The error of the elongation is 2.3 %, by the multi-dimension linear regression model, the relative error are from ? 5 % to ? 15 % respectively . comparing the results, it can be found that the neural network overmatches the multi-dimension linear regression . so that, the model can be used in the production of the marine propeller aluminum bronze 計(jì)算結(jié)果表明,三項(xiàng)輸出的預(yù)測值與實(shí)測數(shù)據(jù)接近,抗拉強(qiáng)度和屈服強(qiáng)度的相對誤差在1的范圍以內(nèi),延伸率的最大絕對誤差為2.3,明顯優(yōu)于多元線性回歸模型5?15的誤差,能較好地滿足工程應(yīng)用的要求,對高強(qiáng)度船用螺旋槳鋁青銅的生產(chǎn)具有一定的指導(dǎo)意義。