Focusing on the defects in fiber structure of corn stalk feed, we proposed a novel production method of ruminant animal fodder by co-treatment with ultrasound and alkalization. By using central composite design with four factors at five levels and response surface analysis, we investigated the effects of ultrasonic power, ultrasonic time, liquid-to-solid ratio and ultrasonic power density in the container on extraction of cellulose and removal of hemicellulose and lignin in corn stalk. In addition, the two-phase regression model was established by taking the contents of lignin, cellulose, and hemicellulose as response values, respectively. The model explained the factor contribution rate of ultrasonic factors to response value and the significant influence on response surface model by the interaction between ultrasonic power, ultrasonic time, liquid-to-solid and ultrasonic power density in the container. The results showed that the determination coefficients of lignin, cellulose and hemicellulose models were 0.69, 0.79 and 0.73, respectively. The process parameters were optimized by the response surface model, and the optimal working parameters of each ultrasonic factor were obtained as ultrasonic power 99 W, ultrasonic time 20 min, liquid-to-solid ratio 7.8:1, and ultrasonic power density in the container 2.05 W/mL. The relative errors of actual values and predicted values were less than 13%, 12% and 14%, respectively. It is indicated that reasonable matching of ultrasonic parameters is helpful to reduce the content of lignin and cellulose of corn stalk feed and increase livestock digestibility, and provides scientific basis for the design of ultrasonic alkalization device and selection of working parameters.