Physical Modeling Approach Research Articles

Creep curve modelling of Austenitic Steel 316LN

Physical based creep models that elucidate the creep deformation behaviour with ongoing microstructural evolution can be a useful tool for the components life assessment as well as the design of improved materials, deployed at high temperature and pressure. In this research work, a creep model that is a combination of physical based model and CDM approach is employed to predict the creep curves of steel 316LN. The microstructure based variables those are different dislocation densities (mobile and forest) are the input parameters. The model provides a provision for the assessment of each microstructural variable, in each time step of the creep deformation. Consequently, other than creep curves, the model also demonstrates the evolution of dislocation density (mobile and forest), dislocation velocity, dislocation mobility and mean free path. Initial values of input parameters (various dislocation densities and mean free path) are obtained from the literature for initializing the model. It can be observed that the predicted creep curves are in reasonable agreement with experimental ones and the evolution of other involved parameters is discussed thoroughly.

A high-resolution physical modelling approach to assess runoff and soil erosion in vineyards under different soil managements

Steep-slope viticulture is a common practice in the Mediterranean basin, and provides landscapes of considerable socio-economic value. However, these complex agricultural systems are intrinsically fragile. One of the main problems is soil erosion due to extreme rainfall events. This may cause a progressive reduction in soil fertility and the occurrence of instabilities and land degradation phenomena. To worsen this condition there is the increasing mechanisation of agricultural management causing soil compaction, and the pressure of climate change, with an intensification of extreme weather events. In this context, vineyard soil management plays a key role, as it can accelerate or mitigate overland flow and soil erosion phenomena. There are various techniques for quantifying these processes, often based on field measurements through prolonged data collection using experimental plots. However, the advent of new technologies in remote sensing opens new frontiers in the acquisition of high-resolution spatial data. Indeed, they could be used in runoff/erosion simulation models and integrated with site-specific data.The aim of this paper is to assess runoff and soil erosion processes in vineyard under four different soil managements. Specifically, four practices were tested: (1) Reference (inter-row managed with standard farm grass cover; RF); (2) Continuous Tillage (bare soil obtained by continuous mechanical weeding using roto-tiller; CT); Nectariferous (a mix of herbaceous species capable of attracting insects favouring inter-row biodiversity; NF); (4) Single Tillage (inter-row weeding once a year using roto-tiller; ST). The research proposes a modelling approach using a physically-based model (SIMWE) using samples of runoff and sediment as an assessment, collected with a low-cost methodology. In particular, a cost-effective approach easily replicable in different contexts (such as developing countries) is sought. In addition to cultivation specifics, areas of soil compacted by the passage of agricultural vehicles were also analysed using the connectivity index. In general, results show an interesting capacity of ST in mitigating soil erosion, as well as for NF. Furthermore, the negative role of wheel tracks as preferential pathways for surface runoff and sediment is highlighted. Finally, the work shows that CT aggravated soil erosion as compared to RF.

Operational output-only data-driven control input estimation and low-order physical modeling for vehicle vibrations

Vehicle vibration is one of the most critical issues hindering the supply of attractive automobiles to customers. This study focuses on drivability related to longitudinal vehicle body vibration excited by driver input at tip-in acceleration. To understand and design the vibration, a physical model with necessary and sufficient fidelity for its reproduction is required. Moreover, estimation of the excitation input (control input) is also vital for this reproduction. However, achieving both the accurate measurement and modeling of the control input is difficult and labor intensive. Therefore, a novel low-order physical modeling approach with control input estimation using operational output-only measurement is proposed here. The proposed approach requires some assumptions, such as inertia information, but requires no input measurement, no preliminary testing, no limitation to a linear model, and no assumptions in estimating the control input. The basic concept is to regard the vehicle system as a relationship (called a vibration transfer structure) between rigid-body motions and restoring forces and moments (called transfer forces). The proposed approach is implemented in five steps: First, the operational output is measured during tip-in acceleration. Second, important motions are extracted from the output using tensor decomposition. Third, a simplified transfer structure is determined using the important motions. Fourth, the control input and the transfer forces are estimated as time-series data using the structure. Finally, by characterizing the transfer forces, a low-order physical dynamic model is obtained. The effectiveness of this approach is demonstrated using two sets of operational tip-in acceleration data.

Cyber–physical system modeling with Modelica using message passing communication

Modelica is an object-oriented modeling language whose design and features facilitate the description of cyber–physical systems (CPS). Message passing communication (MPC), seen as the transmission of impulses of information between model components, eases the description of the discrete-event parts of CPS models. However, Modelica does not currently supports MPC. Modelica supports an equation-based component connection rationale, where Modelica tools automatically transform component connections into model equations, following a physical modeling approach. The differences between MPC and Modelica connections are analyzed. A proposal for supporting MPC in Modelica is presented, inspired by the coupled PDEVS model communication approach. The presented MPC proposal is based on the definition of structures to manage messages, named buffers, interface ports and communication channels. Also, an implementation of the proposed MPC mechanism in the form of a new free Modelica library, named MSGLib, is presented. MSGLib includes functionality to manage and dynamically store messages, and describe component communications. Two examples, a pick and place system and a robotic arm, are presented to demonstrate the use of the library, and its combination with other Modelica models.

Topology optimization with a closed cavity exclusion constraint for additive manufacturing based on the fictitious physical model approach

This paper proposes a topology optimization method that considers the geometric constraint of no closed cavities to improve the effectiveness of additive manufacturing based on the fictitious physical model approach. First, the basic topology optimization concept and level set-based method are introduced. Next, the fictitious physical model for a geometric constraint in the topology optimization framework is discussed. Then, a model for the geometric constraint of no closed cavities for additive manufacturing is proposed. Numerical examples are provided to validate the proposed model. In addition, topology optimization considering the geometric constraint is formulated, and topology optimization algorithms are constructed using the finite element method. Finally, optimization examples are provided to validate the proposed topology optimization method.

Virtual manipulation of tail postures of a gliding barn owl (Tyto alba) demonstrates drag minimization when gliding.

Aerodynamic functions of the avian tail have been studied previously using observations of bird flight, physical models in wind tunnels, theoretical modelling and flow visualization. However, none of these approaches has provided rigorous, quantitative evidence concerning tail functions because (i) appropriate manipulation and controls cannot be achieved using live animals and (ii) the aerodynamic interplay between the wings and body challenges reductive theoretical or physical modelling approaches. Here, we have developed a comprehensive analytical drag model, calibrated by high-fidelity computational fluid dynamics (CFD), and used it to investigate the aerodynamic action of the tail by virtually manipulating its posture. The bird geometry used for CFD was reconstructed previously using stereo-photogrammetry of a freely gliding barn owl (Tyto alba) and we validated the CFD simulations against wake measurements. Using this CFD-calibrated drag model, we predicted the drag production for 16 gliding flights with a range of tail postures. These observed postures are set in the context of a wider parameter sweep of theoretical postures, where the tail spread and elevation angles were manipulated independently. The observed postures of our gliding bird corresponded to near minimal total drag.

Hydrodynamic performances of a wall type breakwater - a physical and numerical approach

In this paper, hydrodynamic characteristics of caisson type breakwater are investigated through physical model approach and a numerical model based on the Volume of Fluid (VOF) is validated. The investigations are carried out for varying wave characteristics and depth of water 0.50 m. In an experimental study, to understand the model and scale effects for the desired wave conditions is a critical task in all-time conditions. So, it is also necessary to develop an appropriate numerical model to understand the hydrodynamics of the selected test model. Using the Volume of Fluid (VOF) method and incompressible open channel fluid flow a 2D numerical wave flume is developed using ANSYS- Fluent platform. The wave boundary conditions are adopted by solving the Reynolds-Averaged Navier Stokes equations (RANS equation) and especially with k—ε model to examine the effects of turbulence on the numerical results. The wave forces, wave runup, and wave reflection characteristics on the test model are measured for different wave characteristics, and results obtained from the numerical investigations are comparable with the experimental results to evident the relevance of the developed numerical model.

Hybrid approach for energy consumption prediction: Coupling data-driven and physical approaches

In recent years, a large number of building energy consumption prediction models, with various intended uses, have been proposed. The majority of these models have either taken a data-driven or a physical modeling approach, with each approach having its own strengths and limitations. Towards leveraging the strengths and reducing the limitations of each approach for improved prediction performance, this paper presents a hybrid machine-learning approach for occupant-behavior-sensitive energy consumption prediction. The proposed approach is composed of three constituent models: (1) a machine-learning model that learns the impact of outdoor weather conditions from simulation-generated data, (2) a machine-learning model that learns the impact of occupant behavior from real data, and (3) an ensemble model that predicts cooling energy consumption based on the outputs of the first two models. The simulation-generated data were created through simulating a set of reference buildings in EnergyPlus. The real data were collected from an office building in Pennsylvania. The proposed hybrid model was validated on an unseen real dataset. It achieved 0.73 kWh RMSE and 9.07% CV in hourly cooling energy consumption prediction, which indicates that the proposed approach is promising.

Chromosome Folding Promotes Intrachromosomal Aberrations under Radiation- and Nuclease-Induced DNA Breakage.

The long-standing question in radiation and cancer biology is how principles of chromosome organization impact the formation of chromosomal aberrations (CAs). To address this issue, we developed a physical modeling approach and analyzed high-throughput genomic data from chromosome conformation capture (Hi-C) and translocation sequencing (HTGTS) methods. Combining modeling of chromosome structure and of chromosomal aberrations induced by ionizing radiation (IR) and nuclease we made predictions which quantitatively correlated with key experimental findings in mouse chromosomes: chromosome contact maps, high frequency of cis-translocation breakpoints far outside of the site of nuclease-induced DNA double-strand breaks (DSBs), the distinct shape of breakpoint distribution in chromosomes with different 3D organizations. These correlations support the heteropolymer globule principle of chromosome organization in G1-arrested pro-B mouse cells. The joint analysis of Hi-C, HTGTS and physical modeling data offers mechanistic insight into how chromosome structure heterogeneity, globular folding and lesion dynamics drive IR-recurrent CAs. The results provide the biophysical and computational basis for the analysis of chromosome aberration landscape under IR and nuclease-induced DSBs.

Numerical Simulation of Flow in Parshall Flume Using Selected Nonlinear Turbulence Models

This study uses a computational fluid dynamics (CFD) approach to simulate flows in Parshall flumes, which are used to measure flowrates in channels. The numerical results are compared with the experimental data, which show that choosing the right turbulence model, e.g., v2−f and LC, is the key element in accurately simulating Parshall flumes. The Standard Error of Estimate (SEE) values were very low, i.e., 0.76% and 1.00%, respectively, for the two models mentioned above. The Parshall flume used for this experiment is a good example of a hydraulic structure for which the design can be more improved by implementing a CFD approach compared with a laboratory (physical) modeling approach, which is often costly and time-consuming.

Characterizing Binding Interactions That Are Essential for Selective Transport through the Nuclear Pore Complex.

Specific macromolecules are rapidly transported across the nuclear envelope via the nuclear pore complex (NPC). The selective transport process is facilitated when nuclear transport receptors (NTRs) weakly and transiently bind to intrinsically disordered constituents of the NPC, FG Nups. These two types of proteins help maintain the selective NPC barrier. To interrogate their binding interactions in vitro, we deployed an NPC barrier mimic. We created the stationary phase by covalently attaching fragments of a yeast FG Nup called Nsp1 to glass coverslips. We used a tunable mobile phase containing NTR, nuclear transport factor 2 (NTF2). In the stationary phase, three main factors affected binding: the number of FG repeats, the charge of fragments, and the fragment density. We also identified three main factors affecting binding in the mobile phase: the avidity of the NTF2 variant for Nsp1, the presence of nonspecific proteins, and the presence of additional NTRs. We used both experimentally determined binding parameters and molecular dynamics simulations of Nsp1FG fragments to create an agent-based model. The results suggest that NTF2 binding is negatively cooperative and dependent on the density of Nsp1FG molecules. Our results demonstrate the strengths of combining experimental and physical modeling approaches to study NPC-mediated transport.

Wave Attenuation Analysis for Artificial Coral Reefs Using a Physical Modelling Approach

Kim, T.; Kwon, Y.; Hong, S.; Kim, J.; Kwon, S., and Lee, J., 2021. Wave attenuation analysis for artificial coral reefs using a physical modelling approach. In: Lee, J.L.; Suh, K.-S.; Lee, B.; Shin, S., and Lee, J. (eds.), Crisis and Integrated Management for Coastal and Marine Safety. Journal of Coastal Research, Special Issue No. 114, pp. 529–533. Coconut Creek (Florida), ISSN 0749-0208. Due to global warming and coastal development, the damages in coastal such as sea level rise and changes in coastline caused by scour and erosion are increasing, Various types of breakwaters are applied to prevent the coastal erosion, however, the problem such as scour around the structure is exacerbated, therefore a fundamental solution is needed. In this study, two-dimensional laboratory experiments were conducted in the wave flume to investigate the wave transmission phenomena of artificial coral reef. Experiments are conducted by analyzing wave transmission dependents on ratio of stem length to water depth(ls/h), wave height and wave period. Wave attenuation appeared to be most dependent on stem length to water depth(ls/h) and structure width. As incident wave height increased and the wave period decreased, the wave attenuation also increased, however, it was insignificant compared to the influence of the inundation rate. Wave attenuation increases dependents on incident wave height and period, however showed insignificant effect compared to submergence rate. In addition wave attenuation rate also affects wave energy reduction in order of ls/h > H0/h > h/L0. Based on results, we analyzed correlation between KC and CD of artificial coral reef to suggest the equation on empirical relationship. As a result, it is judged that estimating wave transmission rate and drag coefficient of artificial coral reef dependents on sectional specification can be utilized properly in working design and application.

Firn changes at Colle Gnifetti revealed with a high-resolution process-based physical model approach

Abstract. Our changing climate is expected to affect ice core records as cold firn progressively transitions to a temperate state. Thus, there is a need to improve our understanding and to further develop quantitative process modeling, to better predict cold firn evolution under a range of climate scenarios. Here we present the application of a distributed, fully coupled energy balance model, to simulate cold firn at the high-alpine glaciated saddle of Colle Gnifetti (Swiss–Italian Alps) over the period 2003–2018. We force the model with high-resolution, long-term, and extensively quality-checked meteorological data measured in the closest vicinity of the firn site, at the highest automatic weather station in Europe (Capanna Margherita, 4560 m a.s.l.). The model incorporates the spatial variability of snow accumulation rates and is calibrated using several partly unpublished high-altitude measurements from the Monte Rosa area. The simulation reveals a very good overall agreement in the comparison with a large archive of firn temperature profiles. Our results show that surface melt over the glaciated saddle is increasing by 3–4 mm w.e. yr−2 depending on the location (29 %–36 % in 16 years), although with large inter-annual variability. Analysis of modeled melt indicates the frequent occurrence of small melt events (<4 mm w.e.), which collectively represent a significant fraction of the melt totals. Modeled firn warming rates at 20 m depth are relatively uniform above 4450 m a.s.l. (0.4–0.5 ∘C per decade). They become highly variable at lower elevations, with a marked dependence on surface aspect and absolute values up to 2.5 times the local rate of atmospheric warming. Our distributed simulation contributes to the understanding of the thermal regime and evolution of a prominent site for alpine ice cores and may support the planning of future core drilling efforts. Moreover, thanks to an extensive archive of measurements available for comparison, we also highlight the possibilities of model improvement most relevant to the investigation of future scenarios, such as the fixed-depth parametrized routine of deep preferential percolation.

Construction and application of an adapted rock physics template (ARPT) for characterizing a deep and strongly cemented gas sand in the Nam Con Son Basin, Vietnam

The Hai Thach field is one of the biggest gas and condensate producing fields in Vietnam. Located in the Nam Con Son Basin, offshore southern Vietnam, the field consists of multiple, stacked reservoir units, for which petrophysical characterization is challenging and a site-adapted rock physics template (ARPT) was proposed to solve this task. As reported in many publications, rock physics templates (RPTs) have been successfully applied for the characterization of shallow unconsolidated and weakly cemented reservoirs with cementation of less than 5% like in the North Sea or some other sites. However, the construction of RPT is very site-dependent. In this study, an approach of rock physical modelling, named the adapted stiff sand model (ASSM), was proposed to construct a site-adapted RPT of Vp/Vs versus acoustic impedance (AIP) for characterizing the upper part of Middle Miocene Sand in the Nam Con Son Basin, which is a 3.5-km deep and strongly cemented gas sand with 10% up quartz-overgrowth and clayey cementation. A key point in this approach was the use of a modified Biot-Gassmann's equation and in situ well log data to calculate the dry moduli of reservoir rock frame. As a result, the elastic and petrophysical parameters of this upper MMS sand, named as MMF30, are well evaluated, i.e., AIp from 9000 to 11,000 m/s*g/cc, Vp/Vs from 1.65 to 1.8, porosity from 12 to 16%, and gas saturation from 50% up, which match well with the core data and were used further in a prestack elastic inversion of a seismic inline that could map well the multiple stacked gas zones of the study reservoir. The rock physical modelling approach in this study could help characterize successfully deep and highly-cemented clayey gas sands in the Nam Con Son Basin.

Fracture Propagation and Hydraulic Properties of a Coal Floor Subjected to Thick-Seam Longwalling above a Highly Confined Aquifer

The high-pressure and water-rich confined aquifer occurring in the Ordovician limestone sequence poses great threats to the routine production of underground longwall mining. Considering the intense cooperation of mining disturbance and water pressure, water-conducting fractures within a coal seam floor can connect the lower aquifer and upper goaf, and this hydraulic behavior is considered the root of water inrush hazard and water loss or contamination. In this paper, the panel 4301 of the Longquan coal mine serves as the case where the panel works closely above the floor with high water pressure. By the combination of physical and numerical modelling approaches, the variation characteristics of fracture development and volumetric strain of floor rocks subjected to mining disturbance are analyzed. A numerical computation model is constructed based on the volumetric strain-permeability equation obtained by curve fitting, and on such basis, the impacts of different mining parameters on floor rock permeability are studied. The results show that the floor rocks experience fracture generation, extension, and convergence procedures as the workface advances along the longitudinal direction, and fractures appearing in front of the workface are more developed. In the whole process of coal seam extraction, the volumetric strain profile exhibits “Λ” shape and an inverted saddle shape before and after overburden strata collapse. By controlling a single variable, the paper reveals that panel height is of greater impact on floor permeability changes than panel length and panel width.

Experimental Investigation of Droplets/wall Interaction Relevant to Entrained-flow Slagging Gasifiers

The phenomenology of droplets/wall interaction relevant to entrained-flow gasifiers has been investigated. The present study reports an experimental characterization of the sticking/spreading/rebound patterns during impact experiments of molten particles. A physical modelling approach was followed, by simulating the slag particles with synthetic wax particles impacted against a target at near-ambient conditions. The phases of impact, spread and deposition were characterized in terms of spread factor for different operating conditions. No rebound occurred under the tested conditions. Experimental data fairly well agreed with theoretical correlations from literature in terms of morphological evolution of the impacted droplets and excess rebound energy after collisions with the walls.

Application of data-driven methods for energy system modelling demonstrated on an adaptive cooling supply system

The efficient and sustainable operation of building energy systems is playing an increasingly important role in most industrialized countries. At the same time, building energy systems are becoming increasingly complex; fault-free and optimal operation, under dynamic boundary conditions, is becoming more and more challenging. There are many approaches in research to address the optimal control problem of building energy systems, such as Rule-based Control, Model Predictive Control, or Adaptive Control. However, most methods rely on models of the system dynamics with high prediction accuracies. This is especially the case in Model Predictive Control, where the model is part of a continuously executed optimization problem; but models are also required when it comes to the optimal design of Rule-based Controllers, the safe pre-training of Adaptive Controllers, or model-based fault detection. A limiting factor for the manual development of physical models, for building energy systems, are the low monetary incentives for engineering services, due to the low energy prices in most countries. In addition, the creation of such models is time-consuming and error-prone, even for domain experts. Another weakness is that changes in the system dynamics are not automatically adapted within the models. These challenges are contrasted by an increasing availability of monitoring-data and computational power in recent years; with machine-learning algorithms, these resources are used in numerous application areas to achieve very promising results. Machine-learning methods can help to obtain data-driven, self-calibrating models, which can be learned from monitoring-data. In this paper, we apply methods for automated data-driven model generation. We demonstrate how machine-learning algorithms together with structured hyper-parameter tuning can be used to model individual subsystems as well as a complete energy supply system. To represent the dynamics of the supply system, it is first decomposed into simple functional relationships, which are aggregated into the overall system after training of the comparatively simple subsystem models. We evaluate the accuracy of the data-driven subsystem models using established metrics for the evaluation of regression models, namely the R2-score and the RMSE. The considered system is integrated into a district cooling network and consists of two compression chillers and an ice storage unit. Our investigations show that the dynamics of the subsystems can be learned with high accuracies, depending on the operation mode and the selected features. The prediction of the power demand of the compression chillers is learned with R2-scores between 0.94 and 0.99 and RMSE values between 2.02 kW and 3.51 kW. Also, the prediction of the percentage of ice formation within the ice storage is learned accurately with a R2-score of 1 and RMSE values between 0.08 % and 0.72 %. The dynamics of the aggregated system also show plausible behavior and can thus be used in future work. This work is part of an ongoing research project with the aim to optimize the operation of the entire campus cooling energy supply system. Our results show that, if detailed monitoring-data are available, data-driven modelling represents a viable alternative to the labor-intensive physical modelling approach. Furthermore, we emphasize the importance of structured hyper-parameter tuning, discuss the specifics of different machine-learning algorithms, and elaborate on possible future developments in this research area.

Surface currents in operational oceanography: Key applications, mechanisms, and methods

ABSTRACT This paper reviews physical mechanisms, observation techniques and modelling approaches dealing with surface currents on short time scales (hours to days) relevant for operational oceanography. Key motivations for this article include fundamental difficulties in reliable measurements and the persistent lack of a widely held consensus on the definition of surface currents. These problems are augmented by the fact that various methods to observe and model ocean currents yield very different representations of a surface current. We distinguish between four applicable definitions for surface currents; (i) the interfacial surface current, (ii) the direct wind-driven surface current, (iii) the surface boundary layer current, and (iv) an effective drift current. Finally, we discuss challenges in synthesising various data sources of surface currents - i.e. observational and modelling – and take a view on the predictability of surface currents concluding with arguments that parts of the surface circulation exhibit predictability useful in an operational context.

A combined physical and DEM modelling approach to investigate particle shape effects on load movement in tumbling mills

The discrete element method (DEM) which is used to simulate granular flows often assumes spherical shape for particles. This assumption is legitimized by the added complexity of non-spherical shape representation, contact detection and computational cost. In this work, the difference between the dynamics of non-spherical and spherical particles was studied in detail by a combined physical and DEM modeling approach. An in-house developed DEM software called KMPCDEM©, which was coded to handle non-spherical particles, was used to simulate the behavior of particles. To calibrate the model parameters, a model tumbling mill (100 cm diameter and 10.8 cm length) with one transparent end was used which made accurate photography possible. The tests were performed at filling of 20% and mill speed of 85% of critical speed with steel balls and wood cubes. In the simulation, each cubical particle was represented with clusters of spheres (with identical size) by particle packing algorithm for contact detection and contact-force calculation. Comparison of the simulation and experimental results showed that the difference between the measured and predicted impact toe, shoulder angle and bulk toe angle were 3, 4 and 5°, respectively. The significant change in the charge movement and structure on account of non-spherical particles was reflected in the amount of in-flight charge, and positions of shoulder, impact toe and bulk toe. It found that there was a 17% difference in the amount of in-flight of charge between cubical and spherical particles. The marked difference was attributed to higher interlocking of non-spherical particles in comparison to spherical balls. The results showed that cubical particles participated 5% more in the high energy impact action compared to that of the spherical particles. The simulation computation time increased by 35 times when the shape of particles changed from spherical to cubical.

Modelling heating and cooling energy demand for building stock using a hybrid approach

The building sector accounts for 30% of final energy consumption and 28% of global energy-related carbon dioxide emissions, with space heating and cooling consuming a large share of total buildings’ energy consumption. Building stock modelling for space heating and cooling energy prediction provides critical insights on the stock energy consumption and aid the building retrofit policy-making process with the evaluation of the energy-saving potential. By combining the physical modelling approach and data-driven approach, a hybrid approach is applicable for modelling the heating and cooling energy consumption of the building stock, including both residential buildings and non-residential buildings. Within this framework, the Urban Modelling Interface (UMI) tool has been used for physical modelling to generate heating and cooling energy use intensity. Then, ten different machine learning models, including Gaussian radial basis function kernel support vector regression, linear kernel support vector regression, polynomial kernel support vector regression, random forests, extreme gradient boosting, ordinary least-squares linear regression, ridge regression, least absolute shrinkage and selection operator, elastic net and artificial neural network, have been applied to predict heating and cooling energy use intensity (EUI). The approach has been demonstrated using a case study in Chongqing, China. The results show that machine learning models can achieve accurate building heating and cooling EUI prediction, with the polynomial kernel support vector regression showing the best accuracy at the level of a single building, and the Gaussian radial basis function kernel support vector regression performing the best at the stock level. Machine learning models generated by proposed hybrid approach not only provide quickly prediction of building space heating and cooling energy consumption at the stock level, but also support building retrofit decision makings by evaluate energy saving potential of various retrofit options.