Research On Hot Deformation Behavior And Evolution Of Grain Boundary Characteristic Distribution In Incoloy 925 Alloy

Incoloy 925 alloy（hereinafter referred to as 925 alloy）is an age-strengthened nickel-based corrosion-resistant alloy developed based on 825 alloy.The age-hardening treatment results in the formation of a diffuse distribution of nanoscaleγ′phases within the austenitic matrix,which makes the alloy have good creep resistance,corrosion resistance and tissue stability,as well as excellent comprehensive mechanical properties.It is widely used in harsh high temperature and corrosive media environment,such as chemical,petroleum,nuclear power and other energy industries.However,its high degree of alloying not only increases the complexity of mechanical behavior and tissue regulation during thermal processing,but also weakens the corrosion resistance of the alloy at the grain boundaries during service.With the booming development of nuclear power industry and petroleum industry in China,the demand for 925 alloy will further increase.Therefore,it is of great theoretical significance and practical value to study the evolution of microstructure of 925 alloy during thermomechanical treatment,optimize the thermal processing parameters of the alloy and improve the service performance of the alloy.In this paper,based on the constant strain rate hot compression test,the introduction of artificial neural network in the thermal deformation of 925 alloy achieves the improvement of the constitutive model,processing map and recrystallization kinetic model,and explores the feasibility of annealing treatment to regulate the grain boundary characteristic distribution after hot deformation with large strain rate.Secondly,the evolution of grain boundary networks at different initial deformation temperatures was studied.Finally,the effects of strain,processing pass and strain path on the evolution behavior and periodicity of the grain boundary network in the iterative thermomechanical treatment were investigated.The main contents and conclusions are as follows:（1）The hot deformation behavior of 925 alloy was systematically studied at deformation temperatures of 900¹150°C and strain rates of 0.01¹0s^-1.Under low strain conditions,for the samples with lower temperature and higher strain rate,the flow stresses show negative strain rate sensitivity.Dynamic recrystallization（DRX）mechanisms include continuous dynamic recrystallization,discontinuous dynamic recrystallization,and particle stimulated nucleation.Discontinuous dynamic recrystallization is the main mechanism of dynamic softening,which can effectively promote the nucleation of DRX grains at the original grain boundaries to form a necklace structure.In addition,the distorted twin grain boundaries also provide nucleation sites for DRX.（2）Three constitutive models were constructed based on the flow stress profile,namely,Arrhenius model,BP neural network model and GA-BP neural network model.The predictability of the three constitutive models was quantified by calculating the statistical index parameters such as correlation coefficient（R）and average absolute relative error（M）.Neural network models exhibit excellent predictability and genetic algorithm effectively improves the stability and generalization of neural network models.The 3D processing map of 925 alloy was further constructed based on reconstructing the3D stress distribution combined with dynamic material theory.The improved 3D processing map exhibits better predictive capability and more accurate response to rheological instability behavior such as dynamic strain aging and adiabatic temperature rise.An EBSD recrystallization grain identification method with higher accuracy than GOS threshold method is proposed based on neural network.The Avrami-DRX recrystallization kinetic model was improved by using the Bi Hill function to describe the changes of adiabatic temperature rise and different nucleation mode under different strain rates,and the data traversal and parameter optimization were carried out by combining with genetic algorithm.Compared with the conventional kinetic equations,the correlation coefficient of the optimized model was improved from 84%to 95%.（3）The feasibility of annealing treatment to optimize the grain boundary characteristic distribution after hot deformation with large strain was investigated.After the annealing treatment at 1075°C/10min,theΣ3ⁿ grain boundary ratio basically recover to the solution treatment state.However,the overall distribution of J₂/（1-J₃）values is relatively low due to the higher nucleation density.As the annealing time was increased to 90 min,the storage energy was further released,which promoted the growth of twin clusters in the fully dynamically recrystallized samples.As the increasing deformation temperature and decreasing the strain rate decrease,theΣ3 grain boundary percentage increases and theΣ3/Σ（9+27）ratio decreases,which promotes the transition from the"new twin formation mechanism"to the"Σ3 regeneration mechanism".To control the nucleation density,the degree and distribution of storage energy after hot deformation with large strain can be regulated by DRX,and the lower strain rate and higher deformation temperature are beneficial to the formation of large-size twin clusters.（4）The correlation between energy storage structure and deformation heat treatment parameters at different deformation temperatures was studied systematically.Cryogenic deformation significantly increases the dislocation density of the deformed matrix and reduces the average free range of dislocations that become entangled.The geometrically necessary dislocations in the larger deformed samples were distributed in a herringbone,ribbon or lamellar pattern.After the subsequent annealing treatment,when the pre-strain is 10%,the optimization of the grain boundary characteristic distribution is significantly improved,with the best effect in the sample annealed at 1075°C,where the percentage ofΣ3ⁿ grain boundaries can reach 68.39%.The cryogenic deformation can effectively reduce the recrystallization critical temperature,and the percentage ofΣ3ⁿ grain boundary in the sample with 10%pre-strain and annealing temperature of 1050°C increases from31.47%to 67.46%.The long-range ordered dislocation cumulative distribution structure in the matrix after cryogenic deformation introduced more strain concentration regions and increased the migration rate of grain boundaries.Under the suitable pre-strain conditions,cryogenic deformation is more favorable to achieve the optimization of grain boundary characteristic distribution.（5）The iterative thermomechanical treatment（ITMP）significantly increased the percentage ofΣ3ⁿ grain boundaries and promoted the generation of incoherent twin boundaries.The periodicity of the grain boundary characteristic distribution after ITMP with 5%strain is related to the dynamic balance of strain accumulation and strain release.The percentage ofΣ3ⁿ grain boundaries fluctuates with the increase of processing passes,and reaches 68.19%by 8 passes of processing.The ITMP with 10%strain promotes the nucleation and growth of strain-free grains,so the percentage ofΣ3ⁿ grain boundaries is basically maintained above 55.79%.The peakΣ3ⁿ grain boundary percentage（69.34%）was reached after 6 iterations of processing.Cross rolling improves the deformation uniformity,activates more slip systems,and promotes the accumulation of strain in grains with different orientations.The original dislocation structure is effectively disrupted by introducing mixed strain.The cross-rolling improves the dislocation blocking state,which can effectively promote the grain boundary reaction and the formation of high-orderΣ3n twin grain boundaries,shortening the periodicity of iterative processing.