Research On Wire Laser Cladding Process And Its Neural Network Expert System

Wire laser cladding is a process for additive manufacturing and remanufacturing.It is different from other processes with high material utilization rate,high processing efficiency,low energy consumption and small heat affected zone.Although laser cladding with wire has experienced decades,in it there are still a large number of problems that have not been solved,which limits its application on a large scale.Among them,there are two major problems.One is that It is difficult to obtain ideal cladding quality,the other is that the process specification is difficult to determine so that the period of determination of process specification is long.In view of the problem that the cladding quality is not up to standard and the flatness of cladding layer is not enough,in this paper the process is studied experimentally with the movable multifunctional laser cladding system produced by Hunan Jiacheng Additive Manufacturing Co.,Ltd.The effects of laser power,scanning speed,wire feeding speed and relative position of wire on the height,width and surface flatness of clad bead.A number of multi-pass laser cladding experiments were performed,and the effects of process variables on the clad layer were disclosed.In this paper,two methods of variable-lap rate and hot wire laser cladding are proposed to improve the inner quality and surface flatness of the clad layer.In a series of experiments,it is found that in variable-lap rate laser cladding or hot wire laser cladding the flatness of cladding layer can be indeed improved,but only in hot wire feeding laser cladding defects such as pores and inclusions in clad layer can be eliminated.It is found that the clad bead microstructure in hot wire feeding laser cladding includes mostly equiaxed crystals,while that in cold wire laser cladding includes mostly columnar crystals.In view of the problem of long R & D cycle and trial and error of cladding process,a neural network expert system for wire laser cladding was developed with SQL Sever Management Studio 18 and visual studio 2019.It is distinguishing that the forward and reverse reasoning-mixed mode was adopted in this expert system,in which the traditional expert experience is used as the forward reasoning basis and the neural network is used as the reverse reasoning basis.The reliability of the neural network expert system for wire laser cladding is improved by the combination of forward and reverse reasoning.Finally,the experiment oriented to engineering practice is carried out to verify the reliability of the neural network expert system for wire laser cladding.Using 405 stainless steel wire of 1.2mm,a high-quality clad layer with a height of 1.5mm and a flatness of 84.5% is processed on the 45# steel hollow pipe with an inner diameter of280 mm and an outer diameter of 312 mm.It is found that the cladding layer has no defects such as cracks and pores,and the maximum error of the height of the cladding layer is 4.1%.The results show that the neural network expert system is effective and reliable.