Methodology For Model Generation Research Articles

Compact Modeling of Advanced Gate-All-Around Nanosheet FETs Using Artificial Neural Network.

As the architecture of logic devices is evolving towards gate-all-around (GAA) structure, research efforts on advanced transistors are increasingly desired. In order to rapidly perform accurate compact modeling for these ultra-scaled transistors with the capability to cover dimensional variations, neural networks are considered. In this paper, a compact model generation methodology based on artificial neural network (ANN) is developed for GAA nanosheet FETs (NSFETs) at advanced technology nodes. The DC and AC characteristics of GAA NSFETs with various physical gate lengths (Lg), nanosheet widths (Wsh) and thicknesses (Tsh), as well as different gate voltages (Vgs) and drain voltages (Vds) are obtained through TCAD simulations. Subsequently, a high-precision ANN model architecture is evaluated. A systematical study on the impacts of ANN size, activation function, learning rate, and epoch (the times of complete pass through the entire training dataset) on the accuracy of ANN models is conducted, and a shallow neural network configuration for generating optimal ANN models is proposed. The results clearly show that the optimized ANN model can reproduce the DC and AC characteristics of NSFETs very accurately with a fitting error (MSE) of 0.01.

Robust empirical aging model generation methodology: Description and validation

The aging behavior of each lithium-ion battery is unique, and its observation requires expensive tests. Consequently, it does not exist a standardized aging model and, in most cases, the available aging data is insufficient to describe completely the aging behavior. This paper proposes a robust aging model generation methodology applicable to any lithium-ion battery with any kind of lab-level aging data availability. This methodology has four phases: firstly, an expert aging study; secondly, the main trend model; thirdly, the integration of the effect of the aging stress factors on the main trend model; fourthly, the verification of the resultant aging model. The robustness of the aging model is provided to the model thanks to the verification process where the accuracy of describing the testing cases, the correctness of describing interpolated operating conditions, and the correctness of describing the aging at extrapolated time periods and cycle numbers are checked. This methodology has been applied to an example aging data set. The result is a customized robust aging model that describes the aging behavior observed on the provided aging data that validates the proposed methodology on a real use case. The proposal exploits the universality of the empirical aging model while minimizing the inherent uncertainty levels, obtaining like this a robust and universal aging modeling methodology.

Virtual Gas Turbines Part II: An Automated Whole-Engine Secondary Air System Model Generation

Abstract The design and analysis of the secondary air system (SAS) of gas turbine engine is a complex and time-consuming process because of the complicated topology and iterative nature of SAS design. The conventional SAS design-analysis model generation process is quite tedious and inefficient. It is still largely dependent on human expertise and thus incurs high time-cost. This paper presents an automated, whole-engine SAS flow network model generation methodology. This method accesses a prebuilt feature-based whole-engine geometry model and transforms the geometry features into the features suitable for SAS flow network analysis. The proposed method extracts both the geometric and non-geometric information from the engine geometry model such as rotational frames, materials, and boundary conditions. Apart from ensuring geometric consistency, this methodology also establishes a bidirectional information exchange protocol between the engine geometry model and the SAS flow network model, which enables to make engine geometry modifications based on SAS analysis results. The application of this feature mapping methodology is demonstrated by generating the SAS flow network model of a modern three-shaft gas turbine engine. This flow network model is generated within a few minutes, without any human intervention, which significantly reduces the SAS design-analysis time cost. The proposed methodology seamlessly links the geometry and the air system modelers of Virtual Gas Turbines simulation framework and thus allows performing a large number of whole-engine SAS simulations, design optimizations and fast redesign activities.

Automation of route identification and optimisation based on data-mining and chemical intuition.

Data-mining of Reaxys and network analysis of the combined literature and in-house reactions set were used to generate multiple possible reaction routes to convert a bio-waste feedstock, limonene, into a pharmaceutical API, paracetamol. The network analysis of data provides a rich knowledge-base for generation of the initial reaction screening and development programme. Based on the literature and the in-house data, an overall flowsheet for the conversion of limonene to paracetamol was proposed. Each individual reaction-separation step in the sequence was simulated as a combination of the continuous flow and batch steps. The linear model generation methodology allowed us to identify the reaction steps requiring further chemical optimisation. The generated model can be used for global optimisation and generation of environmental and other performance indicators, such as cost indicators. However, the identified further challenge is to automate model generation to evolve optimal multi-step chemical routes and optimal process configurations.

GIS based generation of dynamic hydrological and land patch simulation models for rural watershed areas

This paper introduces a GIS based methodology to generate dynamic process model for the simulation based analysis of a sensitive rural watershed. The Direct Computer Mapping (DCM) based solution starts from GIS layers and, via the graph interpretation and graphical edition of the process network, the expert interface is able to integrate the field experts’ knowledge in the computer aided generation of the simulation model. The methodology was applied and tested for the Southern catchment basin of Lake Balaton, Hungary. In the simplified hydrological model the GIS description of nine watercourses, 121 water sections, 57 small lakes and 20 Lake Balaton compartments were mapped through the expert interface to the dynamic databases of the DCM model. The hydrological model involved precipitation, evaporation, transpiration, runoff, infiltration. The COoRdination of INformation on the Environment (CORINE) land cover based simplified “land patch” model considered the effect of meteorological and hydrological scenarios on freshwater resources in the land patches, rivers and lakes. The first results show that the applied model generation methodology helps to build complex models, which, after validation can support the analysis of various land use, with the consideration of environmental aspects.

Development of a geometrically accurate and adaptable finite element head model for impact simulation: the Naval Research Laboratory–Simpleware Head Model

This study demonstrates a novel model generation methodology that addresses several limitations of conventional finite element head models (FEHM). By operating chiefly in image space, new structures can be incorporated or merged, and the mesh either decimated or refined both locally and globally. This methodology is employed in the development of a highly bio-fidelic FEHM from high-resolution scan data. The model is adaptable and presented here in a form optimised for impact and blast simulations. The accuracy and feasibility of the model are successfully demonstrated against a widely used experimental benchmark in impact loading and through the investigation of potential brain injury under blast overpressure loading.

Automatic simulation model generation for simulation-based, real-time shop floor control

This paper presents a structure and architecture for automatic simulation model generation (for very detailed simulation models intended to be used for real-time simulation-based shop floor control). The simulation model code is generated from a shop floor resource model and a shop floor control model. The shop floor resource model provides much of the static information for the simulation model; while a shop level control model provides much of the dynamic information required by the simulation model. The simulation code generated can be used for traditional system analysis, but more importantly, it can also be used to control the manufacturing system by sending and receiving messages using an Ethernet communication link to a high-level task executing system. Six manufacturing systems are used to illustrate and test the validity of the simulation model generation methodology. Finally, factory level planning and scheduling activities using the generated simulation model are described.

Artificial intelligence modelling of control systems

The field of control systems simulation needs artificial intelligence technology as much as expert systems need systems simulation tools. A functional approach for the design of Expert Systems that perform model generations and simulations is proposed. A differential games simulator design is chosen to exemplify the above ideas. The discrete-event approach, based on the geometry of the game is proposed. Results are comparable with the simu lation results obtained using the imperative approach, but the interrogative approach offers faster execution and clearer sim ula tor definition. Knowledge representation of the differential games models is described using Semantic Networks. The model genera tion methodology is a blend of several problem-solving para digms, and the hierarchical dynamic goal system construction serves as the basis for model generation. Prolog-based imple mentation of the system is suggested.