Extrapolation Capability Research Articles

Thermoelastic interaction in the semi-infinite solid medium due to three-phase-lag effect involving memory-dependent derivative

The present study deals with the thermoelastic interaction in a semi-infinite elastic solid with a heat source in the context of three-phase-lag model with memory-dependent derivative. The governing coupled equations, involving time delay and kernel functions are expressed in the vector matrix differential equation form in the Laplace transform domain. The analytical formulations of the problem have been solved by eigenvalue technique. The Honig–Hirdes numerical method is used for the inversion of Laplace transformation. Numerical results are obtained by choosing various types of time delay parameters and kernel functions and graphical representations have been performed accordingly. An extrapolative capability is established by considering the memory-dependent derivative into a three-phase-lag model.

Data-Driven Iterative Feedforward Tuning for a Wafer Stage: A High-Order Approach Based on Instrumental Variables

The feedforward controller plays an important role in the achievement of high servo performance of wafer scanning. In this paper, a novel data-driven feedforward tuning method is developed in the presence of noise. Three distinguished features make it different from the existing methods: first, high extrapolation capability to tasks; second, low requirement on the system model; and especially, third, high noise tolerant capability. These superiorities are achieved by a high-order iterative feedforward tuning algorithm based on instrumental variables. It utilizes error data from all past iterations via an integrator in the learning law, yet without the need of the plant model or the sensitivity function. Furthermore, $H_2$ optimization with specified convergence speed constraint is proposed to design the learning gain. Connections and differences between the proposed algorithm and the existing ones are discussed. Experimental results validate the proposed method and confirm its effectiveness and superiority.

110th Anniversary: Accurate Modeling of the Simultaneous Absorption of H2S and CO2 in Aqueous Amine Solvents

We present results concerning the application of soft statistical associating fluid theory (soft-SAFT), a molecular-based equation, for modeling the absorption of carbon dioxide and hydrogen sulfide in several aqueous amines, including monoethanolamine, diethanolamine, and N-methyldiethanolamine, at conditions of relevance for acid gas separation. The chemisorption of CO2 and H2S in aqueous amines was modeled through the formation of physically bounded CO2-amine and H2S-amine aggregates by explicitly considering strong intermolecular association forces. This approach eliminates the need to consider all speciation reactions, significantly reducing the number of fitted parameters required to describe the absorption process while providing a direct connection between each amine and the chemical reaction with the acid gases. The model predicts the absorption of H2S for different amine concentrations in excellent agreement with experimental data, showing high extrapolative capability. Soft-SAFT is also used to...

Monotonic Smooth Takagi–Sugeno Fuzzy Systems With Fuzzy Sets With Compact Support

In this paper, sufficient conditions on monotonicity of Takagi–Sugeno (T–S) fuzzy systems with linear consequent parts with smooth membership functions with compact support (i.e., equal to zero outside a compact set) are derived. The presented conditions considerably reduce conservatism of those known for T–S fuzzy systems utilizing membership functions with the support that is not compact. Besides, two illustrative examples will demonstrate that the monotonicity requirement imposed on fuzzy systems may significantly improve their extrapolation capability.

Performance prediction of a reciprocating piston expander with semi-empirical models

The aim of this study is to characterize a prototype of reciprocating piston expander integrated into a micro-ORC system test bench (in the kW range of power), installed at the laboratory of the university of Bologna. In order to simulate behavior and performances of the expander in not yet explored operating conditions, two semi-empirical models proposed in the literature have been opportunely adapted to the case of study and calibrated over a full set of available experimental data. One model is based on polynomial correlations on the expander efficiencies, whereas the other one is based on a lumped parameters approach with a more physical sense. Both the models have been evaluated on the error on predict the outputs and compared into performance prediction maps. The preliminary results demonstrate that the polynomial fitting functions model is the most accurate in predicting the outputs in the range of the explored working conditions. However, in order to verify the models extrapolation capability, more experimental points should be collected. The validation of the models outside the calibration range will be object of further investigations.

Artificial neural network modeling of hydrogen sulphide gas coolers ensuring extrapolation capability

Artificial neural network modeling of hydrogen sulphide gas coolers ensuring extrapolation capability

A systematic approach for the thermodynamic modelling of CO2-amine absorption process using molecular-based models

The development of new amine systems for CO2 capture is a topic of high interest because of the limitations current aqueous amine systems have for capturing CO2 at large scale. Having a robust and systematic approach for describing the absorption of CO2 would help accelerating the discovery of high performance amine solvents. In this contribution, a molecular-based equation of state is applied to describe the absorption of CO2 in aqueous solutions of single and blended amines at conditions of relevance for post-combustion CO2 capture. A scheme of implicit reactions is used to describe the formation of carbamate and/or bicarbonate products resulting from the chemical reactions between CO2 and five amines of practical industrial interest. This procedure eliminates the need to specify the detailed equilibrium reactions and significantly reduces the number of parameters required to represent the absorption process. A maximum of two adjustable model parameters (one of which with a linear temperature dependence), optimised for a fixed amine concentration, suffices to represent the absorption of CO2 in aqueous solutions of single amines over a broad range of temperatures (298–413 K) and partial pressures of CO2 (0.1–1000 kPa). The extrapolation capabilities of the model are tested by predicting the absorption of CO2 in aqueous solutions of single amines for different amines concentrations (∼8.5–35 wt%), with modelling results showing good quantitative agreement with solubility, speciation and enthalpy of absorption data available in literature. Furthermore, without introducing any new model parameter, the absorption of CO2 in various amine blends is satisfactorily predicted by considering competing interactions for the reactive sites in the model of CO2. The developed models are then used to assess the CO2 capture performance of selected amine systems in terms of two key process parameters: solvent cyclic capacity and regeneration energy. Results for systems with the same total amine mass concentration show that the highest molar cyclic capacities are obtained for 30 wt% piperazine (0.45 molCO2.molAmine-1), whereas the greatest energy savings for solvent regeneration are estimated for 30 wt% methyldiethanolamine (2.3 GJ.tCO2-1). Moreover, two piperazine–promoted blends showed the potential for reducing up to ∼26% the energy consumption for solvent regeneration and separating up to ∼41% more CO2 in a molar basis when compared to the benchmark 30 wt% monoethanolamine. Altogether, these results demonstrate the feasibility of the developed approach as a reliable platform for the screening of amine solvents as function of key process parameters, and as a valuable tool for process modelling.

Physics-based Learning for Aircraft Dynamics Simulation

The accurate prediction of flight trajectories is crucial for the real-time prognostics of air transportation system. However, the computation costs of predictions can be expensive or even prohibitive especially for a large number of aircrafts in the air traffic system. This study proposes the concept of physics-based learning, a hybrid approach based on data-driven learning and physical models, as a computationally efficient method for the simulation of aircraft dynamics. The physics-based learning integrates the underlying physics of dynamical systems into learning models such as neural networks to reduce the training and simulation costs. The application of physics-based learning for simulating aircraft dynamics is demonstrated using a recently introduced physics-aware network known as the deep residual recurrent neural network (DR-RNN) on a Boeing 747-100 aircraft. The aircraft dynamics are described using a six degrees-of-freedom aircraft model. The DR-RNN is first trained using the simulated responses of the aircraft and then the trained network is used to predict the response of aircraft under arbitrary control inputs and disturbances. The results show that the DR-RNN can accurately predict aircraft responses and has excellent extrapolation capabilities. Moreover, the DR-RNN exhibits superior computation efficiency compared with a classical numerical method, the fourth-order Runge-kutta method, highlighting its suitability in serving as surrogating models for aircraft dynamical systems.

Robustness of Extreme Learning Machine in the prediction of hydrological flow series

Prediction of hydrological flow series generated from a catchment is an important aspect of water resources management and decision making. The underlying process underpinning catchment flow generation is complex and depends on many parameters. Determination of these parameters using a trial and error method or optimization algorithm is time consuming. Application of Artificial Intelligence (AI) based machine learning techniques including Artificial Neural Network, Genetic Programming (GP) and Support Vector Machine (SVM) replaced the complex modeling process and at the same time improved the prediction accuracy of hydrological time-series. However, they still require numerous iterations and computational time to generate optimum solutions. This study applies the Extreme Learning Machine (ELM) to hydrological flow series modeling and compares its performance with GP and Evolutionary Computation based SVM (EC-SVM). The robustness and performance of ELM were studied using the data from two different catchments located in two different climatic conditions. The robustness of ELM was evaluated by varying number of lagged input variables the number of hidden nodes and input parameter (regularization coefficient). Higher lead days prediction and extrapolation capability were also investigated. The results show that (1) ELM yields reasonable results with two or higher lagged input variables (flows) for 1-day lead prediction; (2) ELM produced satisfactory results very rapidly when the number of hidden nodes was greater than or equal to 1000; (3) ELM showed improved results when regularization coefficient was fine-tuned; (4) ELM was able to extrapolate extreme values well; (5) ELM generated reasonable results for higher number of lead days (second and third) predictions; (6) ELM was computationally much faster and capable of producing better results compared to other leading AI methods for prediction of flow series from the same catchment. ELM has the potential for forecasting real-time hydrological flow series.

Modeling performance of response surface methodology and artificial neural network

In recent years, response surface methodology (RSM) which is a statistical technique and artificial neural network (ANN) a soft computing technique have been highly used for modelling, simulation and optimization of several physical processes in engineering. Both RSM and ANN strategies have particular computational properties that makes them suitable for making predictions, but differ in their extrapolation and interpolation capabilities on complex non-linear processes, and thus potentially conflict in their predictive accuracy. This study models and compares the capabilities of RSM and ANN in predicting the tensile strength of a 6 mm thick mild steel gas tungsten arc welded plate based on the effects of input variables such as weld current, weld speed, gas flow rate and filler rod. The RSM and ANN based models for prediction were compared using the coefficient of determination criteria. With a higher value of 0.836, the ANN model proved to be a better modeling technique than the RSM model. Keywords: Soft Computing Techniques, Response Surface Method, Artificial Neural Network

LIQUID-LIQUID EQUILIBRIA FOR SYSTEMS CONTAINING FATTY ACID ETHYL ESTERS, ETHANOL AND GLYCEROL AT 333.15 AND 343.15 K: EXPERIMENTAL DATA, THERMODYNAMIC AND ARTIFICIAL NEURAL NETWORK MODELING

Abstract In this study, the liquid-liquid equilibrium (LLE) data of systems containing ethyl linoleate/oleate/palmitate/laurate, ethanol and glycerol at temperatures ranging from 323.15 to 353.15 K were used to evaluate the performance of the NRTL, UNIFAC, Cubic-Plus-Association Equation of State (CPA EoS), and artificial neural network (ANN) models. The systems evaluated correspond to the most important components formed at the end of the ethanolysis reaction of soybean, palm and coconut oils. The temperature range selected is very important for heterogeneous catalysts, especially for high-pressure systems. The accuracy of the models was evaluated by average global deviation. UNIFAC, UNIFAC-LLE and CPA EoS models showed lower accuracy with deviations of 10.1, 8.01 and 5.95%, respectively. In spite of this predictive limitation, these models show high extrapolation capability for the description of LLE behavior when few experimental data are available in the literature. The ANN model shows the best agreement between experimental and predicted data with an average deviation of 1.12%. In this regard, ANN is offered in this work as an alternative to equations of state and activity coefficient models to be used in a more reliable and less cumbersome way for process simulators of biodiesel production and separation equipment design.

Application of the hybrid ANFIS models for long term wind power density prediction with extrapolation capability.

In this paper, the suitability and performance of ANFIS (adaptive neuro-fuzzy inference system), ANFIS-PSO (particle swarm optimization), ANFIS-GA (genetic algorithm) and ANFIS-DE (differential evolution) has been investigated for the prediction of monthly and weekly wind power density (WPD) of four different locations named Mersing, Kuala Terengganu, Pulau Langkawi and Bayan Lepas all in Malaysia. For this aim, standalone ANFIS, ANFIS-PSO, ANFIS-GA and ANFIS-DE prediction algorithm are developed in MATLAB platform. The performance of the proposed hybrid ANFIS models is determined by computing different statistical parameters such as mean absolute bias error (MABE), mean absolute percentage error (MAPE), root mean square error (RMSE) and coefficient of determination (R2). The results obtained from ANFIS-PSO and ANFIS-GA enjoy higher performance and accuracy than other models, and they can be suggested for practical application to predict monthly and weekly mean wind power density. Besides, the capability of the proposed hybrid ANFIS models is examined to predict the wind data for the locations where measured wind data are not available, and the results are compared with the measured wind data from nearby stations.

Review and Thermodynamic Modeling with NRTL Model of Vapor–Liquid Equilibria (VLE) of Aroma Compounds Highly Diluted in Ethanol–Water Mixtures at 101.3 kPa

A review of the vapor–liquid equilibrium data of aroma compounds highly diluted in hydroalcoholic mixtures at 101.3 kPa is presented. The study includes 44 aroma compounds present in distilled beverages from seven chemical families: acetals, alcohols, carbonyl compounds, carboxylic acids, esters, furans, and terpenes. The equilibrium data are modeled using the ideal gas hypothesis (with a correction term for dimerization in the case of carboxylic acids) and the NRTL model. A set of binary interaction parameters is generated, and the quality of the representation is evaluated. A classification of the aroma compounds in terms of their relative volatility with respect to ethanol and water is proposed over the whole ethanol composition range in the liquid phase. Finally, a comparison with the representation obtained when using interaction parameters calculated from binary and ternary mixture data at high concentrations is performed in order to evaluate the extrapolation capability of the NRTL model.

A comparison of the effect of empirical and physical modeling approaches to extrapolation capability of compressor models by uncertainty analysis: A case study with common semi-empirical compressor mass flow rate models

Some semi-empirical compressor models are claimed to be more accurate at extrapolation conditions than their empirical counterparts which have a long history of industrial applications due to their uses of physical principles, but it is unknown how much improvement the principles can bring to the modeling of extrapolation scenarios quantitatively. This paper studies the effect of the number of empirical coefficients and physical principles on model accuracy and uncertainty by comparing the estimation of five regression models of compressor mass flow rates. The choice of model training data follows the industrial norm, and model accuracy and uncertainty are calculated. The quantitative results show that the use of neither empirical coefficients nor physical principles guarantees good accuracy and reliability. If a coefficient is redundant to explain the behavior of the phenomenon, regardless of its empirical or physical origin, it should be removed to reduce model inaccuracy in extrapolation scenarios.

Saturated liquid density equation for pure refrigerants including CFCs, HCFCs, HFCs, HCs, HFOs, HFEs, PFAs and ISs based on the scaling law and the law of rectilinear diameter

This work aims to develop a simple saturated liquid density equation (SLDE) with good modeling, extrapolation and prediction capabilities for pure refrigerants based on the well-known scaling law and law of rectilinear diameter. Saturated liquid densities, critical temperatures and critical densities of 53 refrigerants from 8 different categories including chlorofluorocarbons (CFCs), hydrochlorofluorocarbons (HCFCs), hydrofluorocarbons (HFCs), hydrocarbons (HCs), hydrofluoroolefins (HFOs), hydrofluoroethers (HFEs), perfluoroalkanes (PFAs) and inorganic substances (ISs) have been studied using the proposed SLDE. The equation exhibits a good modeling behavior with a total mean absolute relative deviation (MARD) of 0.105%, and presents reasonable extrapolation accuracy, especially the extrapolations from medium temperature range to low and high temperature ranges, with total MARDs of 0.286% and 0.467%, respectively. The three parameters in the equation were generalized for different categories to better predict saturated liquid densities. The generalized equation shows a more competitive prediction performance than the commonly used generalized SLDEs, with a total MARD of 0.783%. Furthermore, critical temperatures and critical densities were simultaneously predicted using the generalized equation with total MARDs less than 0.743% and 0.998%, respectively.

Artificial neural networks with an infinite number of nodes

A new class of Artificial Neural Networks is described incorporating a node density function and functional weights. This network containing an infinite number of nodes, excels in generalizing and possesses a superior extrapolation capability.

Autoregressive modeling of cycle-to-cycle correlations in homogeneous charge compression ignition combustion

Homogeneous charge compression ignition or gasoline controlled auto-ignition combustion is characterized by a strong coupling of consecutive cycles, which is caused by residuals from the predecessor cycle. Closed-loop combustion control is considered a promising technology to actively stabilize the process. Model-based control algorithms require precise prediction models that are calculated in real time. In this article, a new approach for the transient measurement of the auto-ignition process and the data-driven modeling of combustion phasing and load is presented. Gasoline controlled auto-ignition combustion is modeled as an autoregressive process to represent the cycle-to-cycle coupling effects. The process order was estimated by partial autocorrelation analysis of steady-state operation measurements. No significant correlations are found for lags that are greater than one. This observation is consistent with the assumption that cycle coupling is mainly caused by the amount of exhaust gas that is directly transferred to the consecutive combustion. Because steady-state operation results in a hard coupling of actuation and feedback variables, only minor variations of the test data can be achieved. The steady-state tests delivered insufficient data for the generalized modeling of the transient autoregressive effects. A new transient testing and measurement approach is required, which maximizes the variation of the predecessor cycle’s characteristics. Dynamic measurements were performed with the individual actuation of the injection strategy for each combustion cycle. A polynomial model is proposed to predict the combustion phasing and load. The regression analysis shows no overfitting for higher polynomial orders; nevertheless, a first-order polynomial was selected because of the good extrapolation capabilities of extreme outliers. The prediction algorithm was implemented in MATLAB/Simulink and transferred to a real-time-capable engine control unit. The verification of the approach was performed by test bench measurements in dynamic operation. The combustion phasing was precisely predicted using the autoregressive model. The combustion phasing prediction error could be reduced by 53% in comparison to a state-of-the-art mean value-based prediction. This work provides the basis for the development of a closed-loop autoregressive model-based control for gasoline controlled auto-ignition combustion.

Optimal Estimation of Rational Feedforward Control via Instrumental Variables: With Application to a Wafer Stage

AbstractIterative control enables a significant control performance enhancement by learning feedforward command signals from previous tasks in a batch‐to‐batch fashion. The aim of this paper is to develop an approach to estimate the parameters of rational feedforward controllers that provide high performance and extrapolation capabilities towards varying tasks. An instrumental variable‐based algorithm is developed that leads to unbiased parameter estimates and optimal accuracy in terms of variance. Furthermore, a noncausal implementation of rational feedforward controllers is proposed, aiming to improve performance by means of pre‐actuation. Simulation and experimental results are presented to confirm that optimal accuracy is obtained with the proposed approach, and show the advantages of pre‐actuation in terms of performance.

Local up-sampling and morphological analysis of low-resolution magnetic resonance images

Limitations in the resolution of acquired images, which are due to sensor manufacturing and acquisition conditions, are reduced with the help of algorithms that enhance the spatial resolution by assigning pixel values that are interpolated or approximated from known pixels. We propose a variant of the moving least-squares approximation for image up-sampling, with a specific focus on biomedical MR images. For each evaluation point, we locally compute the best approximation by minimizing a weighted least-squares error between the input data and their approximation with an implicit function. The proposed approach provides a continuous approximation, an accuracy and extrapolation capabilities higher than previous work, and a lower computational cost. As main application, we consider the up-sampling of low field MR images, where the volumetric and meshless properties of the approximation allow us to easily process images with anisotropic voxel size by rescaling the image and inter-slices resolution. Finally, we include the resolution rescaling into a pipeline that performs a morphological characterization of 3D anatomical districts, which has been developed with a focus on rheumatoid arthritis evolution and provides a more accurate segmentation as an input to quantitative analysis.

Extrapolative Scaling Expression: A Fitting Equation for Extrapolating Full Data Matrixes From Limited Data.

Scaling analysis of several thousand Nb3Sn critical-current measurements is used to derive the extrapolative scaling expression (ESE), a fitting equation that can quickly and accurately extrapolate (or interpolate) limited datasets to obtain full three-dimensional dependences of on magnetic field , temperature , and mechanical strain . Unlike nonextrapolative fitting equations, the ESE relation is determined completely by fundamental raw scaling data from master pinning-force curves. The results show that extrapolation errors with ESE approach typical measurement errors. The scaling expression is simple and robust, providing straightforward extrapolation capability for conductor characterization and magnet design. The ESE relation offers the prospect for extrapolations in several new areas, including: a reduction in the measurement space required for full characterization to about one fifth the size; ability to combine data from separate temperature and strain apparatuses; extrapolation of transport data from above 4 K to lower temperatures, where heating effects and instabilities are problematic for transport measurements; and extrapolation of full datasets from as little as a single curve when several core parameters have been measured in similar conductors (particularly applicable to qualifying production wires). Accuracies are evaluated for concatenations of these different extrapolation capabilities. Examples are given for practical Nb3Sn conductors, including those for high luminosity-LHC magnets, ITER, and cryo-cooled NMR magnets.