Hybrid Models

description152 papers

group6 followers

lightbulbAbout this topic

Hybrid models refer to analytical frameworks that integrate multiple methodologies or approaches, often combining qualitative and quantitative techniques, to enhance understanding and prediction in complex systems. These models leverage the strengths of different paradigms to address research questions more comprehensively and effectively.

lightbulbAbout this topic

Key research themes

1. How can hybrid modeling integrate top-down and bottom-up approaches to better represent technological and economic interactions in energy systems?

This research area focuses on overcoming the limitations of conventional top-down and bottom-up models in energy-economy systems by combining the detailed technological specificity of bottom-up models with the economic feedbacks captured by top-down models. The hybrid approaches seek to reflect induced technical changes, consumer and firm decision-making more realistically, and monetary-economic feedbacks, thus offering more robust policy analysis tools for energy environment transitions.

Hybrid Modeling: new Answers to old Challenges introduction to the Special issue of The Energy Journal

by 晟瑄张 and

2019

Key finding: This paper synthesizes a workshop's findings illustrating that hybrid modeling methods integrate detailed technological substitution processes of bottom-up models with the economic structure and trade feedbacks from... Read more

articleView Paper downloadDownload

keyboard_arrow_downShow more

2. What methodologies enable effective hybrid modeling and prediction of dynamical systems by blending mechanistic and data-driven representations?

This theme addresses the challenge of leveraging the strengths of parametric mechanistic models and nonparametric data-driven methods to improve modeling and forecasting of complex dynamical systems. Hybrid modeling methodologies replace a subset of mechanistic model equations with nonparametric representations trained from data, enhancing robustness and predictive accuracy, especially under parameter uncertainty, noise, or chaotic system behavior.

Hybrid modeling and prediction of dynamical systems

by Kevin Flores

2021, PLoS computational biology

Key finding: This work presents a systematic hybrid modeling framework wherein subsets of mechanistic model equations are substituted with data-driven nonparametric components, combining advantages of both approaches. Experimental studies... Read more

articleView Paper downloadDownload

keyboard_arrow_downShow more

3. How can component-oriented hybrid dynamic system modeling using meta-programmable environments enhance compositionality and simulation of complex physical systems?

This research area investigates developing hybrid dynamic physical system models through object-oriented, component-based approaches supported by meta-programming tools. The aim is to enable modular construction of complex, multi-domain hybrid models (combining continuous and discrete dynamics), improve model reusability, and automate synthesis of simulation artifacts to handle system complexity effectively.

Component-oriented modeling of hybrid dynamic systems using the Generic Modeling Environment

by Gautam Biswas

2024, Fourth Workshop on Model-Based Development of Computer-Based Systems and Third International Workshop on Model-Based Methodologies for Pervasive and Embedded Software (MBD-MOMPES'06)

Key finding: The paper demonstrates the formulation of hybrid physical system models based on bond graphs within the Generic Modeling Environment (GME) by defining reusable component libraries. It shows that through plug-and-play... Read more

articleView Paper downloadDownload

keyboard_arrow_downShow more

4. What formal constraint-based approaches facilitate verification of hybrid systems by discovering inductive invariants expressed as complex polynomial inequalities?

Focusing on verification of hybrid systems, this theme studies constraint-based techniques that systematically search for inductive invariants, including boolean combinations of polynomial inequalities, via formulation as quantified SMT constraints. Innovative use of mathematical tools such as Farkas lemma allows transforming the verification problem into satisfiability problems solvable by modern SMT solvers, enabling formal safety proofs for complex hybrid dynamical models.

Constraint-Based Approach for Analysis of Hybrid Systems

by Ashish Tiwari

2024, Lecture Notes in Computer Science

Key finding: The authors develop a novel constraint-based framework translating inductive invariant verification for hybrid systems into an ∃∀ quantified formula over the reals, which is then transformed via a generalized Farkas lemma... Read more

articleView Paper downloadDownload

keyboard_arrow_downShow more

5. How can hybrid AI models combining neural networks, genetic algorithms, and fuzzy logic be modularly designed and configured to tackle complex computational problems?

This theme explores the development of unified software platforms that facilitate the construction and experimentation with hybrid AI models by modularly combining diverse AI algorithms. The goal is to harness complementary capabilities of neural networks, evolutionary optimization, and fuzzy logic within an extensible and interchangeable block-based software tool, promoting rapid prototyping and improved problem-solving performance across domains.

Creating Hybrid AI Models with Bang

by Roman Neruda

2025

Key finding: This paper introduces Bang, a modular and extensible software framework that encapsulates AI methods such as neural networks, genetic algorithms, and fuzzy logic into interchangeable building blocks connected via data... Read more

articleView Paper downloadDownload

keyboard_arrow_downShow more

6. What frameworks and conceptualizations enable hybrid simulation and modeling to serve as transdisciplinary research enablers across diverse disciplines?

This research direction addresses the integration of multiple modeling and simulation (M&S) methods and cross-disciplinary knowledge to develop hybrid models that span disciplinary boundaries. The focus is on establishing conceptual frameworks to facilitate synthesis of diverse theories, techniques, and domain data into coherent transdisciplinary models supporting complex decision-making and system analysis.

Hybrid models as transdisciplinary research enablers

by Andreas Tolk and

2021, European Journal of Operations research

Key finding: The authors propose a conceptual framework for hybrid modeling that emphasizes cross-disciplinary research collaboration and transdisciplinary alignment of M&S methods with domain theories and data. Using examples from... Read more

articleView Paper downloadDownload

keyboard_arrow_downShow more

7. How can union models efficiently represent and support reasoning over large model families with variation across spatial and temporal dimensions?

This area investigates formalizing large families of related models (arising from versions, configurations, or uncertainties) into a single union model that retains exact membership information and compactly encodes commonalities. Such union models enhance efficiency and scalability in analysis and reasoning tasks over entire model families in domains including regulatory compliance and product line engineering.

Union Models for Model Families: Efficient Reasoning over Space and Time

by Sanaa Alwidian

2023, Algorithms

Key finding: The paper formalizes union models as the unified attributed typed graph encompassing all members of a model family, annotated for individual versions and configurations. The approach enables compact representation and... Read more

articleView Paper downloadDownload

keyboard_arrow_downShow more

8. What are the benefits and methodological requirements of adopting model pluralism in scientific practice and philosophy?

Model pluralism recognizes the necessity of utilizing multiple diverse models to capture various aspects, scales, and purposes within scientific investigations. This perspective challenges monistic views focusing on single models and calls for appreciation of sets of models with differing functions, promoting pragmatic strategies for explanation, prediction, and intervention in complex phenomena.

Model Pluralism

by Walter Veit

2019, Philosophy of Social Science

Key finding: This paper philosophically argues that scientific modeling fundamentally involves plural sets of models with diverse epistemic roles, reflecting the complexity of natural phenomena. Adopting model pluralism dissolves problems... Read more

articleView Paper downloadDownload

keyboard_arrow_downShow more

All papers in Hybrid Models

Social Entrepreneurship and The Hybrid Model - a triple bottom line approach

by Lars Føleide

As the economy continues to struggle, and solutions at either side of the spectrum appears to have their respective challenges, attention are now being directed towards the promise of The Hybrid Model. This study examines through an... more

Figure 1: Index of Health and Social Problems in relation to income inequality in rich countries (Wilkinson and Pickett, 2009).

descriptionView Paper arrow_downwardDownload

A hybrid model coupled with singular spectrum analysis for daily rainfall prediction

by K.W. Chau

2010, A hybrid model coupled with singular spectrum analysis for daily rainfall prediction

"A hybrid model integrating artificial neural networks and support vector regression was developed for daily rainfall prediction. In the modeling process, singular spectrum analysis was first adopted to decompose the raw rainfall data.... more

descriptionView Paper arrow_downwardDownload

Review of Soft Computing Models in Design and Control of Rotating Electrical Machines

by K.W. Chau and

2019, Energies

Rotating electrical machines are electromechanical energy converters with a fundamental impact on the production and conversion of energy. Novelty and advancement in the control and high-performance design of these machines are of... more

Figure 1. Taxonomy of soft computing methods used in making the initial database [47,48]. or “tabu search” or “cuckoo search” or “simulated annealing” or “Bayesian network” or “Markov logic network” or “rough set”.

Step.5 reclassifying the papers into individual tables according to the popular SC models used

Figure 3. The number of published articles on rotating electrical machines from 1963-2017: 483,320 document results. (Source: Scopus & Web of Science). Exploring the literature on rotating electrical machines shows continued progress on the design and advancement of rotating electrical machines. Almost half a million documents in this particular realm show the importance of this topic and the dependent technologies. Figure 3 shows the progress in some literature. It is also apparent that the use of SC models has been started from the late 80s and early 90s and boosted the design advancement.

Figure 4. Analysis of the literature on soft computing in design, control, and development of rotating electrical machines from 1987-2017. Total of 30,382 document results. (Source: Scopus & Web of Science).

Figure 5. Analysis of the literature by based on countries (Source: Scopus & Web of Science). Analysis of the literature based on countries showed that China, USA, India, Iran, and the UK are among the top five active regions on the advancement of rotating electrical machines using SC methods (see, Figure 6).

Figure 7. Literature types on the on rotating electrical machines using soft computing, from 1987-2017. (Source: Scopus & Web of Science).

Figure 8. The subject area of rotating electrical machines using soft computing, from 1987-2017. The math, computer science, and energy subject areas had a rise (Source: Scopus & Web of Science). Figure 8. The subject area of rotating electrical machines using soft computing, from 1987-2017.

Figure 9. Results of the study by Aguilar [49] in terms of (a) mean deceleration and (b) distance. In the paper by Amulya, in 2016 [50] a steam turbo-generator condition assessment problem was considered in which the condition is determined by fuzzy logic-based vibration analysis. Recommendations for improving the performance of the conventional pitch angle control strategy for power regulation in wind turbines are given in Ben Smida (2016) [51] in which the pitch angle is based on fuzzy logic. The study was developed using fuzzy logic and proportional integral derivative PID controllers. The comparison of the performance of the two systems was performed using mean absolute error (MAE) and mean percentage error (MPE). Such that, a fuzzy logic controller with low MAE (29.02%) and MPE (0.0088%) values provided a best-controlling ability compared that for PID controller which can make a smooth controlling process.

Figure 10. Comparing the time values for the study by Aruna (2016) [61]. An interesting application of neuro-fuzzy systems is described in Arumugom (2017) [60] where the authors attempt to design a DVR with energy conservation using a MIC and vibration energy harvester (VEH). Aruna [61] proposed in 2016 a hybrid method using ANFIS for the speed control of interior permanent magnet synchronous motor (IPMSM). Their results demonstrate that their solution enables better disturbance rejection compared to the classic PID controller. The settling, arising and overshooting times of the proposed controlling method has been compared with PID, MR-PID, and MWNN-particle swarm optimization (PSO) controlling algorithms. Figure 10 indicates the time values for each term. The related results indicate that the proposed controlling method is the best technique to eliminate the nonlinearity with high reliability, and sustainable performance compared with that for the other techniques.

A vast amount of literature on nonlinear control solutions have been publishing in recent years. or instance, Li et al. (2005) [65] propose a control technique using a robust model reference controller ith the combination of the hybrid fuzzy controller. The core of the control strategy is based on the assical Lyapunov technique which is a complicated technique, but the control signal includes fuzzy, assic and robust terms. The fuzzy logic controller was compared with the hybrid-fuzzy controller sing root mean squared error (RMSE) values in terms of output tracking for DC servo motor and verted pendulum. The results have been presented in Figure 11. Based on results, the hybrid yntrolling method had the best performance in both terms. In order to limit the current and voltage harmonics, and to improve total harmonic distortior and true power factor in Fatih Kececioglu, (2017) [66], a neuro-fuzzy controller is applied in a hybric passive filter for PWM rectifiers. Gnanaprakasam (2015) [67] proposed a hybrid approach for the detection and classification of the vibration signal of an induction motor in which the fault detectior process including the extraction of significant features of vibration signals is carried out by using the S-transformation algorithm. After, the ANFIS classification technique was employed to classify the signal into its faulty or the normal state. S-transform-ANFIS, S-transform- radial basis functior neural network (RBFNN), S-transform-FFNN, and DWT-RBFNN were compared in terms of accuracy sensitivity and specificity for IR faulty condition, centered or fault, opposite or fault, orthogonal o1 fault, and BB fault conditions. The general results have been presented in Figure 12. In all conditions S-transform-ANFIS has the highest accuracy, sensitivity, and specificity compared with those for the other methods.

In the paper by Hossain (2015) [68] three nonlinear control strategies are compared, namely the fuzzy logic controller, SNC, and ANFIS-based controller. Their results show that the proposed fuzzy logic controller-, SNC-, or ANFIS-based VR-FCL are effective in improving the transient stability of a large-scale hybrid power system consisting of a doubly-fed wind power generator (DFIG)-based wind farm, a PV plant, and SG.

Figure 13. Results for the study by Vanchinathan (2018) [78]. TUVeTly Ul DY UL tUtlalhls EICCU Id) MWLaACI Le Ustolott. Design issues of electrical machines are usually raising complicated or multi-objective, ill-definec optimization problems. The application of Evolutionary-based solution or stochastic optimizatior methods can handle these problems by reducing the computational costs and simultaneously providing near-optimal solutions. For instance, Mamede (2018) [76] presents the combination of genetic algorithm (GA) and differential evolution (DE) techniques that allows maximizing the average torque and torque density and minimizing copper loss of single-phase switched reluctance machine (SPSRM). S. B. Tamilselvi, 5S. (2018) [77] proposes an evolutionary optimization procedure for the design of two, namely the surface-mounted and interior mounted permanent-magnet motor (PMMs configurations using DE algorithm. Optimization tasks are the core of most control problems alsc The metaheuristic optimization methods are effective alternatives in such cases, especially in sensorles: drives. Vanchinathan (2018) [78] introduce the bat algorithm (BA) in the parameter tuning issue of < fractional—order PID controller for speed control of sensorless control of brushless DC electric motors Modified genetic algorithm (MGA), ABC, and BA methods were employed in FOPID controller ir three loads (0, 50 and 100% of full load) and the simulation was performed using MATLAB/Simulin| software. Based on results, the proposed BA method, have the lowest steady-state error (%), peal time (s), rise time (s) and settling time(s) compared with those for the other methods in each load separately (Figure 13). It seems that BA method has the best performance at 100% in case of settling rise and peak times but in case of steady-state error, the condition of 0% provides the best performance for BA technique. The detailed results have been presented in Figure 13. The finding of Virtic (2016) [79] also supports that the evolutionary optimization with the analytical evaluation of objective functions significantly shortens the computational time required for design optimization in comparison with the finite element methods (FEM).

Figure 15. Results for the study by Kari (2018) [93]. presented in Figure 15. As is clear, the proposed method has the highest average consistency, accuracy, and reliability.

Figure 16. Results for the study by Kazemdehdashti (2018) [94] in terms of (a) root mean squared error (RMSE) and (b) time. Kazemdehdashti (2018) [94] introduces a new density estimator for optimal power flow problem of networks containing renewable energy sources, etc., using generalized cross-entropy method. Figure 16 presents the results of the study in terms of RMSE (a) and time (b) for comparing the proposed method for 14-bus. As is it shown, the RMSE of the proposed method is lower than that for the other methods, but this has increased the processing time. The two-point estimation method (TPEM) has a lower processing time and lowers accuracy.

In general, it can be claimed that hybrid methods can be the most successful methods than single methods because these methods can take advantage of several methods simultaneously and reduce the defects in each method. With the advancement of artificial intelligence and soft computing methods, the trend towards hybrid methods increases for the reasons given. There are still many hybrid methods that have not been used, which suggests that exploration of these methods be increased which can clarify the dark points in this and increase the orientation and usage by the researchers.

Table 1. Simple fuzzy systems in rotating the electrical machine.

Table 2. Adaptive neuro-fuzzy systems in electrical machines.

Table 3. Hybrid techniques applied in electrical machine design and control.

Table 4. Artificial neural networks. Considering the ANNs-based solutions in the field of electrical machines the study of Bouchiba (2017) [72] provides a fresh example of the NN sliding mode controller for the control for multi-machine web winding systems. The advantage of this technique is that it can significantly reduce the chattering

Table 5. Evolutionary computing and metaheuristics. Optimization problems, especially ill-posed or multi-objective optimization tasks are common in control and design problems in most of the engineering problems, including electrical machine and drive systems. Most of the applications found in the latest paper employ the evolutionary computing

Table 6. Application of ant colony in electrical machines. The characteristics of some individuals are different, but the unity of the colony is very effective. Each ant moves in an isolated and random way, but recognizing its pathways marked with pheromone is also likely to follow them. Meanwhile, an increase in their pheromone concentration by pheromone emissions, thereby increases the concentration of the pheromone attractiveness. Thus, the frequencies of frequently used routes increase the pheromone’s level while being neglected paths. The pheromone emitted by the ants continuously vaporizes, that is, confirmation on the given route (releasing a new pheromone dose) without decreasing the pheromone level (until contains pheromone). If there are two paths leading to food, the shorter route is more likely to turn the ants (as they arrive sooner and later back). Thus, they often reinforce the pheromone level, thereby providing higher “attractiveness” for the route. However, this will attract more ants to the shorter journey. After a while, the shorter way will be taken by most ants, while the longer path for pheromone levels will decrease to a minimum level. The ant colony metaheuristic shares the search task between the ant agents. These agents have very basic subtleties and, to a certain extent, simulate the behavior of true ants. “Artificial ants” (agents) build on solutions based on appropriate problem-specific constructive heuristics. A pheromone matrix determines the order of the building blocks of the solution used to construct good solutions. The values stored in the pheromone matrix are taken into account by the other agents on a probability basis when constructing his solutions. The first ant colony algorithm was applied for the traveling salesman problem (TSP). Since then, it has been widespread in engineering practice. Table 6 includes application of ant colony in rotating electrical machines.

Table 7. Tabu and cuckoo search techniques. The gbest-guided cuckoo search algorithm with the feedback control strategy and constraint domination rule for the security and economic operation of the power system is discussed in Chen et al. (2017) [84]. Results of the proposed method were compared with the results of another method developed by researchers in this field. The comparison results indicate that the proposed approach could provide high-quality, feasible solutions for different problems. q [he tabu search proposed by Glover (1986) [82] is based entirely on the methods described above and their direct development. In the immediate environment of our current solution, we are inves we always look for t direction in which the search was going and in which d lists t gone inspired by the breed solutions and the se solution. The main d the most interesting tigating new, be hat are marked by taboo tabs. Limitations on taboo tter solutions. In the tabu search method we are always trying to keep moving, so he best solution in the environment, even if it is worse than the solution being tested. Since this method has a lot to explore, we have to keep in mind that we need to select the irection it was not yet. This so-called taboo abs helping to get through the roads already by, increasing the search effectiveness. The cuckoo method [83] is a similar optimization method ing behavior of cuckoo species. In t publications collected in Table 7. his approach, the eggs represent the possible arch is performed by the quality of the eggs, which leads to finding the best ifferences and advantages of the discussed approaches can be found in some of domination rule for the security and economic operation of the power system is discussed in

Table 8. Simulated annealing. SA is an efficient tool for various optimization problems. It is shown by Chiu et al. (2016) [88]; that SA can be applied in the optimization task of space-constrained base-vibration systems. Based on the results, in the case of increasing the base-vibrating frequency, the extracted electrical energy will increase. Consequently, based on the buckling and fatigue analysis, the employed approach for an optimal designed one-mass vibration-based electromagnetic energy harvester is quite efficient in maximizing the energy.

Table 9. Application of probabilistic reasoning in the field of electrical machines and power systems. The dissolved gas analysis (DGA) approach is the basic tool for fault detection of power transformers. Kari (2018) [93] introduces an ANFIS system for these purposes. The outputs of the ANFIS model are evaluated by the Dempster—Schafer method which is originally developed for medical reasoning problems. The results of consistency, accuracy, and reliability have been

Table 10. Application of Bayesian methods in the field of electrical machines and power systems. Mansouri et al. (2015) [97] applies the Bayes method in a nonlinear time-varying state and parameter estimation task of induction machines (Ims). Based on results, the Bayes method has higher accuracy. Based on the results of the second comparative study, for all methods, estimation accuracy depends on estimating the model parameters as well as the convergence of the estimated parameters and states.

Table 11. Application of hybrid methods in the field of electrical machines and power systems. Dai et al (2016) [98] propose a new hybrid quantum-behaved particle swarm optimization-based solution for improving the precision of parameter identification (five parameters including stator resistance, stator inductance, rotor resistance, rotor inductance, and mutual inductance of stator and rotor) of the DFIG. Results indicated that the proposed algorithm accelerated the computing process by reducing the processing time. For the optimization of direct-drive permanent magnet synchronous generators, McDonald, (2017) [99] used a hybrid genetic algorithm and pattern search process and has found that the surface-mounted permanent magnet generator produces the lower cost of energy.

descriptionView Paper arrow_downwardDownload

A hybrid model coupled with singular spectrum analysis for daily rainfall prediction

by Gurjeet Singh

2010, Journal of Hydroinformatics

descriptionView Paper arrow_downwardDownload

Propiedades residuales con modelos híbridos de ecuaciones de estado cúbicas y redes neuronales

by José Abraham Betancourt Moreno

Residual properties using hybrid models of cubic equations of state and neural networks This paper presents the prediction of residual thermodynamic properties of hybrid models using state equations and discusses the consistency of... more

descriptionView Paper arrow_downwardDownload

Grammar or lexicon. Or: Grammar and lexicon? Rule-based and usage-based approaches to phonological variation

by Frans Hinskens

Please cite this article in press as: Hinskens, F., et al., Grammar or lexicon. Or: Grammar and lexicon? Rule-based and usage-based approaches to phonological variation. Lingua (2014), http://dx.

descriptionView Paper arrow_downwardDownload

Prediction of Hydropower Generation Using Grey Wolf Optimization Adaptive Neuro-Fuzzy Inference System

by K.W. Chau and

2019, Energies

Hydropower is among the cleanest sources of energy. However, the rate of hydropower generation is profoundly affected by the inflow to the dam reservoirs. In this study, the Grey wolf optimization (GWO) method coupled with an adaptive... more

descriptionView Paper arrow_downwardDownload

A hybrid System Dynamics – Agent Based model to simulate Complex Adaptive Systems: a new methodological framework for sustainability analysis

by Aled Jones and

Sustainability analysis represents a form of Complex Adaptive Systems (CAS) because it involves multiple sectors and agents displaying non-linear and non-rational interacting behaviours characterized by feedbacks and time lags. Thus, it... more

descriptionView Paper arrow_downwardDownload

A HYBRID INTELLIGENT MODEL FOR SOFTWARE COST ESTIMATION

by Ali Bou Nassif

2013

Accurate software development effort estimation is critical to the success of software projects. Although many techniques and algorithmic models have been developed and implemented by practitioners, accurate software development effort... more

descriptionView Paper arrow_downwardDownload

A survey of the use of cellular automata and cellular automata-like models for simulating a population of biological cells

by Nat Gab

The term Markov chain is sometimes used to refer to a discrete time Markov process.

a stable fixed point and p ~ .37 is instable. ‘able 2.2 shows the plot of equation 2.2 in the p, p(t +1) plane. It can be seen that p

Figure 3.1 Time progression of a CPM-like CA simulation

descriptionView Paper arrow_downwardDownload

Interactive simulation of object-oriented hybrid models, by combined use of Ejs, Matlab/Simulink and Modelica/Dymola

by Jose Sanchez

2004

Easy Java Simulations (Ejs) is a freeware, open source, Java-based tool intended to create interactive dynamic simulations. The use of Ejs, together with Matlab/Simulink and Modelica/Dymola allow us to combine the best features of each... more

descriptionView Paper arrow_downwardDownload

Implementation and Validation of a Surface Tension Model for the Multi-Scale Approach GENTOP

by Emilio Baglietto and

Multiphase flows encountered in the nuclear industry are largely of a complex nature, and knowledge of the accurate distribution of the void fraction is of utmost importance for operation of the reactor under steady, transient, and... more

In principle, damping should be larger in a coarse grid due to the discretization error in the convective part which acts as a diffusive term (numerical viscosity) in the momentum equation. As it was observed by Strublj in their two-dimensional cases, also here we observed a slightly changing with grid refinement which remains in the range of +5 %. As it can be observed from refining the mesh (128 x 128) in the cubic case, after 10 s of calculation the ethanol droplet is still oscillating. In this case, the oscillating period calculated from the simulation is around 0.536 s, which differs only in 0.19 % from the analytical solution (Fig. 2). This shown an improvement on the results of around 1.49 % by refining the mesh by a factor of 4.

Figure 3. Oscillating frequency for an initially elliptical ethanol droplet in air (32 x 32) The ellipsoidal test case is naturally more challenging. When using the coarser mesh, the oscillating period obtain from the calculation is 1.12 s while the analytical solution is around 1.55 s which give a difference between the two results of 23 % (Fig. 3). Further, it appears as if the droplet slightly shifted position in the axial direction, which could be occurring due to parasitic air velocities affecting the droplet interface [13]. When the mesh is refined to 128 x 128, the droplet is no longer displaced. As with the refinement in the cubic case, the damping decreases obtaining a non-trivial oscillation even after 10 s of calculation. In this case, the oscillating period calculated from the simulation is around 1.25 s, which differs in 19.3% from the analytical solution (Fig. 4). This shown an improvement on the results of around 4.8% by the refinement of the mesh, which again goes accordingly to the previously discussed observations from Strublj.

Figure 4. Oscillating frequency for an initially elliptical ethanol droplet in air (128 x 128)

Figure 5. Initial state for calculations of contact angle for a residing bubble the outside of the free surface, until the residing angle of 30 degrees has been achieved. At this point the force changes direction and acts to maintain the pre-defined contact angle (Fig. 6).

Figure 6. Final state for calculations for a residing bubble with a contact angle of 30°

Figure 7. Final state for calculations for a residing bubble with a contact angle of 90° In the second case, a residing angle of 90 degrees was established. In a similar manner as in the previou test, the contact angle was successfully achieved within the first 0.1 s and maintained during the rest of th 0.6 s of calculation (Fig. 7). Finally, an extreme case with a contact angle of 179 degrees was tested to chec the implementation limits, obtaining a final residing angle of around 168 degrees (Fig. 8).

3.3. Rising Bubble Morphology and Velocity Validation Tomiyama [15] has rigorously quantified the shapes and velocities of single bubbles rising through a vertical pipe. The experiments were performed with air-water and a 24.8 mm pipe length. The ratio, A, of equivalent bubble diameter (dg) to the diameter of the tube (Dr) was varied from 0.2 to 2.0 in these experiments. In this work, the test case for A = 0.6 for stagnant and pipe flow have been simulated and are discussed. When bubbles have a ratio A greater than 0.6, they are usually classified as Taylor bubbles and their terminal velocity is no longer a function of A, but of Dr. In these cases, the terminal velocity can be expressed as in Eq. 18, where Fr represents the Froude number which is a function of the Eé6tvos and Morton numbers.

Figure 9. Morphology of a single rising bubble of 1 = 0.6 without surface tension The first test was performed on a representative mesh, using 24 cells in the radial direction, and a time step of le-4 s. Since stagnant flow was chosen, it assumes laminar flow conditions. A constant drag coefficient of 0.44 was set for this calculation. In order to separate the effects of all other forces in GENTOP and study only the effect of the surface tension implementation, the inhomogeneous “free surface” model inherit from CFX with aggressive interface compression scheme was used. For this first case, the surface tension has been deactivated. As the bubble rise in the channel, the shape and interface is highly disturbed, and the gas structure is in constant oscillation. This is consistent with the behavior observed in the GENTOP validation on the MT-Loop pipe flow cases [1]. The terminal velocity was found to be 0.219 m/s, deviating from the theoretical solution by only 1.38 %. In terms of the morphology of the bubble, Fig. 9, shown a very deformed structure which is certainly not expected for this case. A comparison using the Tomiyama analytical model in this case, would yield unphysical results.

Figure 11. Morphology study of a single rising bubble of 1 = 0.6 with the surface tension implementation and V_ = 1.0 m/s

free surface region on GENTOP instead of the full AIAD model could be sufficient for the proper prediction of velocity distribution for the resolved gas in vertical flow conditions. Figure 12. Morphology study of a single rising bubble of 1 = 0.6 with the surface tension implementation using the full GENTOP concept

descriptionView Paper arrow_downwardDownload

Developing an intelligent expert system for streamflow prediction, integrated in a dynamic decision support system for managing multiple reservoirs: A case study

by Hamed Zamani Sabzi (Zamanisabzi)

Since fresh water is limited while agricultural and human water demands are continuously increasing, optimal prediction and management of streamflows as a source of fresh water is crucially important. This study investigates and demonstrates how data preprocessing and data mining techniques would improve the accuracy of streamflow predictive models. Based on easily accessible Snow Telemetry data (SNOTEL), four streamflow prediction models – autoregressive integrated moving average (ARIMA), artificial neural networks (ANNs), a hybrid-model of ANN and ARIMA (ANN-ARIMA), and an adaptive neuro fuzzy inference system (ANFIS) – were developed and utilized in a streamflow prediction process on Elephant Butte Reservoir. Utilizing the statistical correlation analysis and the extracting importance degrees of predic-tors led to efficiently select the most effective predictors for daily and monthly streamflow to Elephant Butte Reservoir. For the daily prediction time step, by preprocessing the historical data and extracting and utilizing the extracted climate variability indices through data mining techniques, the ANFIS model achieved a superior streamflow prediction performance for Elephant Butte Reservoir compared to the other three evaluated prediction models. Additionally, for predicting monthly streamflow to the Elephant Butte Reservoir, ANFIS showed significantly higher accuracy than the ANNs. As an optimal application of the developed predictive expert systems, successful integrating the prediction models in integrated reservoir operations balanced the need for a reliable supply of irrigation water against losses through evaporation. The optimal operation plan significantly minimizes the total evaporation loss from both reservoirs by providing the optimal storage levels in both reservoirs. This study provides the conceptual procedures of non-seasonal (ARIMA) model, and since the model is univariate, it demonstrates a strongly-reliable inflow prediction when existing information is limited to streamflow data as a predictor.

descriptionView Paper arrow_downwardDownload

Prediction of petroleum reservoir properties using different versions of adaptive neuro-fuzzy inference system hybrid models

by J Labadin

Inference System (ANFIS) hybrid model and two innovative hybrid models in the prediction of oil and gas reservoir properties. ANFIS is a hybrid learning algorithm that combines the rule-based inferencing of fuzzy logic and the... more

ANFIS, proposed by [23], is a result of an intelligent combination of the learning capabilities of Artificial Neural Networks (ANN) and the reasoning capabilities of Fuzzy Logic as a hybrid intelligent system. It is a network structure that implements Fuzzy Inference System (FIS), a knowledge representation where each fuzzy rule describes a local behavior of the system, and employs hybrid learning rules for training. It is a class of adaptive networks which enjoys many of the advantages claimed by ANN and the linguistic interpretability of FIS while combining the gradient descent and the least-squares method technologies. The neuro-fuzzy model of ANFIS is a multilayer neural network-based fuzzy system. Its basic architecture is shown in Figure 1 with the system having a total of five layers [24]. The main objective of ANFIS is to integrate the best features of fuzzy systems, through the representation of prior knowledge into a set of constraints (network topology) to reduce the optimization search space; and ANN through the

Fatai A. Anifowose is a PhD student in the Faculty of Computer Science and Information Technology, Universiti Malaysia Sarawak. Presently, he works as a Research Engineer in the Center for Petroleum and Minerals, Research Institute, King Fahd University of Petroleum and Minerals Abdulazeez Abdulraheem is an Associate Professor in the department of Petroleum Engineering, King Fahd University of Petroleum and Minerals, Saudi Arabia. He earned a Ph.D. degree in Civil Engineering with specialization in Rock Mechanics from the University of Oklahoma, Norman, USA in 1994, a Master degree in Civil Engineering from the Indian Institute of Science, Bangalore, India, in 1985, and a Bachelors degree in Civil Engineering from the Osmania University, Hyderabad, India, in 1983. His areas of research interests include theoretical and experimental rock mechanics, constitutive (material behavior) modeling, and numerical simulation using finite element and finite difference, soil mechanics, and structural dynamics. Author Biographies

Figure 2. Conceptual Framework of the FN-T2FL Hybrid Model [22]

Figure 5. Comparison of Input MFs with Constant Output MF. Figure 4. Input MFs with Linear Output MF showing Over-fitting. Fi

Figure 7. Optimal Radii for ANFIS-SC with Site 1 Well 2 Dataset Figure 6. Optimal Radii for ANFIS-SC with Site 1 Well 1 Dataset

Figure 8. Optimal Radii for ANFIS-SC with Site 1 Well 3 Dataset

Figure 9. Optimal Radii for ANFIS-SC with Site 2 Well 1 Dataset

Figure 10. Optimal Radii for ANFIS-SC with Site 2 Well 2 Dataset

Figure 12. Optimal Clusters for ANFIS-FCM with Site 1 Well 1 Dataset

Figure 11. Optimal Radii for ANFIS-SC with Site 2 Well 3 Dataset

Figure 13. Optimal Clusters for ANFIS-FCM with Site 1 Well 2 Dataset

Figure 14. Optimal Clusters for ANFIS-FCM with Site 1 Well 3 Dataset

Figure 15. Optimal Clusters for ANFIS-FCM with Site 2 Well 1 Dataset

Figure 16. Optimal Clusters for ANFIS-FCM with Site 2 Well 2 Dataset

Figure 17. Optimal Clusters for ANFIS-FCM with Site 2 Well 3 Dataset

Figure 18. Testing CC Comparison for Porosity Datasets Figure 19. Testing RMSE Comparison for Porosity Datasets

Figure 20. Training and Testing Run Time Comparison for Porosity Datasets

Figure 22. Testing RMSE Comparison for Permeability Datasets Figure 23. Training and Testing Run Time Comparison for Permeability Datasets Figure 21. Testing CC Comparison for Permeability Datasets

Table 1. Optimal Number of Clusters for Datasets.

descriptionView Paper arrow_downwardDownload

INTERACTIVE SIMULATION OF OBJECT-ORIENTED HYBRID MODELS, BY COMBINED USE OF EJS, MATLAB/SIMULINK AND MODELICA/DYMOLA

by Manuel Berenguel and

2004, European Simulation Multiconference

Figure 1: Model of a Process The model shown in Fig. | will be used to illustrate the discussion. The voltage applied to the pump (¥v) is an input variable. The cross-sections of the tank (A) and the outlet hole (a), the pump parameter (k) and the gravitational acceleration (g) are time-independent properties of the physical system.

Iwo input variables to the DymolabBlock block are used to model the interactive changes in the state: [state[:] and CKstate (see Fig. 2). The array [state contains the values used to reinitialize the model state. The signal CKstate is used to trigger the state re-initialization event. The state re-initialization is performed using a built-in Modelica operator: reinit(x,expr). It re- initializes an state variable (x) with the value obtained of evaluating an expression (expr), at an event instant. The output-variable array of the DymolaBlock block (O[:]) contains the variables representing the actual value of the state, in addition to the other variables linked to the view elements. Ejs uses the value of this output array (O[:]) to refresh the simulation view. When the user changes the value of one or several state variables, Ejs writes the new state in [state array and changes the value of CKstate (from one to zero or from zero to one) in order to trigger the re-initialization event.

Figure 3: Schematic Description of the Proposed Modeling Methodology for Interactive Simulation

In order to reformulate the model as shown in Equation (1), with x, ={n}, a dummy variable Q and an equation (h =) are included in the model (H. Elmqvist: personal communication, 1995). These changes do not modify the model: the equation h = is used to calculate @ . The index of the extended model is greater than one (Brenan et al. 1996). The model index can be reduced to one or zero by applying the Pantelides algorithm (Pantelides 1998) and by including the differentiated equations in the model (Mattsson and Soderlind 1992). Then, the dummy derivatives are defined so that the user’s state selection is fulfilled. The result of these manipulations is shown below, where the variable to evaluate from each equation is written within square brackets.

Figure 4: Simulink Model of the Quadruple-Tank Process and its Control System Figure 5: View of the Quadruple-Tank System

descriptionView Paper arrow_downwardDownload

Hybrid convolution model and its application in predictive pH control

by Janos Abonyi and

1999, Computers & Chemical Engineering, vol. 23, pp. S227–S230, DOI: 10.1016/S0098-1354(99)80056-8

This paper presents a new method for synthesising chemical process models that combines prior knowledge and fuzzy models. The hybrid convolution model consists of a fuzzy model based steady-state, and an impulse response model based... more

descriptionView Paper arrow_downwardDownload

Prediction of daily water level using new hybridized GS-GMDH and ANFIS-FCM models

by K.W. Chau

2021, Engineering Applications of Computational Fluid Mechanics

Accurate prediction of water level (WL) is essential for the optimal management of different water resource projects. The development of a reliable model for WL prediction remains a challenging task in water resources management. In this... more

descriptionView Paper arrow_downwardDownload

The Use of a Hybrid Model in Web-Based Education: The Global Campus Project

by Natalia Kozlova

2003, Web-Based Education: Learning from Experience

Education is one of the key sectors that benefited from the continuous developments and innovations in information and communication technology. The changes have affected the concepts of teaching, the methodologies used in class and... more

descriptionView Paper arrow_downwardDownload

Physical, Data-Driven, and Hybrid Approaches to Model Engine Exhaust Gas Temperatures in Operational Conditions

by Andrea Coraddu and

2021, Ships and Offshore Structures

Fast diesel engine models for real time prediction in dynamic conditions are required to predict engine performance parameters, to identify emerging failures early on, and to establish trends in performance reduction. In order to address... more

Figure 11: Scatter plot (measured vs predicted) and trend in time for 7“ using a DDM with A = 0.

Figure 1: Holland Class Oceangoing Patrol Vessels.

Figure 2: Propulsion System Layout for the Holland Class Oceangoing Patrol Vessels.

where J is the usual regularization of KRLS and @ € [0, 09), instead, is another hyperparameter that forces the shared model to be close to the task specific mod- els. Basically the MTL problem of Eq. (20) is a concatenation of three learning problems solved with KRLS plus a term which tries to keep a relation between all the three different problems. Then, it is possible to state the MTL version of Eq. (12), as follows

where p = [p1,..., Pri’. The solution of this problem is again equivalent to solv- ing a linear system By exploiting the kernel trick as in KRLS, it is possible to reformulate Prob- lem (20), as follows

Figure 5: How to take into account the dynamics in DDMs and HMs.

(e) Mean exhaust gas temperature after tur- (f) Mean exhaust gas temperature before tur- bine. bine.

Figure 7: PM steady state operating conditions.

Figure 8: PM Steady-State operating condition.

(a) Mean exhaust gas temperature after tur- (b) Mean exhaust gas temperature before tur- bine. bine.

Figure 10: Scatter plot (measured vs predicted) and trend in time for 7 using a PM with A = 0.

Figure 12: Scatter plot (measured vs predicted) and trend in time for 7“ using an HM with A = 0

Figure 13: Scatter plot (measured vs predicted) and trend in time for T@ using a DDM with A = 20 which is the best one as shown in Table 9.

Figure 14: Scatter plot (measured vs predicted) and trend in time for 74 using an HM with A = 20 which is the best one as shown in Table 9.

Table 8: Indexes of performances (MAE, MAPE, and PPMCC) of the different models (PMs, DDMs and HMs) for different A € {0, 10, 20,30} for 7“. Note that A = 0 means that the authors dc not exploit time series information from the past, for A > O there is no PM result as described ir Section 4. Figure 9: PM Dynamic operating condition.

Table 9: Indexes of performances (MAE, MAPE, and PPMCC) of the different models (PMs, DDMs, and HMs) for different A € {10, 20, 30} for T?. Note that A = 0 means that the authors do not exploit time series information from the past, for A > 0 there is no PM result as described in Section 4. Table 10: Top 10 feature in X, ranked in descending importance, of the different models (DDMs and HMs) for different A € {0, 10, 20, 30} for T4. Note that, for A > 0, the importance does not refer to the single feature in X but the entire past temporal series.

descriptionView Paper arrow_downwardDownload

PI-LSTM: Physics-Infused Long Short-Term Memory Network

by Shubhendu Kumar Singh and

2019, 2019 18th IEEE International Conference On Machine Learning And Applications (ICMLA)

We introduce a novel machine learning based fusion model, termed as PI-LSTM (Physics-Infused Long Short-Term Memory Networks) that integrates first principle Physics-Based Models and Long Short-Term Memory (LSTM) network. Our architecture... more

Fusion of two levels of physics, with our data-driven LSTM network, yields two types of hybrid models namely Hybrid 1 and Hybrid 2. In a later section of this paper it is shown that the quality of physics model or its abstraction ability in its closeness to the real system behavior, will ultimately help in improving the prediction accuracy of the overall hybrid model. The hybrid model aims to determine the state of the IPS with respect to different spatial and temporal coordinates. The state vector includes X, V,0,w, which are the position, the velocity of the cart, the angle and the angular velocity of the pendulum, respectively. The state space equations used for both the partial and the complete physics models are: where M., Mp, l, g are the mass of cart, the mass of the pole, the length of the pole and the gravitational acceleration,

The inverted-pendulum system (IPS) is one of the most popular dynamical systems examples of an unstable system being used in machine learning domain (e.g., reinforcement learning) or controls. In IPS problem, the angle and angular velocity of a pendulum are controlled by moving the cart Fig. 2. The position of the cart and the force applied to it are bounded. Fig. 2. Inverted Pendulum Problem Two types of modified IPS partial physics models are considered while formulating the problem statement. In both models, average wind speed is considered, which differs from the real wind speed. The only difference between these two models is the coefficient of friction involved.

Our physics model traces its roots to a model proposed by [20]. In this model, a non-linear ODE is used to model tumor growth and stability criterion for the system. This model is generalized by considering 12 states, two of the states being the number of stem and fully-differentiated cells. Semi- differentiated cells can have 10 divisions before acquiring ully-differentiated state. The rest of the ten states are charac- terized by the number of cells corresponding to each of these 0 divisible cellular conditions. These states are governed by (17) to (21). The cellular automata model classifies the cells into three categories: Stem cells, semi-differentiated cells, and fully differentiated cells. Their cellular differentiation mechanism is depicted in Fig. 3. Only the stem cell has the ability to divide infinitely while the other two types have limited divisibility. Stem cells can divide symmetrically to produce new stem cells known as daughter cells, and hence can increase the tumor size. Also, through asymmetric division, they can produce differentiated cells, which can only divide to a certain extent. Semi-differentiated cells can divide up to some extent, while the fully differentiated ones cannot divide at all. Therefore the probability to divide for each cell, pg, and the probability of having symmetric or asymmetric division, pas are considered as parameters for the model. Differentiated cells die due to factors such as lack of glucose or attack from immune cells. Thus, in the cellular automata model, the rate of cell death a is one of the important parameters. In this model, the overall running time is also one of the input parameters.

Fig. 4. Violin plot for generalizability, extrapolability, and robustness of inverted pendulum

Fig. 5. Inverted Pendulum-Hyperparameter Tuning of Hybrid Model

Fig. 7. Inverted Pendulum-Regression plot of Data-driven Model

Fig. 6. Inverted Pendulum-Regression plot of Hybrid1 Model

Fig. 8. Inverted Pendulum-Performance comparison over test set

Fig. 11. Violin plot for sensitivity to the number of sample of cancer cell growth model

Fig. 9. Violin plot for sensitivity to the number of sample of inverted pendulum

Fig. 10. Violin plot for generalizability, extrapolability, and robustness of cancer cell growth model

INVERTED PENDULUM PROBLEM: COMPLETE (REAL) AND PARTIAL PHYSICS MODELS Cancer cells exhibit uncontrolled growth and no longer respond to neural signals that control cell growth inside the body. Cancer cells originate within the tissues and keep on dividing. In the later stages, these cells metastasize, meaning they will break through tissues and spread to nearby tissues and other organs. Our objective is to model the growth of cancer cells, i.e., to estimate the size of the tumor based on its initial size and growth rate of cells. TABLE I

descriptionView Paper arrow_downwardDownload

Prediction of Hydropower Generation Using Grey Wolf Optimization Adaptive Neuro-Fuzzy Inference System

by farhad hooshyaripor

Energies

descriptionView Paper arrow_downwardDownload

PI-LSTM: Physics-Infused Long Short-Term Memory Network

by Amir Behjat

2019, 2019 18th IEEE International Conference On Machine Learning And Applications (ICMLA)

descriptionView Paper arrow_downwardDownload

A Hybrid Modeling and Simulation Methodology for Dynamic Physical Systems

by José Sepúlveda

2002, SIMULATION

This article presents the design and implementation of a robust simulation methodology for hybrid models of physical systems that encompasses behavior patterns that undergo discrete transitions between modes of continuous behavior... more

descriptionView Paper arrow_downwardDownload

The Market for Picasso Prints: An Hybrid Model Approach

by marilena loatelli

2015

ABSTRACT: Two basic approaches have been used by the literature focusing on the return to holding artistic works: the hedonic price model and the repeat-sales model. This paper provides a procedure for jointly estimating the two models in... more

descriptionView Paper arrow_downwardDownload

A Hybrid Intelligent Model for Software Cost Estimation

by Danny Ho

Journal of Computer Science

descriptionView Paper arrow_downwardDownload

Hybrid Models

Key research themes

1. How can hybrid modeling integrate top-down and bottom-up approaches to better represent technological and economic interactions in energy systems?

2. What methodologies enable effective hybrid modeling and prediction of dynamical systems by blending mechanistic and data-driven representations?

3. How can component-oriented hybrid dynamic system modeling using meta-programmable environments enhance compositionality and simulation of complex physical systems?

4. What formal constraint-based approaches facilitate verification of hybrid systems by discovering inductive invariants expressed as complex polynomial inequalities?

5. How can hybrid AI models combining neural networks, genetic algorithms, and fuzzy logic be modularly designed and configured to tackle complex computational problems?

6. What frameworks and conceptualizations enable hybrid simulation and modeling to serve as transdisciplinary research enablers across diverse disciplines?

7. How can union models efficiently represent and support reasoning over large model families with variation across spatial and temporal dimensions?

8. What are the benefits and methodological requirements of adopting model pluralism in scientific practice and philosophy?

Related Topics

All papers in Hybrid Models