Construction of a Knowledge Base for Simulation and Control of Large Scale and Complex Systems. Applications

Fine-Grained Parallel and Distributed Spatial Stochastic Simulation of Biological Reactions

Catalytic materials and chemistry development using a synergistic combination of machine learning and ab initio methods

Local structure-preserving algorithms including multi-symplectic, local energy- and momentum-preserving schemes are proposed for the generalized Rosenau–RLW–KdV equation based on the multi-symplectic Hamiltonian formula of the equation. Each of the present algorithms holds a discrete conservation law in any time–space region. For the original problem subjected to appropriate boundary conditions, these algorithms will be globally conservative. Discrete fast Fourier transform makes a significant improvement to the computational efficiency of schemes. Numerical results show that the proposed algorithms have satisfactory performance in providing an accurate solution and preserving the discrete invariants.

The specification and distributed control of discrete event robotic manufacturing systems using Petri nets are considered, and a methodology of decomposition and coordination is presented for hierarchical and distributed control. First, a task specification is defined as a Petri net model at the conceptual level, and then transformed to the detailed Petri net representation of the subtasks. Finally, the overall Petri net is decomposed and the constituent subnets are assigned to local Petri net based controllers. The controllers are coordinated by the coordinator so that the decomposed transitions fire at the same time. System coordination algorithm through communication between the coordinator and the controllers, is presented. By the proposed method, modeling, simulation and control of large and complex manufacturing systems can be performed consistently using Petri nets.

The methods of modeling and control of discrete event robotic manufacturing cells using Petri nets are considered, and a methodology of decomposition and coordination is presented for hierarchical and distributed control. Based on task specification, a conceptual Petri net model is transformed into the detailed Petri net model, and then decomposed into constituent local Petri net based controller tasks. The local controllers are coordinated by the coordinator through communication between the coordinator and the controllers. Simulation and implementation of the control system for a robotic workcell are described. By the proposed method, modeling, simulation, and control of large and complex manufacturing systems can be performed consistently using Petri nets.

Solving electronic structure problems represents a promising field of applications for quantum computers. Currently, much effort is spent in devising and optimizing quantum algorithms for near-term quantum processors, with the aim of outperforming classical counterparts on selected problem instances using limited quantum resources. These methods are still expected to feature a runtime preventing quantum simulations of large scale and bulk systems. In this work, we propose a strategy to extend the scope of quantum computational methods to large scale simulations using a machine learning potential trained on quantum simulation data. The challenge of applying machine learning potentials in today’s quantum setting arises from the several sources of noise affecting the quantum computations of electronic energies and forces. We investigate the trainability of a machine learning potential selecting various sources of noise: statistical, optimization, and hardware noise. Finally, we construct the first machine learning potential from data computed on actual IBM Quantum processors for a hydrogen molecule. This already would allow us to perform arbitrarily long and stable molecular dynamics simulations, outperforming all current quantum approaches to molecular dynamics and structure optimization.

Data mining with a multitude of methodologies is a good basis for the integration of intelligent systems. Small, specialised systems have a large number of feasible solutions, but developing truly adaptive, and still understandable, systems for highly complex systems require domain expertise and more compact approaches at the basic level. This paper focuses on the integration of methodologies in the smart adaptive applications. Statistical methods and arti cial neural networks form a good basis for the data-driven analysis of interactions and fuzzy logic introduces solutions for knowledge-based understanding the system behaviour and the meaning of variable levels. Efficient normalisation, scaling and decomposition approaches are the key methodologies in developing large-scale applications. Linguistic equation (LE) approach originating from fuzzy logic is an efficient technique for these problems. The nonlinear scaling methodology based on advanced statistical analysis is the corner stone in representing the variable meanings in a compact way to introduce intelligent indices for control and diagnostics. The new constraint handling together with generalised norms and moments facilitates recursive parameter estimation approaches for the adaptive scaling. Well-known linear methodologies are used for the steady state, dynamic and case-based modelling in connection with the cascade and interactive structures in building complex large scale applications. To achieve insight and robustness the parameters are de ned separately for the scaling and the interactions. Introduction EK Juuso Intelligent Methods in Modelling and Simulation of Complex Systems 2 SNE 24(1) – 4/2014 ON 1 Steady-State Modelling multiple input, multiple output MIMO response surface methodology RSM multiple input, single output MISO Fuzzy set theory Extension principle Type-2 fuzzy Figure 1: Methodologies for modelling of complex system. EK Juuso Intelligent Method of Modelling and Simulation in Complex Systems SNE 24(1) – 4/2014 3 O N Linguistic fuzzy models Takagi-Sugeno (TS) fuzzy models Singleton models Fuzzy relational models Figure 2: Combined fuzzy modelling. Artificial neural networks ANN Linear networks multilayer perceptron MLP backpropagation learning Neurofuzzy systems Figure 3: A fuzzy neuron. function expansion Approximate reasoning 2 Decomposition Methodologies EK Juuso Intelligent Methods in Modelling and Simulation of Complex Systems 4 SNE 24(1) – 4/2014 ON Decomposition Figure 4: Decomposition for modelling. Hierarchical clustering Partitioning-based clustering algorithms fuzzy clustering Fuzzy c-means (FCM) Subtractive clustering Neural clustering Robust clustering number of clusters Composite local model Linear parameter varying (LPV) models Piecewise affine (PWA) systems Fuzzy models Multiple neural network systems mixed approach EK Juuso Intelligent Method of Modelling and Simulation in Complex Systems SNE 24(1) – 4/2014 5 O N 3 Adaptive Nonlinear Scaling Membership 3.1 Working point and feasible ranges Figure 5: Nonlinear scaling [28] EK Juuso Intelligent Methods in Modelling and Simulation of Complex Systems 6 SNE 24(1) – 4/2014 ON 3.2 Membership definitions linguistic range linguistic values • the corner points (Figure 5) are good for visualisation; • the parameters suit for tuning; • the coefficients are used in the calculations. 3.3 Adaptation of nonlinear scaling EK Juuso Intelligent Method of Modelling and Simulation in Complex Systems SNE 24(1) – 4/2014 7 O N 4 Intelligent Systems 4.1 LE models 4.2 Hybrid LE systems EK Juuso Intelligent Methods in Modelling and Simulation of Complex Systems 8 SNE 24(1) – 4/2014 ON extension principle linguistification

Current simulation tools for multibody dynamics are not problem dependent, they use the same modeling process to all cases regardless of their characteristics. In addition, real-time simulation of small multibody systems is achievable by existing simulation tools, however, real-time simulation of large and complex systems is not possible with existing methods. This is a challenge that needs to be addressed before further advances in mechanical simulation with hardware-in-the-loop and man-in-the-loop, as well as virtual prototyping are made possible. This paper addresses the issue of how the modeling process-dependent versus independent coordinates, and descriptor form versus state-space form of the equations of motion-affects the dynamic simulation of multibody systems and how it may be taken into account and added to the concept of intelligent simulation. With this new concept all the factors involved in the simulation process — modeling, equations, solution, etc. — are chosen and combined depending upon the characteristics of the system to be simulated. It is envisioned that this concept will lead to faster and more robust real-time simulators.

The context of our work is the complex systems simulation. Complex systems modelling and simulation require to take into account the interactions between the different components. For instance, an airport ground traffic simulation requires to model the planes with the pilots, the control tower, the

<div class="section abstract"><div class="htmlview paragraph">The objective of this paper is two-fold. Firstly, provide guidance to best implement end to end traceability from program requirements to physical implementation, and Secondly provide techniques to review and understand large scale complex systems. Even with a Digital Engineering Environment (DEE) being an enabler towards applying Systems Engineering practices to develop large scale complex systems, many organizations are unclear on the methodology for modeling their architectures and enabling stakeholders to easily review, understand and assess those architectures. An architecture can be a conceptual, logical or physical architecture, depending on the system’s lifecycle state. For the context of this paper, the modeling environment is any System’s Modeling Language (SysML) based tool along with modeling tools for electrical, mechanical and software development and product life cycle management tool. The intended audience is any engineering organization defining end-to-end architecture within a DEE, and all stakeholders tasked with reviewing conceptual, logical and physical architecture. The outcome of this paper is to provide engineering organizations with guidance on underlying principles for modeling and understanding or assessing architecture that describe large complex systems.</div></div>

Complex adaptive systems (CAS) exhibit properties beyond complex systems such as self-organization, adaptability and modularity. Designing models of CAS is typically a non-trivial task as many components are made up of sub-components and rely on a large number of complex interactions. Studying features of these models also requires specific work for each system. Moreover, running these models as simulations with a large number of entities requires a large amount of processing power. We propose a framework that consists of a modeling language, analysis tools, and simulation engine, called CASTLE. We propose a language, Complex Adaptive Systems Language (CASL), which is designed for simple creation of CAS models while remaining domain agnostic. In particular, an extension to CASL, called CASL-SG, that introduces the concept of 'semantic grouping' allows for large scale simulations to execute on relatively modest hardware. A component of our framework, the observation module, aims to provide an extensible and expandable set of metrics to study key features of CAS such as aggregation, adaptability, and modularity, compatibility with existing analysis tools, while also allowing for more domain-specific techniques.

Multi-scale methods such as wavelet and multigrid schemes are efficient algorithms for the numerical solution of large classes of partial differential equations (PDEs). In this chapter we illustrate the basic ideas behind iterative methods for the solution of large linear systems which arise from the discretisation of PDEs. Besides describing the concepts of wavelet preconditioning, we summarise the main ingredients of multigrid methods. Examples from computational geosciences will show the importance of multi-scale methods for the simulation of large and complex systems.

Overview: New directions for simulation and modeling of complex systems

This study examines the performance of integration methods for hybrid simulation of large and complex structural systems in the context of structural collapse due to seismic excitations. The target application is not necessarily for real-time testing, but rather for models that involve large-scale physical sub-structures and highly nonlinear numerical models. Four case studies are presented and discussed. In the first case study, the accuracy of integration schemes including two widely used methods, namely, modified version of the implicit Newmark with fixed-number of iteration (iterative) and the operator-splitting (non-iterative) is examined through pure numerical simulations. The second case study presents the results of 10 hybrid simulations repeated with the two aforementioned integration methods considering various time steps and fixed-number of iterations for the iterative integration method. The physical sub-structure in these tests consists of a single-degree-of-freedom (SDOF) cantilever column with replaceable steel coupons that provides repeatable highlynonlinear behavior including fracture-type strength and stiffness degradations. In case study three, the implicit Newmark with fixed-number of iterations is applied for hybrid simulations of a 1:2 scale steel moment frame that includes a relatively complex nonlinear numerical substructure. Lastly, a more complex numerical substructure is considered by constructing a nonlinear computational model of a moment frame coupled to a hybrid model of a 1:2 scale steel gravity frame. The last two case studies are conducted on the same porotype structure and the selection of time steps and fixed number of iterations are closely examined in pre-test simulations. The generated unbalance forces is used as an index to track the equilibrium error and predict the accuracy and stability of the simulations.

Machine-learning potentials (MLPs) trained on data from quantum-mechanics based first-principles methods can approach the accuracy of the reference method at a fraction of the computational cost. To facilitate efficient MLP-based molecular dynamics and Monte Carlo simulations, an integration of the MLPs with sampling software is needed. Here, we develop two interfaces that link the atomic energy network (ænet) MLP package with the popular sampling packages TINKER and LAMMPS. The three packages, ænet, TINKER, and LAMMPS, are free and open-source software that enable, in combination, accurate simulations of large and complex systems with low computational cost that scales linearly with the number of atoms. Scaling tests show that the parallel efficiency of the ænet-TINKER interface is nearly optimal but is limited to shared-memory systems. The ænet-LAMMPS interface achieves excellent parallel efficiency on highly parallel distributed-memory systems and benefits from the highly optimized neighbor list implemented in LAMMPS. We demonstrate the utility of the two MLP interfaces for two relevant example applications: the investigation of diffusion phenomena in liquid water and the equilibration of nanostructured amorphous battery materials.