M. Vasudevan

Artificial neural network approach for estimating weld bead width and depth of penetration from infrared thermal image of weld pool

Abstract In this article an artificial neural network based system to predict weld bead geometry using features derived from the infrared thermal video of a welding process is proposed. The multilayer perceptron and radial basis function networks are used in the prediction model and an online feature selection technique prioritises the features used in the prediction model. The efficacy of the system is demonstrated with a number of welding experiments and using the leave one out cross-validation experiments.

Genetic-Algorithm-Based Computational Models for Optimizing the Process Parameters of A-TIG Welding to Achieve Target Bead Geometry in Type 304 L(N) and 316 L(N) Stainless Steels

The weld-bead geometry in 304LN and 316LN stainless steels produced by A-TIG welding plays an important role in determining the mechanical properties of the weld and its quality. Its shape parameters such as bead width, depth of penetration, and reinforcement height are decided according to the A-TIG welding process parameters such as current, voltage, torch speed, and arc gap. Identification of a suitable combination of A-TIG process parameters to produce the desired weld-bead geometry required many experiments, and the experimental optimization of the A-TIG process was indeed time consuming and costly. Therefore it becomes necessary to develop a methodology for optimizing the A-TIG process parameters to achieve the target weld-bead geometry. In the present work, genetic algorithm (GA)-based computational models have been developed to determine the optimum/near optimum process parameters to achieve the target weld-bead geometry in 304LN and 316LN stainless steel welds produced by A-TIG welding.

Measurement of residual stresses in austenitic stainless steel weld joints using ultrasonic technique

Abstract A methodology has been developed using a non-destructive ultrasonic technique for measuring surface/subsurface residual stresses in 7 mm thick AISI type 316LN stainless steel weld joints made by activated tungsten inert gas and multipass tungsten inert gas welding processes. Measurement of residual stresses using an ultrasonic technique is based on the effect of stresses on the propagation velocity of elastic waves. Critically refracted longitudinal L CR wave mode was employed and accurate transit time measurements were made across the weld joints. Quantitative values of the longitudinal residual stresses across the weld joints were estimated from the measured transit times and predetermined value of acoustoelastic constant for AISI type 316LN stainless steel. The nature of the residual stress profiles and their variations across the two types of weld joints were compared and interpreted.

Intelligent Modeling for Optimization of A-TIG Welding Process

An intelligent model combining artificial neural network (ANN) and genetic algorithm (GA) has been developed for determining the optimum process parameters for achieving the desired depth of penetration and weld bead width during Activated Flux Tungsten Inert Gas Welding (A-TIG) welding of type 316LN and 304LN stainless steels. First, ANN models correlating process parameters with depth of penetration and weld bead width have been developed. There was good correlation between the measured and models predicted depth of penetration as well as the weld bead width for both training and test data. A GA code was developed in MATLAB in which the objective function was evaluated using the ANN models. The optimized values for GA parameters such as crossover rate, population size, and mutation probability were identified. The developed GA model produced multiple outputs such as current, torch speed, voltage, and arc gap for the same target depth of penetration and bead width, and validation was carried out by experiments. There was good agreement between the target values and the actual values of depth of penetration and weld bead width obtained for both the stainless steels.

Validation of processing maps for 304L stainless steel using hot forging, rolling and extrusion

The development of a microstructure in 304L stainless steel during industrial hot-forming operations, including press forging (mean strain rate of 0.15 s(-1)), rolling/extrusion (2-5 s(-1)), and hammer forging (100 s(-1)) at different temperatures in the range 600-1200 degrees C, was studied with a view to validating the predictions of the processing map. The results have shown that excellent correlation exists between the regimes exhibited by the map and the product microstructures. 304L stainless steel exhibits instability bands when hammer forged at temperatures below 1100 degrees C, rolled/extruded below 1000 degrees C, or press forged below 800 degrees C. All of these conditions must be avoided in mechanical processing of the material. On the other hand, ideally, the material may be rolled, extruded, or press forged at 1200 degrees C to obtain a defect-free microstructure.

Ultrasonic velocity measurements for characterizing the annealing behaviour of cold worked austenitic stainless steel

Precise measurements of shear wave velocities have been made in 20% cold worked and annealed samples of alloy D9 (Ti-modified austenitic stainless steel) to characterize the microstructural changes during annealing. The variation in wave velocity with annealing time exhibited a three stage behaviour. In the first stage, a slight increase in the velocity during recovery regime has been observed. This is followed by a significant increase in the velocity in the recrystallization regime (second stage) and reaching saturation on completion of recrystallization (third stage). Microstructural observations using optical metallography confirmed these three stages. The maximum variation in velocity is observed only when the polarization or the propagation direction of the shear wave is alingned with the cold working direction. Variation in shear wave velocity during annealing process, in general, is just opposite to that of the variation in longitudinal wave velocity. A number of velocity measurements were made by changing the propagation and polarization directions of the shear waves with respect to the cold working direction. Based on these measurements, a suitable velocity ratio parameter is suggested for determining the degree of recrystallization during annealing of cold worked alloy D9.

Effect of Activated Flux on the Microstructure and Mechanical Properties of 9Cr-1Mo Steel Weld Joint

Tungsten inert gas (TIG) welding process is generally used to produce high quality weld joints of 9Cr-1Mo steel. However, there is limitation associated with the depth of penetration achievable in single pass autogenous welding. Specific activated flux has been developed in the present work to enhance the depth of penetration up to 6 mm in single pass by A-TIG welding. 9Cr-1Mo steel A-TIG weld joint using activated flux was made in single pass welding while the multipass TIG weld joint using modified 9Cr-1Mo filler wire was made in seven passes. The enhancement in depth of penetration during A-TIG welding process for this steel was attributed to arc constriction. The strength properties of the A-TIG weld joint was superior to that of the multipass TIG weld joint. The multipass TIG weld joint exhibited slightly improved impact toughness than the A-TIG weld joint in PWHT condition. Therefore, there was no degradation in the microstructure and mechanical properties of the weld joint produced by A-TIG welding process compared to that of the weld joint produced by conventional TIG welding process in plain 9Cr-1Mo steel.

Assessment of Residual Stresses and Distortion in Stainless Steel Weld Joints

Welding introduces significant residual stresses in the welded structure/component due to non-uniform heat distribution during heating and cooling cycle. To control, reduce, or beneficially redistribute the residual stresses in weld joints, the stress distribution needs to be known. In the present study, weld joints of 10 mm thick 316LN stainless steel were made by multi-pass TIG and A-TIG welding processes and their residual stresses distribution and distortion values were measured and compared. While V-groove edge preparation was required for making multi-pass TIG weld joint, square-edge preparation was sufficient for making single pass A-TIG weld joint. Ultrasonic nondestructive technique based on the critically refracted longitudinal waves (LCR waves) has been used for the quantitative surface/sub-surface residual stress measurements in the weld joints. Distortion measurements were carried out before and after welding using height gauge. Peak tensile residual stress and the angular distortion values were lower in the A-TIG weld joint compared to that of the multipass TIG weld joint.

Delta ferrite prediction in stainless steel welds using neural network analysis and comparison with other prediction methods

The ability to predict the delta ferrite content in stainless steel welds is important for many reasons. Depending on the service requirement, manufacturers and consumers often specify delta ferrite content as an alloy specification to ensure that weld contains a desired minimum or maximum ferrite level. Recent research activities have been focused on studying the effect of various alloying elements on the delta ferrite content and controlling delta ferrite content by modifying the weld metal compositions. Over the years, a number of methods including constitution diagrams, Function Fit model, Feed-forward Back-propagation neural network model have been put forward for predicting the delta ferrite content in stainless steel welds. Among all the methods, neural network method was reported to be more accurate compared to other methods. A potential risk associated with neural network analysis is over-fitting of the training data. To avoid over-fitting, Mackay has developed a Bayesian framework to control the complexity of the neural network. Main advantages of this method are that it provides meaningful error-bars for the model predictions and also it is possible to identify automatically the input variables which are important in the non-linear regression. In the present work, Bayesian neural network (BNN) model for prediction of delta ferrite content in stainless steel weld has been developed. The effect of varying concentration of the elements on the delta ferrite content has been quantified for Type 309 austenitic stainless steel and the duplex stainless steel alloy 2205. The BNN model is found to be more accurate compared to that of the other methods for predicting delta ferrite content in stainless steel welds.

Bayesian neural network analysis of ferrite number in stainless steel welds

Abstract Bayesian neural network (BNN) analysis has been used in the present work to develop an accurate model for predicting the ferrite content in stainless steel welds. The analysis reveals the influence of compositional variations on ferrite content for the stainless steel weld metals, and examines the significance of individual elements, in terms of their influence on ferrite content in stainless steel welds, based on the optimised neural network model. This neural network model for ferrite prediction in stainless steel welds has been developed using the database used to generate the WRC-1992 diagram and the first authors laboratory data. The optimised committee model predicts the ferrite number (FN) in stainless steel welds with greater accuracy than the constitution diagrams and the other FN prediction methods. Using this BNN model, the influence of variations of the individual elements on the FN in austenitic stainless steel welds is also determined, and it is found that the change in FN is a non-linear function of the variation in the concentration of the elements. Elements such as Cr, Ni, N, Mo, Si, Ti, and V are found to influence the FN more significantly than the other elements present in stainless steel welds. Manganese is found to have a weaker influence on the FN. A noteworthy observation is that Ti influences the FN more significantly than does Nb, whereas the WRC-1992 diagram considers only the Nb content in calculating the Cr equivalent.