Elsevier

Optics and Lasers in Engineering

Volume 86, November 2016, Pages 62-74
Optics and Lasers in Engineering

Welded joints integrity analysis and optimization for fiber laser welding of dissimilar materials

https://doi.org/10.1016/j.optlaseng.2016.05.011Get rights and content

Highlights

  • An optimization method for dissimilar laser welded joints integrity is proposed.

  • The integrity of welded joints in the dissimilar welding is analyzed and evaluated.

  • The relationship between process parameters and weld bead integrity is established.

  • The results show that the method is effective at obtaining desired welded joints.

Abstract

Dissimilar materials welded joints provide many advantages in power, automotive, chemical, and spacecraft industries. The weld bead integrity which is determined by process parameters plays a significant role in the welding quality during the fiber laser welding (FLW) of dissimilar materials. In this paper, an optimization method by taking the integrity of the weld bead and weld area into consideration is proposed for FLW of dissimilar materials, the low carbon steel and stainless steel. The relationships between the weld bead integrity and process parameters are developed by the genetic algorithm optimized back propagation neural network (GA-BPNN). The particle swarm optimization (PSO) algorithm is taken for optimizing the predicted outputs from GA-BPNN for the objective. Through the optimization process, the desired weld bead with good integrity and minimum weld area are obtained and the corresponding microstructure and microhardness are excellent. The mechanical properties of the optimized joints are greatly improved compared with that of the un-optimized welded joints. Moreover, the effects of significant factors are analyzed based on the statistical approach and the laser power (LP) is identified as the most significant factor on the weld bead integrity and weld area. The results indicate that the proposed method is effective for improving the reliability and stability of welded joints in the practical production.

Introduction

The availability of welded joints between dissimilar materials is continuously growing in the different applications such as the power, automotive, chemical and spacecraft industries. The greater flexibility, cost incurred towards costly and scarce materials are the attractive properties which dissimilar joints can provide. Despite of the advantages, the welding of dissimilar materials is not the easy work as the similar materials welding. The new issues, carbon migration, differences in thermal expansion coefficients, difficulty in heat treatment and electrochemical property variations, are usually existed in the dissimilar materials welding [1]. Compared with the conventional welding methods [2], the laser welding has been more widely used in industrial application. These process characteristics like high power density, high penetration, high productivity and narrow affected zone are particularly beneficial for the welding of dissimilar materials [3], [4]. The laser welding as a valid way for the dissimilar materials welding is demonstrated by many researchers. Because of dissimilar materials differences in the physical, mechanical and metallurgical properties [5], [6], the welding defects are often associated with welded joints. The lack of fusion, underfill, undercut and other related weld imperfections decrease the weld bead integrity directly. During the fiber laser welding (FLW) of dissimilar materials, the welding process parameters, such as the laser power (LP), welding speed (WS), focal position (FP), gap (GAP) and shielding gas (SG), greatly affect the integrity of weld bead. Therefore, adopting the proper process parameters is essential for improving the joints integrity and welding quality of dissimilar materials.

To ensure the high quality welding, the weld bead shape affected by the process conditions is studied by some researchers. Since the complicated relationships between welding process parameters and bead geometry, the welding parameters are often determined by the welder's experience, charts and handbooks in the practical production [7]. While they are not efficient for the continuously updated welding process due to the required time consuming inspection and trial. To obtain the desired weld bead and improve the welding quality, design of experiments (DOE) and statistical techniques are widely used in the welding through developing the relationship models between input variables and output responses. Tarng et al. [8] obtained the set of optimal welding process parameters using Taguchi method. Kumar et al. [9] employed Taguchi method and regression model to optimize the magnetic arc oscillation welding process parameters of AA 5456 aluminum alloy welds to improve the mechanical properties. The influences of welding current, welding speed, amplitude and frequency on mechanical properties were studied. Shojaeefard et al. [10] optimized the rotational speed, tool tilt angle and traverse speed in friction stir welding adopting Taguchi L9. The tensile strength of welded joints was set as the optimal objective. As the developing mathematical models are time consuming and cumbersome in expressing the non-linear characteristics between the inputs and outputs, the artificial intelligence method is as an effective tool and introduced into the welding field. Katherasan et al. [11] proposed flux cored arc welding parameters optimization by artificial neural networks (ANN) and particle swarm optimization (PSO) algorithm. During the optimization process, the neural network illustrated good predicting capacity and the expected penetration, width and reinforcement of the weld bead were achieved. Dutta et al. [12] modeled TIG welding process using conventional regression analysis and neural network-based approaches. The results indicated that the genetic-neural approach could yield more adaptive predictions compared to that of the conventional regression analysis method. Yang et al. [13] reported the response surface methodology and back propagation neural network (BPNN) for the process parameters optimization in the lap welding. It showed that the average error from BPNN is less than that from regression models derived from the response surface methodology. Sathiya et al. [14] used ANN and genetic algorithm (GA) to optimize weld bead geometry and mechanical property in the laser welding. The optimal results were in good agreement with the actual values from the experiments. It is observed from the above studies that the artificial intelligence method is capable of identifying the optimal process with reasonable high accuracy in the welding of similar materials. However, the weld bead integrity is not considered in the optimization process.

Only a few investigations of welded joints integrity optimization in the dissimilar laser welding have been reported in the literatures. Ruggiero et al. [15] studied the weld-bead profile optimization of low carbon steel and austenitic steel in the laser welding. However, the weld integrity is not as the optimal objective and optimized in their research. Torkamany et al. [16] investigated the optimization of laser welding parameters for a sound weld with full penetration along the dissimilar interface. They could only achieve minimizing both the middle width and area of the weld bead. Furthermore, Sun et al. [17] reported the butt welding of Al/steel by using a 10 kW fiber laser welding system. This study only indicated that the good weld appearance was obtained at the appropriate welding parameters from several specimens analysis. Their work was focused on the feasibility of butt joining of aluminum and steel using laser filler wire.

This paper proposes an optimization method for improving weld bead integrity in the FLW of dissimilar materials using the artificial intelligence and statistical technique. The integrity of weld bead and the process parameters effects are analyzed. The relationship between the weld bead integrity and process parameters is established by the genetic algorithm optimized back propagation neural network (GA-BPNN). The predicted results from GA-BPNN are optimized by PSO with the objective of maximizing the weld bead integrity and minimizing the weld area. The optimal results are verified on weld profiles, microstructure characteristics, microhardness and tensile strength properties. Due to the limitation of judging the effects of the process parameters on the responses in the optimizing process, the significant factors effects on the integrity and area of weld bead are analyzed based on the signal to noise (S/N) ratio and analysis of variance (ANOVA).

The structure of the paper is organized as follows: The welded joints integrity is analyzed and experimental details to study the weld bead integrity are described in Section 2. Section 3 introduces establishing the relationships between the process parameters and output responses based on GA-BPNN to predict the weld bead integrity and the PSO algorithm used for process parameters optimization. In Section 4, the verification of the proposed optimization method of weld bead integrity on the weld macro profile, microstructure, microhardness and tensile strength properties, and impacts of the process parameters are discussed. The conclusions of the current study are offered in Section 5.

Section snippets

The integrity analysis of welded joints

The integrity of the welded joints is highly related to the welding quality and the safe operation in the industrial production [18]. In particular, the dissimilar metals with different thermal conductivity, thermal expansion coefficient and other physical properties are negative for the formation of good joint integrity. Elmesalamy et al. [19] reported that the welded joints with the integrity decreased to 80% failed in the weld bead during the tensile tests. They thought that the weld bead

Optimization methodology

The engineering problems are usually solved by traditional optimization methods. However, they are not efficient or inclined to obtain a local optimal solution for dealing with the practical problems with large search space. Although BPNN is suitable for nonlinear problems predicting, it is very sensitive to the initial values and often oscillation and hard to be convergence during the training process. To improve the performance of the BPNN, GA which has the good optimization capacity is

Validation of the results

To confirm validity of the optimal results, three groups of test experiments were carried out. Since the accuracy grade of the optimized process parameters is much more than that of the experimental condition in practical, the values closing to the optimal results are selected in the feasible solution as presented in Table 8. The weld bead from the optimized process parameters is shown in Fig. 7 and the geometry parameters, WF, WB and BH, are obtained by using the same experimental procedure as

Conclusion

The demand for FLW of dissimilar materials is growingly increasing in the automotive, power, chemical and space industries. The process parameters have significant influence on the weld bead integrity and hence the welding quality. This paper proposes an optimization method considering the integrity of the weld bead and weld area comprehensively in the FLW of the low carbon steel (Q235) and stainless steel (SUS301L-HT). The purpose of the present research is to improve the weld bead integrity

Acknowledgments

This research has been supported by the National Natural Science Foundation of China (NSFC) under Grant no. 51323009, the National Basic Research Program (973 Program) of China under Grant no. 2014CB046703, the National Natural Science Foundation of China (NSFC) under Grant nos. 51505163 and 51421062, and the Fundamental Research Funds for the Central Universities, HUST: Grant no. 2014TS040. The authors also would like to thank the anonymous referees for their valuable comments.

References (41)

Cited by (53)

View all citing articles on Scopus
View full text