vikas saraf

© ATI 2012. All Rights Reserved.

Vikas Saraf

Artificial Neural Networks to Predict Tensile Propertiesin Titanium Forgings

An ICME Approach: Integrated Computational Materials Engineering

October 7-10, 2012 • Atlanta, Georgia

© ATI 2012. All Rights Reserved. 2© ATI 2012. All Rights Reserved.

Predicting Tensile Properties in Titanium Forgings

• ICME in action

• Ti6-4 processing

• Models developed

• Integration and results

• Summary

ICME: Integrated Computational Materials Engineering

Overview

© ATI 2012. All Rights Reserved.© ATI 2012. All Rights Reserved.

Optimal Solution

Billet Requirement

Forging/HT/Machining requirement

Engine designrequirement

Machined

In-engine

OEM shape design

ICME in Supply chain

EA GP7000

© ATI 2012. All Rights Reserved.© ATI 2012. All Rights Reserved.Titanium in Boeing 777: 10% of the structural weight (Source: ITA).

Advanced Material Usage- Titanium in Boeing 777

Courtesy: The Boeing Company

© ATI 2012. All Rights Reserved. 5© ATI 2012. All Rights Reserved.

ß-transus

α-ß

Primary processing

Secondary processing

Evolving microstructure

Ti6‐4 Processing

© ATI 2012. All Rights Reserved.© ATI 2012. All Rights Reserved.

Ti6-4 HT1 Temp= ß+

Alpha Prime phase basketweave structure within prior beta grains.

This structure results from fast cooling from Heat Treating in the Beta phase field.

Alpha-Beta structure refined through conversion and forging process.

Heavy Cross-Section thicknesses Cooling rate is too slow…

Reduced Cross-Section thicknesses Cooling rates better for high strength.

Ti6‐4 Processing

© ATI 2012. All Rights Reserved. 7© ATI 2012. All Rights Reserved.

• Phase Equilibrium/ Beta approach curve

• Phase Field/ Alpha and beta volume fraction

• Crystal Plasticity– Yield surface– Strain partitioning– Texture

• Primary alpha growth/ volume fraction

• Prior beta grain size

• Secondary alpha growth (lath width)

• Variant selection (αs transformation)

• Texture formulation (Kearns number)

• RT Tensile properties

Predictability

© ATI 2012. All Rights Reserved.© ATI 2012. All Rights Reserved.

Neural Network model, Ti6-4

Fv- volume Fraction

Predictability

Artificial Neural Network Training using PatternMaster® Program

© ATI 2012. All Rights Reserved.© ATI 2012. All Rights Reserved.

Validation

UTS; ±5%

Neural Network model, Ti6-4

Predictability

© ATI 2012. All Rights Reserved.© ATI 2012. All Rights Reserved.

Production resultsTi 6-4 pancake forged in production environmentα-ß forged; α-ß heat treated; aged

Measurement locations

Process Simulation• Billet conversion• Secondary near-net forging• Heat Treatment

Data Generation• Beta approach curve• Flow stress generation• Strain partitioning• Initial Texture

Prediction• α, ß growth• α Fv• α-lath• ND α-texture

Post-Processing• RT Tensile properties

© ATI 2012. All Rights Reserved.© ATI 2012. All Rights Reserved.

Volume fraction of primary alpha Grain size (radius) of primary alpha

Thickness of secondary alpha lath

Production results

LocationBCMCRCBTMT

0.070.070.000.220.00

0.070.020.210.010.09

0.262.500.370.980.69

αp-size (µm) αp-Fv αs-width (µm)Measured-Predicted Measured-Predicted Measured-Predicted

© ATI 2012. All Rights Reserved.© ATI 2012. All Rights Reserved.

Production resultsUTS

1 3

2 4 5

© ATI 2012. All Rights Reserved.© ATI 2012. All Rights Reserved.

Production resultsYield Strength

1 3

2 4 5

© ATI 2012. All Rights Reserved.© ATI 2012. All Rights Reserved.

Production resultsElongationReduction in Area

With texture

Without texture

1 3

2 4 5

© ATI 2012. All Rights Reserved.© ATI 2012. All Rights Reserved.

Neural Network model, Ti6-4

(-) Nominal (+) (-) Nominal (+)

Upper bound

Lower bound

© ATI 2012. All Rights Reserved.© ATI 2012. All Rights Reserved.

Neural Network model, Ti6-4

0 2x 0 2x

Upper bound

Lower bound

© ATI 2012. All Rights Reserved.© ATI 2012. All Rights Reserved. Measured Avg / Range

Radial Axial Hoop

148.5/3

150/0

156/0

147.5/1

146.5/1

157/0

135.5/1

136/0142/0

134/2

133.5/1

144/0

UTS

YS

α-ß forged; α-ß heat treated; agedTensile strength prediction; Shaft forging Shen, Saraf, Furrer; TMS 2006

© ATI 2012. All Rights Reserved.© ATI 2012. All Rights Reserved.

Radial Hoop

144ksi143ksi

143ksi

147ksi147ksi149ksi

139ksi

Tensile strength prediction; Complex forging

α-ß forged; Mill annealed

Shen, Saraf, Furrer; TMS 2006

© ATI 2012. All Rights Reserved.© ATI 2012. All Rights Reserved.

Infrastructure to predict complex microstructural evolution and, subsequently, its effect on tensile properties, has been developed.

Simple geometrical shapes produced in an industrial environment confirm the predictability and application of the developed infrastructure.

Complex, near-net-shape aerospace components have shown good correlation between prediction and measurements.

Summary

© ATI 2012. All Rights Reserved.© ATI 2012. All Rights Reserved.

Acknowledgement

Development and integration of models was funded by the USAF under the auspices of Materials Affordability Initiatives.

Michael Glavicic, Tom Broderick, Vasisht Venkatesh, Todd Morton,Yoji Kosaka, Ron Wallis, Vikas Saraf

(AIPT Team members from Rolls-Royce, GE, Pratt & Whitney, Boeing, Timet, Wyman Gordon and ATI Ladish respectively)

And

Fan Zhang, Wei-Tsu Wu, Ravi Shankar, Ayman Salem,Yunzhi Wang, Donald Boyce

(Sub-contractors from CompuTherm, Scientific Forming Technologies Corp., Materials Resources, The Ohio State University and Cornell University, respectively)