基于BP神经网络模型的Ni-TiN镀层耐磨性预测研究

Predicative Study on the Wear Resistances of Ni-TiN Coatings Using BP Neural Networks

采用超声电沉积方法,在C470型压缩机阀片表面制备Ni-TiN镀层。利用扫描电镜、X射线衍射仪和摩擦磨损试验机研究Ni-TiN镀层表面形貌、组织结构及耐磨性,并采用BP神经网络模型预测Ni-TiN镀层的磨损量。结果表明,BP神经网络模型的最佳结构组成为3×9×1,其预测值与实验值的拟合度R=0.99938,相对误差最大值与最小值分别为1.67%和0.63%。当TiN粒子浓度为8 g/L、超声波功率180 W、电流密度4 A/dm2时,Ni-TiN镀层表面犁沟较浅,磨损量较小。Ni-TiN镀层中存在Ni和TiN相,镍的衍射峰分别位于44.82°、52.22°和76.78°,TiN的衍射峰分别位于38.48°、42.82°和66.54°。

Ni-TiN coatings were prepared on the surface of C470 type compressor valve plate by ultrasonic electrodeposition method. The surface morphology,structure and wear resistance of Ni-TiN coating were studied by the scanning electron microscope（ SEM）,X-ray diffraction instrument（ XRD） and abrasion tester. The wear loss mass of Ni-TiN coatings were forecasted by BP neural network model. The results indicate that the structure of the BP model is 3 × 9 × 1,and the fitting similarity is 0. 99938. The maximum and minimum relative deviation is 1. 67% and 0. 63%,respectively. SEM and XRD results show that NiTiN coatings have a shallow furrows surface and less wear loss when the TiN particles at a concentration of8 g / L,ultrasonic power of 180 W,the current density of 4 A / dm^2. There were nickel and TiN phases in the coatings,and nickel diffraction peaks were located at 44. 82°,52. 22° and 76. 78°,respectively. TiN diffraction peaks were located 38. 48°,42. 82° and 66. 54°,respectively.