digital twin of ceramic processing - lucideon
TRANSCRIPT
Digital Twin of Ceramic Processing
Jingzhe Pan
School of Engineering
University of Leicester
Challenges for the ceramics industry
▪ High rejection rate
e.g. 60% of the amount of revenue generated [1]
▪ Excessively high cost of post-sintering machining
e.g. 60–90% of total cost of finished product [2]
1) J Bhamua, KS Sangwan, Reduction of Post-kiln Rejections for improving Sustainability in Ceramic Industry: A Case Study,
Procedia CIRP, 26 (2015), 618 – 623.
2) AN Samant and NB Dahotre, Laser machining of structural ceramics – A review, Journal of the European Ceramic Society
29 (2009) 969–993.
Computer modelling of ceramic processing
▪ Spray drying and die filling
▪ Injection moulding/powder compaction
▪ Drying
▪ Sintering
H. Riedel 1997Fraunhofer-Institut für
Werkstoffmechanik, Wöhlerstr.
Freiburg, Germany
Finite element modelling of
sintering deformation
Jingzhe Pan, International Materials Reviews 2003 Vol. 48 No. 2
Work by Pan’s team
co-firing of bi-layered beam
Work by Pan’s team
nonuniform initial density
Work by Pan’s team vs experiment
Work by Pan’s team
cracking during constrained sintering
Challenges for computer modelling of sintering
➢ no chemistry/material input
Challenges for computer modelling of sintering
➢ extremely sensitive constitutive law
Thermodynamics dictates that
ሶ𝜀𝑖𝑗 =𝜕Ψ
𝜕𝜎𝑖𝑗
- strain rate potentialΨ
Challenges for computer modelling of sintering
➢ extremely sensitive constitutive law
Challenges for computer modelling of sintering
➢ extremely sensitive constitutive law
H. Riedel 1997Fraunhofer-Institut für
Werkstoffmechanik, Wöhlerstr.
Freiburg, Germany
Finite element modelling of
sintering deformation
Requires individual measurement
of the constitutive properties
Digital Twin
A digital twin of sintering (Leicester ongoing work)
𝜎𝑖𝑗
ሶ𝜀𝑖𝑗L
ρ
Training an artificial neural network to learn a constitutive law
𝜀ሶ𝑖𝑗 =ϵ0ሶ
σ0 𝐿0𝐿 3
3
2𝑐 ρ 𝑠𝑖𝑗 + 3𝑓 ρ σm − σs δ𝑖𝑗
𝜎𝑖𝑗
ሶ𝜀𝑖𝑗 = ?L =1.6 µm, ρ = 70%, 𝜎𝑠 = 1.0MPa
Work by Venkat Ghantasala, PhD student; Shuihua Wang, PDRA
0.896 0.735 0.606 0.658 0.802 0.912
Training an artificial neural network to learn the constitutive law
𝜎𝑖𝑗
ሶ𝜀𝑖𝑗L
ρ
Training an artificial neural network to take chemistry
and material inputs
𝑐𝑖
Training an artificial neural network to learn from
manufacturing data
Concluding remark
• A digital twin turns the manufacturing process of advanced
ceramics into a material laboratory, such that issues are
identified and resolved, and the process is optimised.
• Simulation-based control under real-time constraint is
possible.
• The digital twin can communicate with production through
the Internet of Things (5G), opening the door for separation
of skills in manufacturing and mathematical modelling.