Inference Model of 304 Stainless Steel Sheet Welding Quality Based on Improved FNN-BP Network
For the current development direction of the intelligent setting of laser welding parameters in the field of laser welding,the presumption module of welding parameters for intelligent welding systems has become a hot research object.After analyzing the influence of welding process parameters on welding quali-ty,a laser welding quality inference model based on an improved fuzzy expert system and BP neural net-work is built,which consists of two parts,namely a fuzzy inference of welding quality based on welding speed,welding power and out-of-focus amount and a BP modified neural network based on predicted val-ues,plate thickness,peak power,and duty cycle.The fuzzy inference of welding quality is firstly based on the existing manual experience to fuzzify the welding parameters and establish the welding rule base,then the fuzzy inference type is determined through analysis,and finally the fuzzy inference output welding qual-ity prediction value;BP neural network correction,based on the plate thickness and other parameters for dif-ferent plate thickness under the weld image quality score and flatness difference to correct the prediction value,to obtain more accurate inference value.Through the experiments,the stainless steel sheet intelligent laser welding system proved to have certain feasibility and important engineering significance.
NETL
NSTL
万方数据
