ABSTRACT. Modelling for simulation, modelling for optimization, and modelling for control follow the same principles, but have to meet different requirements. Simulation models should represent the behavior of the system under consideration for a predefined set of test cases faithfully, with the accuracy usually being measured in the time domain, i.e. by looking at the differences of the stationary values or trajectories of some key variables. Modelling for optimization is different in that the goal is that the computed optimum of the model and the optimum of the real system should match, and the model should support efficient global numerical optimization. This implies that the model must be accurate around the global optimum, but over the full range of the variables, a medium accuracy is sufficient. The problem for modelling of course is that without the knowledge of the region of the global optimum, it is not possible to build such a model, and hence modelling and optimization should be interleaved.

When using models for control, the relationship between model accuracy and control performance is even more intricate than for optimization. One can control a plant satisfactorily using a coarse model (in the simplest case, the sign of the gain is sufficient), while model errors can also lead to poor performance and instability. For linear time-invariant control loops, the classical robust control theory from the 1980s tells us that a good model for control is a model that describes the dynamic behavior accurately near the gain-crossover frequency, and qualitatively correctly for lower frequencies. So what a good model is for the purpose of control depends on how fast we want to control the system.

For high-performance control, techniques based on online optimization, i.e. model-predictive control (MPC) have become the dominant technology in the last decades, due to their ability to handle constraints, nonlinear systems, and economically motivated cost functions (Engell, 2007). In contrast to the classical theory of robust control, robustness to model errors for such control strategies is difficult to analyze and, similar to stability, is usually handled using a constructive approach, i.e. by building controllers that have certain robustness properties for a given description of the uncertainty of the model. min-max robust MPC in which the performance is optimized for the worst case model is the best known representative of this approach. This however comes at the price of a high conservatism.

To build good models is a costly endeavor. Therefore, both in modelling for optimization and in modelling for control, one is interested in techniques that provide good performance without huge modelling efforts. In the presentation, we discuss two recent approaches to reducing the negative effects of model errors in optimization and control. For real-time optimization, we outline the so-called modifier adaptation approach, which adds a data-based local model to a global model and updates it iteratively to ensure convergence to the true optimum of the real plant (Gao and Engell 2005, Gao, Wenzel and Engell, 2016). For control, the multi-stage MPC approach is discussed in which the future information on the realization of the model uncertainty is included in the optimization that is performed at the current time step to reduce the conservatism of the controller (Lucia, Finkler and Engell, 2013).

Engell, S. (2007): Feedback control for optimal process operation. Journal of Process Control 17, 203-219. Gao, W., Engell, S. (2005): Iterative set-point optimization of batch chromatography. Computers and Chemical Engineering 29, 1401-1409. Gao, W., Wenzel, S., Engell, S. (2016): A reliable modifier-adaptation strategy for real-time optimization. Computers and Chemical Engineering 91,318-328. Lucia, S., Finkler, T., Engell, S. (2013): Multi-stage nonlinear model predictive control applied to a semi-batch polymerization reactor under uncertainty. Journal of Process Control 23 1306-1319.

ABSTRACT. The thermo-mechanical processing history together with the steel composition defines the final microstructure, which in turn produces the macroscopic mechanical properties of the final product. In many industrial processes it is therefore of paramount importance to find the optimal thermal path that produces the desired microstructure. In the current study an optimization method has been developed to calculate the optimal thermal path for producing desired amounts of microstructural constituents (ferrite, bainite, martensite) of a medium carbon, low-alloy steel, and a low carbon microalloyed steel. The optimization is performed for two separate industrial processes: induction hardening of a pipeline steel and a water cooling of hot rolled steel strip. The optimization workflow consists of first setting the desired amounts of microstructural constituents, and subsequent optimization of the thermal path, which produces these desired amounts. For the water cooling of a steel strip we additionally employed previously developed tool to calculate the cooling water fluxes that are needed to realize the optimized cooling path in water cooling line after hot rolling. To demonstrate the applicability of the method, we present results that were obtained for different case studies related to the industrial processes.

ABSTRACT. In the current study a previously developed cellular automata (CA) model is applied to test how the shape of the growing bainite sheaves as well as the nucleation and growth rates affect the Kolmogorov-Johnson-Mehl-Avrami (KJMA), Austin-Rickett (AR) and the generalized Lee-Kim (LK) model, which encompasses both the KJMA and AR models. The KJMA, AR and LK models are applied to describe the mean field evolution of the phase transformation kinetics. The shape of the growing sheaf affects the impingement phenomena, which then influences the transformation kinetics, even though the overall growth rate for different shape aspect ratios can be the same. Similarly, the different nucleation and growth rates affect the impingement. Systematic simulation studies were performed with different aspect ratios of the growing sheaves as well as different ratios of nucleation versus growth, to see the effect of these factors on the overall transformation kinetics.

ABSTRACT. An OpenModelica model of the SSAB Raahe Lime kilns has been developed in order to help in formulation of operation strategy and to choose important parameters to measure and monitor. The model is a dynamic simulation model describing the calcination process of limestone in the lime kiln.

ABSTRACT. The world is using a lot of materials in day-to-day life. This requires a lot of mining of ores from the ground. As amount of ore is remarkable, also amount of waste rock and for that reason amount of tailings is huge. The water content of tailings is a subject to decrease. One potential technology for that is a paste thickener. In this paper, a multivariate linear regression model using paste line pressure difference as output variable is described. The model can be utilized for the development of new control strategy. Another model was formed using rake torque as output variable.

ABSTRACT. Lignocellulosic biomass is a sustainable and renewable source for both renewable solids like biochar and biomethane by anaerobic digestion (AD). To improve the total utilization of the biomass, pretreatment methods such as hot water extraction (HWE) have been developed. HWE produces a hydrolysate rich in hemicellulosic sugars. Each sugar has its own characteristics that need to be understood to have a better overview of AD processing of lignocellulosic biomass extracts. A study of the AD process in batch reactors with synthetic substrates composed of hemicellulosic sugars was performed. The process was modelled using the standard IWA Anaerobic Digestion Model No. 1 (ADM1). Simulations were also performed using three strategies: 1) varying the stoichiometry for monosaccharide degradation to the VFAs acetate (fac,su), butyrate (fbu,su) and propionate (fpro,su), 2) varying the maximum uptake rate of the monosaccharide degrading organisms (kmsu) and 3) including a first order rate limiting step.

The ADM1 model is a good tool to predict biogas production from AD of hemicellulosic sugars. The simulation results showed moderate to good agreement with the experimental results. Adjusting the maximum substrate uptake rate of monosaccharides or the stoichiometry for monosaccharide degradation to VFAs during digestion did not improve the simulations significantly. They did however improve significantly upon incorporating a rate limiting step in the form of a first order kinetic rate expression. The extra step accounts for the observed lag-phase of some sugars during AD, thereby simulating possible effects based on types of microorganisms present and diffusion limitations.

ABSTRACT. This study focuses on exctracting ethylene and propylene from other componenents in the pyrolysis gas mixture of plastic waste. Selective absorption of ethylene and propylene from the gas mixture by using a silver nitrate solution, was selected as a promising technology. A lab scale set-up was built and experimental tests were performed using a model gas mixture. Aspen HYSYSV10 was used to model and simulate the absorption process. The model was validated against experimental data. The validated model was further used to identify improvements for the separation process and increase the recovery of ethylene and propylene. The results from the simulated improved process show that the amount of ethylene and propylene in the product gas could be significantly increased. In the experimental study, only 25% of ethylene and propylene in the feed was captured, whereas the simulation of the improved process indicates that the recovery of the monomers could be almost quantitative (99%).  However, the product gas from the separation system contained an undesirable high amount of CO and CO₂. These gases act as pollutants of the polyolefin reaction  and further studies are needed to obtain pure olefin gases.

ABSTRACT. Condensation in raw biogas during compression is a problem because the CO2, water and H2S in the liquid phase is very corrosive. Raw biogas typically contains 60 mol-% methane, 40 mol-% CO2, is saturated with water and may contain contaminants as H2S. Under compression up to 300 bar, it is a question whether condensation (mostly water) will occur. In case of H2S, it is of interest whether it has influence on the dew point temperature. It is also of interest how much H2S will condense in the liquid phase. The aim of this work is to calculate the dew point or condensation under different conditions with varied temperature, pressure and gas composition and using different equilibrium models. Phase envelopes are calculated which show the two-phase area for temperature and pressure for a specified composition. Traditionally, gas mixtures of methane, CO2, H2S and water are calculated in a process simulation program with standard models like Peng-Robinson (PR) and Soave-Redlich-Kwong (SRK). There is traditionally only one constant binary parameter for each component pair. In the process simulation tools Aspen HYSYS and Aspen Plus, the binary parameter for water/CO2 and water/H2S can be made temperature dependent. Earlier evaluations indicate that this may increase the accuracy of condensation detection. Other models are also available in Aspen HYSYS. For mixtures of methane and CO2 with up to 1 mol-% H2S (a high value for biogas), the different models gave similar results. Under normal ambient temperatures (above 0 °C), a mixture with more than 40 mol-% methane will not give any condensation. When the H2S increased from 0 to 1 mol-%, the dew point temperature increased with only 1.0 K. When raw biogas is cooled or compressed, water will condense first. Some H2S will dissolve in the water, and the amount of calculated H2S dissolved in water varied slightly with the model and on the selected binary parameter for water/H2S. For biogas simulation, it is recommended to select a binary parameter that fits the experimental data for H2S solubility in water.

ABSTRACT. This study intends to develop a simple mathematical model that contributes to the integration of Microbial Electrosynthesis (MES) in an anaerobic digester to reduce CO2 to CH4. Open-source modelling language Modelica was used to build the simple model. The MES internal resistances are important parameters for the model and Electrochemical Impedance Spectroscopy (EIS) experiment was employed to estimate the resistances and distinguish the contribution from each resistance element. The model preliminary simulations show that it is possible to determine the voltage required to keep the potential difference across the cathode biofilm within optimal conditions. The system is sensitive to the effects of biofilm development on electron transfer at both electrodes, which implies effects on the electron flow from anode to cathode. The model will be a useful tool for extrapolating experimental results and to enhance our understanding of MES.

ABSTRACT. : In Northern Europe we have many CHP plants operating with biomass and waste as fuel. As more wind power and solar power is introduced the operating hours are reduced and thereby also the capital burden is distributed on fewer annual hours. At the same time we see a strong demand to replace fossil oil for many different purposes by renewable alternatives. Here biomass and waste are the major resources available and to produce liquid or gaseous bio-based products at existing CHP plants make sense. In this study we have simulated system solutions to identify energy and material balances as well as rough economical figures. The products assumed are primarily fuels like diesel, Hydrogen and methane, but also other organic compounds can be considered. Today PREEM and St1 are planning large scale production of primarily bio-diesel, or HVO, where liquid products from both pulp and paper industry and CHP plants will be suitable feed-stocks. We also have compared hydrogen production in gasifiers to electrolysis, and even a combination of these as oxygen from the electrolyzer can be used for the gasification, to avoid ballast of nitrogen in the product gas. The paper will identify optimal solution under different conditions with respect to both electricity and raw material costs, as well as capital cost.

ABSTRACT. Overheating of synchronous generators may lead to shortened generator lifespan, thus strict constraints are imposed on their operation. A dynamic model of the generator temperature may allow for better monitoring of the generator condition, and thus more flexible operation. Øyvang (2018) considered the combination of a thermal model of an air-cooled generator with advanced control to help ride-through grid problems: By implementing a model-based online monitoring and control system, the temperature development in critical locations in the synchronous generator were kept under control. In addition, exploiting the generator’s full thermal capacity lead to improved performance. Pandey et al. (2019) considered various improved thermal generator models, in combination with model fitting and state estimation. Still, the studies so far have used simple, counter-current heat exchanger models with constant Stanton numbers, which allows for an analytic, explicit heat exchanger description.

In practice, heat capacity and heat transfer coefficients in the heat exchanger vary with fluid temperature and velocity. To handle this case, it is necessary to solve a nonlinear two-point boundary value problem numerically for each time step when solving the thermal synchronous generator model. The iterative solution of the boundary-value problem will dramatically increase the overall solution time for the generator model, and it is of interest to study how the solution time can be reduced.

The general idea pursued in this paper is to develop a correction term to the analytic, explicit heat exchanger model. This is done by solving the nonlinear boundary value problem multiple times off-line, and then fitting a regressor model of the correction term. This combined mechanistic, analytic model with a data-driven correction term is termed a hybrid heat exchanger model. We consider both linear regression and nonlinear regression (e.g., forward neural network) to find the correction term. Both the accuracy and the execution speed of the hybrid heat exchanger model will be discussed, as well as the potential for reduced computation time when solving the thermal generator model. The work is carried out using the modern computer science language Julia with its easy-to-use machine learning tools.

References:

Pandey, Madhusudhan & Øyvang, Thomas & Lie, Bernt. (2020). State Estimation of a Thermal Model of Air-cooled Synchronous Generator. 190-197. 10.3384/ecp20170190.

Thomas Øyvang (2018). Enhanced power capability of generator unites for increased operational security. PhD thesis, University of South-Eastern Norway, Faculty of Technology, Natural Sciences and Maritime Sciences University of South-Eastern Norway N-2018 Porsgrunn Norway, December 2018. ISBN: 978-82-7206-503-3 (print) ISBN: 978-82-7206-504-0 (online).

Bernt Lie (2018). Solution, Project, FM1015 Modelling of Dynamic Systems. University of South-Eastern Norway, November 2018.

ABSTRACT. In this study, a model of a single shaft gas turbine (GT) is developed by using artificial intelligence (AI). A recurrent neural network (RNN) is employed to train the datasets of the GT variables in Python programming environment by using Pyrenn Toolbox. The resulting model is validated against the Test datasets. Thirteen significant variables of the gas turbine are considered for the modelling process. The results show that the RNN model developed in this study is capable of performance prediction of the system with a high reliability and accuracy. This methodology provides a simple and effective approach in dynamic simulation of gas turbines, especially when real datasets are only available over a limited operational range and using simulated datasets for modelling and simulation purposes is unavoidable.

ABSTRACT. Biomass fired boilers usage is increasing due to supportive policies and economic trends. Fluidized bed technology is identified as proper solution for lower quality fuels such as biomass. Moisture and heating value can vary significantly in biomass fuels. Without real-time information on their variation, they are a disturbance to the system. These disturbances affect the system steady state and decrease operational efficiency. Proper characterization of the disturbance enables the use of Feed-Forward control. Feed-Forward makes use of the knowledge about the updated condition of the fuel and can act towards reducing the impact of the fuel on offsetting the system. Feed-Forward Model Predictive Control is proposed as new control strategy. Comparison is made between the existing control strategy and the new proposed solution. Control performance is evaluated on three process outputs, in three different scenarios. Adding feed-forward signal for fuel moisture improves control performance in both controllers, while ultimately Feed-Forward Model Predictive Control shows the best performance in most comparison metrics.

ABSTRACT. Large amounts of digital data requires to be stored and for that data centers are built around the globe. Electric power is the main energy input and heat is the main energy output from these data centers. This work is about utilization of the excess heat which is the by-product of data center operation. To connect the heat from data centers to a district heating network, a heat pump might be necessary to increase the temperature of the heat. The economic potential for different conditions and different heat recovery solutions are evaluated. Simulations and economical optimization at different conditions in Aspen HYSYS were carried out. Especially three alternatives were evaluated. The first is an alternative without a heat pump in which the cooling water leaves the data center at 80 ºC and enters the district heat network at 70 ºC. The second is an alternative with a slight temperature increase in the heat pump. The cooling water temperature from the data center is 65 ºC and the temperature to the district heat system is 70 ºC. The third is an alternative with a higher temperature increase in the heat pump. The cooling water temperature from the data center is 65 ºC and the temperature to the district heat system is 80 ºC. The COP (Coefficient of Performance) in a heat pump for these alternatives were calculated using the refrigerant R-22 in the simulation program Aspen HYSYS. The estimated economic potential for each alternative was calculated by estimated values on electricity cost and district heat price. In one alternative, the electricity cost was specified to 0.1 EUR/kWh, and the district heat price was specified to 0.05 EUR/kWh. For the alternatives using heat pumps, the capital cost was estimated assuming that the heat pump investment was dominating. The COP for the two heat pump alternatives were calculated to be 8.7 and 5.4, respectively. A large data center facility with recovered waste heat of 200 GWh/year were calculated over a period of 10 years. For the specified conditions, the net present value was calculated to be large and positive for all the alternatives. As expected, the most economical alternative was without a heat pump, and the most economical heat pump was the one with the highest COP. Sensitivity calculations were performed to show dependencies of temperatures, district heating price, electricity cost, heat pump cost, COP and pipeline cost.

ABSTRACT. When we are studying gas explosions it is a challenging task to find the position, the velocity, and the shape of the flame and the shock wave. In lab experiments, state-of-the-art has been to use the sensors like pressure transducers to measure velocities. One drawback of this technique is the limited number of sensors that can be placed inside a test rig. An alternative to these sensor-based measurements is to use a high-speed camera whereupon the position, shape, and velocity can be derived from the recorded images.

An Unnormalized Optimal Transport framework was implemented in Python to provide the route of propagation by the recorded high-speed video. This route was then used for front tracking by three different methods. These methods are classified as Divergence, Transport, and Matrix method. The Transport and the Divergence method were analyzed with both synthetic images and recorded high-speed video frames. Unfortunately, the Matrix method was only applicable to noise-free synthetic images. The Transport method provided better results than the Divergence method. The velocity of the shock wave and the shock angles were therefore calculated using the front tracked from Unnormalized Optimal Transport in combination with the Transport method. Preliminary results indicate that our findings are in accordance with results obtained with sensor-based measurements. Moreover, the Unnormalized Optimal Transport framework can also uncover velocities between the location of the transducers.

ABSTRACT. The global electric mobility is rapidly expanding. Hence, the demand for lithium-ion batteries are also increasing fast. Therefore, understanding the energy requirements in this rapidly growing industry is crucial for accurate estimation of the environmental impact as well as identifying the appropriate measures to reduce energy requirements. The biggest contribution to greenhouse gas emissions is the cell manufacturing process. The drying of the electrode slurry is the most energy intensive step in the cell manufacturing process. Therefore, we develop a dynamic model for electrode drying, that can be for analysis of different drying technologies, energy requirement calculations, and optimization and control of the process.

ABSTRACT. Pipe components such as elbow, tees, and weldolets are produced at various places in the world (China, Italy and USA). Although the component is produced in accordance with standard ASME (American Society of Mechanical Engineers) B16.9, measurements show variation in geometry. In piping component design, the geometries are assumed as perfect in accordance with producer’s specification. Unfortunately, ASME B16.9 specification is limited on the details and allows for deviations within a product. For example, a long radius bend has outer diameter and bend radius. Often a specification does not include more than the minimum standards which allows for significant wiggle room for the fabrication process. This requires a conservative approach in design that results in incorrect dimensioning of the components. This leads to extreme scenarios from unnecessary use of materials and waste of limited resources to weak fittings that can break before the expected lifetime is reached. Hydrogen Induced Stress Cracking (HISC) is a major concern when piping components made of duplex and super-duplex are used for subsea installations. HISC occurs due to hydrogen diffusion into the piping which may under high stress causes the pipe component to crack. The focus of this article is based on the standard ASME B16.9 long radius elbows. The pipe size and schedule range is NPS 2-8 and SCH 80-120-160, made of duplex and super duplex materials. A series of measurements were taken utilizing an ultrasonic thickness gage. Measurements were taken at critical locations of the elbow at 5 cross sections including the intrados and extrados. From the results of the measurements, the critical characteristics and dimensions are identified. The results are also compared with the nominal straight pipe dimensions in ASME. The deviation is taken as basis for further analysis using Finite Element Method. A model of the elbow is developed using SolidWorks. Series of simulations were carried out using ANSYS. The simulations compare the strength of the pipe fittings against their straight pipe counterparts with regards to internal pressure. This study gives an overview of the deviations and utilization based on the measurements taken against the standard ASME B16.9 geometry.

ABSTRACT. The aim of this paper is to estimate the parameters of the model of Pressure Peaking Phenomena (PPP). This project focuses on the investigation of the overpressures arising from the ignited hydrogen releases in 14.9m3 enclosure (explosive chamber) through a 4mm nozzle. The various ventilation areas and mass flow rates were applied in 31 tests. The controlled variables for experiments are mass flow rate (MFR, g/s), ventilation area (Av, m2), and time of hydrogen releases (t, s). The Bayesian approach was implemented in the parameter estimation using Markov chain Monte Carlo method for simulations. The discharge coefficient and heat loss coefficient has been analyzed and gave by posterior distribution.

ABSTRACT. In this research project, acoustic chemometrics was assessed as a new method for both classification and prediction of flow rate of five selected gas types. The gas types were selected to span different densities as much as possible while at the same time being relatively safe to use. The five gas types were Argon, Helium, Carbon dioxide, Nitrogen and Air. The research questions were 1) Can measurements of the vibrations in a gas control valve in combination with signal processing and unsupervised learning be used to classify the five gases mentioned above? 2) Can the vibrations in the gas valve in combination with supervised learning be used to determine the flow rate of the five gases? 3) Can a simple low cost piezo disk provide signals comparable to that of an industrial accelerometer? The results show that it is possible to classify the five gas types based on principal component analysis with three components. The gas flow rate could also be predicted for all five gases based on partial least squares regression with an average error of 2-5%. The Piezo disk could not be used for gas classification, but for prediction of gas flow rates it was comparable to the accelerometer. All the prediction models were validated based on independent data.

ABSTRACT. A feasibility study was carried out to assess the possibility of developing prediction models for monitoring drying conditions of wood coatings in one of Europe’s largest and most modern coating plants for exterior cladding. These models were based on data from real-time Process Analytical Technology (PAT) sensors, measuring airflow and air direction, temperature and relative humidity). The study revealed that the information from the PAT sensors gave sufficient input to accurately model the complex drying conditions and their interrelations. Modelling was carried out using both Principal Component Analysis (PCA) and PLS-regression in both its PLS1 and PLS2 manifestations. In addition, the diagnostic prediction performance RMSEP between PLS1 and PLS2 models were not significantly different. This is advantageous for an industrial implementation concerning re-calibration operations: PLS1 requires 40 separate calibrations whereas PLS2 requires only one, because PLS1-R is a regression of a singular output variable (y-variable) and PLS2-R of several simultaneous, correlated output variables. While a single calibration based on PLS2 will take approximately one hour, the PLS1 approach will take more than a week.

ABSTRACT. Intelligent methodologies are beneficial in developing un- derstandable multimodel simulation solutions. Nonlinear scaling extends these applications by facilitating compact nonlinear approaches already at the basic level. Compos- ite local models can continue using linear methodologies for various case-based models. The flexible handling of the new structures and recursive tuning are the keys in adapting the systems in varying operating conditions. The recursive tuning of the scaling functions has two levels: smooth adaptation and strong shape changes. Fuzzy set systems further extend application areas of the models by combining composite local models in a flexible way. The extensions of the data-based methodologies are suitable for developing these adaptive applications via the follow- ing steps: variable analysis, linear models and intelligent extensions. Evolutionary computation is used in the tun- ing of the resulting complex models both the scaling and interactions. Complex problems are solved level by level to keep the domain expertise as an essential part.

ABSTRACT. This study reports a numerical analysis of the performance of green facades in different geographical locations and seasonal conditions. A mathematical model from a previous study is implemented and combined with the modified convective heat transfer coefficients from a recent study of the literature to simulate the transient heat transfer through bare walls and green facades with climbing vegetation. Implicit Finite Difference Method (FDM) based solver is used to perform the numerical simulations. Climate data are taken from relevant weather stations of Oslo and Rome and typical meteorological year (TMY) values are used for this purpose together with variable thermo-physical properties of air. An energy budget analysis reveals that the short-wave radiation term and convective heat transfer term are predominating compared to the other terms involved in the energy balance equation for summer time. The results show that the green walls are most effective in summer seasons with high levels of solar radiation, as most of the cooling effect is credited to the vegetation blocking the solar radiation. In cooler seasons, the benefit is less prominent. Furthermore, an analysis of the effects of the different models of convective heat transfer coefficients is presented.

ABSTRACT. To implement large-scale agent-based simulations, developers historically relied on C and C++ due to performance, while struggling to deal with tedious explicit memory management. This struggle translates into software defects and lower developer productivity. More recently, desire to harness multi-core systems via concurrent software complicates design and implementation when memory is shared among compute cores. When we faced this situation, we were looking for a system programming languages as fast as C and C++ but without caveats around memory management. Between Go and Rust, we chose Rust language which guarantees safety in memory management even for concurrency, without a run-time or garbage collector. In this paper, we intend to share our experience with Rust to build a framework for agent-based epidemiological simulations. Our experiments show promising results for a million agents, all running on commodity-class hardware. Key outcomes of this whole exercise are: 1. Performance 2. Flexibility 3. Robustness.

ABSTRACT. This paper is an investigation of automated netlist synthesis for ternary-valued n-ary logic functions, based on a static ternary gate design methodology. We present an open-source C++ implementation, which outputs a ready-to-simulate SPICE subcircuit netlist file for ternary-valued n-ary function circuits. A circuit schematic of the 3-operand carry is demonstrated as synthesized by the netlist generator.

We investigate a holistic (non-compound) approach to designing balanced full-adders by using 3-operand functions as compared to a traditional 2-operand compound design methodology. Three gate-level design approaches (compound, non-compound and hybrid) for the balanced full-adder have been simulated in HSPICE and are compared to each other and the state-of-the-art with simulation results.

Furthermore, we propose to standardize the ternary functions by indexing them. This indexing system allows for the convenience of referencing any possible logic function with no ambiguity. This indexing is necessary as most ternary functions do not have semantic names (e.g. AND, OR) and the amount of unique 3-valued functions grows exponentially with higher arity.