidistant multi-point gray deformation prediction model, an optimization non-equidistant multi-point gray deformation prediction model based on multivariate total least-squares is proposed. By combining the example analysis, the results show that compared to non-equidistant multi-point gray deformation prediction model based on least-squares parameter estimation that only considers the observed value exists error, this model can suppress the error of modeling data, and improve the fitting and prediction accuracy of the model. It is suitable for application in deformation prediction.

Key words: non-equidistant multi-point gray model, multivariatetotal least-squares, least-squares, deformation prediction