关键词: 无取向硅钢, 铁损, 多元线性回归, LightGBM算法, 新能源汽车, 驱动电机

Abstract: Combined with the systematic pilot test research and electron backscatter diffraction (EBSD) detection means, the influence of "composition, microstructure and texture" of 2.7 %~3.6 % Si -0.5 %~2.0 %  Al -0.3 %~0.9 % Mn non-oriented silicon steel on the iron loss was investigated, and the main influencing factors were clarified. SPSS software was used to carry out the primary and secondary statistical analysis on the correlation of the factors affecting iron loss. On this basis, the multiple linear regression method and LightGBM algorithm were used to establish the prediction model of non-oriented silicon steel loss, and the fitting ability of the model was evaluated by the mean square error and average error of the prediction results. The results showed that: the textures of finished sheets are dominated by {111}〈112〉 and {114}〈481〉.The increase of Si, Al and Mn content increases the resistivity of nonoriented silicon steel, and the iron loss P_1.5/50 decreases accordingly.With the increase of grain size of the finished sheets, the iron loss P_1.5/50 decreases significantly.There are twenty-three main factors affecting the loss of nonoriented silicon steel by "compositionmicrostructuretexture".The mean square error of iron loss prediction model established by multiple linear regression method and LightGBM algorithm can reach 8.76×10^-3 and 3.81×10^-5, respectively, and the average error rate can reach 3.231 % and 0.210 %, respectively. The models established by the two methods can be used to predict the iron loss of non-oriented silicon. However, the prediction accuracy of the LightGBM algorithm is higher than that of the multiple linear regression model, and the interpretation of the prediction results of the multiple linear regression model is higher than that of the LightGBM algorithm.

Key words: non-oriented silicon steel, iron loss, multiple linear regression, lightGBM algorithm；new energy vehicle, drive motor