Effect of process parameters on process efficiency and inertia friction welding behavior of the superalloys LSHR and Mar-M247

Abstract

The relationships between the inertia friction welding process parameters (initial angular velocity, ωₒ, and moment of inertia, I, of the flywheel and the axial compression force, P) and welding behavior (welding time, sample upset and flash formation, efficiency and kinetics of friction-induced sample heating) of dissimilar superalloys were determined. The results showed that the initial kinetic energy of the flywheel, E_o = Iωₒ²/2, should not be considered as a key parameter for the design of the IFW process. Only a fraction of this energy (denoted as ‘sample energy’, E_S), which is process-parameters dependent, is used to heat the workpieces at the weld interface. Together with I and P, E_S controls the upset length, the duration of the welding process, temperature profiles near the weld interface and the weld quality. Optimization of the inertia friction welding for better quality, reduced energy consumption and materials saving requires information about the process efficiency η = E_S/E_o and its dependence on the process parameters. Critical sample energy per unit surface area (≈ 79 MJ/m²) was identified above which good quality metallurgical bonds were produced between LSHR and Mar-M247 alloys.

Introduction

During the inertia friction welding (IFW) process, the kinetic energy of the welding machine flywheel is converted into frictional heat at the joining surfaces resulting in local material softening and axial/radial metal flow under the action of an applied axial compression force. The local weld upset moves plasticized material, together with surface contaminants and oxides, into the flash and brings nascent material from each component into contact, facilitating the formation of a metallurgical bond. Detailed information about this process is provided in a review paper by Attallah and Preuss (2012).

The efficiency η of the inertia friction welding (IFW) process is defined as the fraction of the total flywheel kinetic energy which is dissipated at the faying surfaces of the welded components, η = E_S/E_o. Mahaffey et al. (2016) found that, at the same initial (total) flywheel kinetic energy E_o, an increase in the flywheel moment of inertia I decreased the fraction of the kinetic energy lost to parasitic sinks within the IFW machine. Senkov et al. (2017) conducted direct measurements of the friction torque and dissipated energy at the joining surfaces during IFW using a MTI–120 B welder and reported that η depends greatly on process parameters such as the moment of inertia I, the initial (total) kinetic energy Eₒ (or initial angular velocity ωₒ) of the flywheel, and the axial compression force P. In particular, they found that η can vary from ∼20% to 72% with the remaining energy consumed by the rotating parts of the welding equipment/machine bearings. The parasitic energy losses increase (and thus η decreases) with an increase in Eₒ (or ωₒ) (for IFW trials conducted at fixed values of I and P) or P (at fixed I and ωₒ) and with a decrease in I (at fixed P and E_o). Senkov et al. (2017) explained the strong effect of IFW parameters on η by the strong and different dependence of the friction torque of the machine bearings (machine torque, M_M) and the torque between the faying surfaces of the welded components (sample torque, M_S) on the angular velocity ω of the flywheel. In particular, Senkov et al. (2017) reported that M_M has a maximum at ω = ωₒ and decreases linearly to a minimum, static value $M_M^0$ with a decrease in ω to zero. M_S, on the other hand, exhibits low values at the beginning of welding but rapidly increases as ω decreases to zero. As a result of such different behavior of M_M and M_S, higher fraction of the flywheel energy is converted to heat at the joining surfaces (and thus higher efficiency is achieved) when welding with lower ω.

These recent findings make the results of previous IFW models, which are based on the assumption that the process efficiency is constant and it does not depend on IFW parameters, questionable. For example, Wang et al. (2005) assumed in their calculations that the efficiency of IFW is 85%, without any explanation of how this value was derived. Mohammed et al. (2009) compared upset and temperature curves of the energy based model at various flywheel mechanical efficiencies with those measured experimentally (at a single set of IFW parameters) and found good agreement when η was between 90 and 95%. Wang et al. (2014) conducted comprehensive finite element modeling of the IFW process at different values of E_o, I and P while assuming that η = 90% and independent on the process parameters. Bennett (2015) compared the experimental temperature profiles during the heating stage of a single IFW process with those calculated from finite element modeling and estimated the process efficiency value of ∼70% and then used this value across all of the models presented in his paper. The agreement of the output of such models (e.g., temperature, upset length, flash development, etc.) with experimental data was generally achieved by adjusting the friction coefficient and/or the input flow stress data. Although these simulations may indirectly account for the effect of process efficiency on welding behavior for a limited, specific range of parameters, assumptions related to efficiency likely increase uncertainty when the modeling approach is applied to other welding conditions or to the design of production-scale operations. Therefore, the conclusions drawn from the results of these models about the specific effects of I, Eₒ, ωₒ and P on welding behavior such as the welding time, the decrease in workpiece length (upset length), flash formation, and the width of the heat affected zone, require additional experimental verification. Unfortunately, none of the previously-reported experiments provides sufficient information for such analysis.

In the present work, IFW of two dissimilar superalloys, LSHR and Mar-M247, was conducted using a wide range of welding conditions, and the effect of ωₒ, I, P, and η on the welding time, upset length, sample friction torque, and the kinetics of the friction-induced sample heating were quantified. The experimental results were compared with recent finite-element-modeling predictions for the IFW process.