Achieving superior polymeric components through additive manufacturing(AM)relies on precise control of rheology.One rheological property particularly relevant to AM is melt viscosity(η).ηis influenced by polymer che...Achieving superior polymeric components through additive manufacturing(AM)relies on precise control of rheology.One rheological property particularly relevant to AM is melt viscosity(η).ηis influenced by polymer chemistry,molecular weight(M_(w)),polydispersity,shear rate([Math Processing Error]),and temperature(T).The relationship ofηwith M_(w),[Math Processing Error],and T is captured by parameterized equations.Several physical experiments are required to fit the parameters,so predictingηof new polymer materials in unexplored physical domains is laborious.Here,we develop a Physics-Enforced Neural Network(PENN)model that predicts the empirical parameters and encodes the parametrized equations to calculateηas a function of polymer chemistry,M_(w),polydispersity,[Math Processing Error],and T.We benchmark our PENN against physics-unaware Artificial Neural Network(ANN)and Gaussian Process Regression(GPR)models.We demonstrate that the PENN offers superior values ofηwhen extrapolating to unseen values of M_(w),[Math Processing Error],and T for sparsely seen polymers.展开更多
基金supported by the Office of Naval Research(ONR)through Grant N00014-21-1-2258the National Science Foundation(NSF)DMREF Grant 2323695.A.J.acknowledges S.Shukla for help with the fingerprinting of polymer blends.A.J.acknowledges L.Chen and C.Kuenneth for valuable discussions on benchmarking ML models.
文摘Achieving superior polymeric components through additive manufacturing(AM)relies on precise control of rheology.One rheological property particularly relevant to AM is melt viscosity(η).ηis influenced by polymer chemistry,molecular weight(M_(w)),polydispersity,shear rate([Math Processing Error]),and temperature(T).The relationship ofηwith M_(w),[Math Processing Error],and T is captured by parameterized equations.Several physical experiments are required to fit the parameters,so predictingηof new polymer materials in unexplored physical domains is laborious.Here,we develop a Physics-Enforced Neural Network(PENN)model that predicts the empirical parameters and encodes the parametrized equations to calculateηas a function of polymer chemistry,M_(w),polydispersity,[Math Processing Error],and T.We benchmark our PENN against physics-unaware Artificial Neural Network(ANN)and Gaussian Process Regression(GPR)models.We demonstrate that the PENN offers superior values ofηwhen extrapolating to unseen values of M_(w),[Math Processing Error],and T for sparsely seen polymers.
