Introduction to the Dynamic Data Driven Applications Systems (DDDAS) Paradigm

Unsupervised learning of history-dependent constitutive material laws with thermodynamically-consistent neural networks in the modified Constitutive Relation Error framework

“…Before presenting the results of neural network training with the mCRE framework, this section aims to show parameter identification of the hardening modulus in a simple 1D toy problem. This section is associated with an open-access code 1 and can be helpful for a better understanding of the mCRE concept, hyperparameters influence and tuning. Figure 5 illustrates the problem that is addressed.…”

“…To effectively combine the advantages of numerical simulations and physical measurements, there is a growing trend to develop novel processes that enable real-time, dynamic information exchange between the physical system and its corresponding numerical simulator (referred to as a virtual twin). This approach, known as Dynamic Data Driven Application Systems (DDDAS) [1,2], aims to achieve the following objectives:…”

“…t e x i t s h a 1 _ b a s e 6 4 = " U d M b i T 0 + 2 u P 9 j F 6 K 0 c / T K A K1 6 D M = " >A A A J W 3 i c r V b L b t N Q E J 0 + g C Q U S E G s 2 F h E S O k m i n k v K y E k l k E l a a W m i n z d 6 8 S q X / U L K i s f w t e w h U 9 g w R 7 + g p m x 0 z T E i W 8 k b C W Z e 2 b O 8 T y u H Y v A s a O 4 2 / 2 5 t b 2 z e + v 2 n V q 9 c X f v 3 v 0 H z f 2 H g 8 h P Q l P 2 T d / x w x N h R N K x P d m P 7 d i R J 0 E o D V c 4 8 l h c v C P / c S r D y P a 9 T / F V I M 9 c Y + z Z l m 0 a M U K j 5 o u h E J E 9 d o 1 R J i + n 2 t A V / p d M M y L N 0 K z E M y l I 8 y 0 N P U L I I B p l u j 4 d N V v d T p c P b d n Q C 6 M F x d H z 9 3 d r M I R z 8 M G E B F y Q 4 E G M t g M G R H i e g g 5 d C B A 7 g w y x E C 2 b / R K m 0 E B u g l E S I w x E L / B 7 j K v T A v V w T Z o R s 0 2 8 i o O f E J k a P E O O j 3 E h 2 n Q 1 j f 0 J K x O 6 S j t j T c r t C n 9 F o e U i G s M E 0 S r e L F K V R 7 X E Y M F b r s H G m g J G q D q z U E m 4 K 5 S 5 d q O q G B U C x M g + R 3 + I t s n M W Z 8 1 5 k R c O / X W Y P 9 v j i S U 1 m Y R m 8 C f t V k K P M f M l I g 6 n O 2 s O v J R7 1 d P L M M I t 8 J P 2 p e Y y / q o m Q 7 F e I h 9 5 r m 6 n I u H f c g Q j 5 k Z 8 R R o P e E 8 y R L Y 6 f e I T t f y 6 S q 5 Q s 6 / W S e t c 5 / N u 8 3 b Q G 0 5 n 2 r l a m 2 K H z N a r T + P V d H O e 7 W o Y + F a p Y M W H J V y j 5 S 4 v V J u T 4 k b l H I D J W 5 S y k 2…”

NN‐mCRE: A modified constitutive relation error framework for unsupervised learning of nonlinear state laws with physics‐augmented neural networks

“…To exploit the best between numerics and measurements, the trend is to design new processes in which physical systems and their numerical simulator (twin) dynamically and seamlessly exchange information in real‐time through a feedback loop. This paradigm, called dynamic data driven application systems (DDDAS) 1,2 aims to continuously predict the evolution of the physical phenomena of interest and drive the system accordingly (inference phase); to dynamically update the computational model using in situ measurements which are assimilated on‐the‐fly (training phase). …”

A DDDAS-Based Impact Area Simulation Study of Highway Abnormalities

“…Traffic congestion will be reduced, and the highway will be smooth if we can precisely assess the breadth and growth trends of abnormal occurrences and devise targeted traffic management techniques. Therefore, for expressway control and efficient operation, it is crucial to comprehend the extent of the impact of abnormal occurrences [2].…”