Due to their high energy intensity, buildings play a major role in the current worldwide energy transition. Building models are ubiquitous since they are needed at each stage of the life of ...buildings, i.e. for design, retrofitting, and control operations. Classical white-box models, based on physical equations, are bound to follow the laws of physics but the specific design of their underlying structure might hinder their expressiveness and hence their accuracy. On the other hand, black-box models are better suited to capture nonlinear building dynamics and thus can often achieve better accuracy, but they require a lot of data and might not follow the laws of physics, a problem that is particularly common for neural network (NN) models. To counter this known generalization issue, physics-informed NNs have recently been introduced, where researchers introduce prior knowledge in the structure of NNs to ground them in known underlying physical laws and avoid classical NN generalization issues.
In this work, we present a novel physics-informed NN architecture, dubbed Physically Consistent NN (PCNN), which only requires past operational data and no engineering overhead, including prior knowledge in a linear module running in parallel to a classical NN. We formally prove that such networks are physically consistent – by design and even on unseen data – with respect to different control inputs and temperatures outside and in neighboring zones. We demonstrate their performance on a case study, where the PCNN attains an accuracy up to 40% better than a classical physics-based resistance-capacitance model on 3-day long prediction horizons. Furthermore, despite their constrained structure, PCNNs attain similar performance to classical NNs on the validation data, overfitting the training data less and retaining high expressiveness to tackle the generalization issue.
Display omitted
•Physics-based building zone temperature models are cumbersome.•Classical Neural Network models fail to capture the underlying physics.•Novel Physically Consistent Neural Network (PCNN) models are proposed.•PCNNs are mathematically shown to retain physical consistency by design.•They show superior accuracy to classical physics-based methods.
Buildings constitute more than 40% of total primary energy consumption worldwide and are bound to play an important role in the energy transition process. To unlock their potential, we need ...sophisticated controllers that can understand the underlying non-linear thermal dynamics of buildings, consider user comfort constraints and produce optimal control actions. A crucial challenge for developing such controllers is obtaining an accurate control-oriented model of a building. To address this challenge, we present a novel, data-driven modeling approach using physics informed neural networks. With this, we aim to combine the strengths of two prominent modeling frameworks: the interpretability of building physics models and the expressive power of neural networks. Specifically, we use measured data and prior information about building parameters to realize a neural network model that is guided by building physics and can model the temporal evolution of room temperature, power consumption as well as the hidden state, i.e., the temperature of building thermal mass. The main research contributions of this work are: (1) we propose two new variants of physics informed neural network architectures for the task of control-oriented thermal modeling of buildings, (2) we show that training these architectures is data-efficient, requiring less training data compared to conventional, non-physics informed neural networks, and (3) we show that these architectures achieve more accurate predictions than conventional neural networks for longer prediction horizons (as needed for effective control). We test the prediction performance of the proposed architectures using both simulated and real-word data to demonstrate (2) and (3) and argue that the proposed physics informed neural network architectures can be used for control-oriented modeling.
•We propose data-driven models designed for thermal control in buildings.•We develop and analyze variants of physics-informed neural networks (PhysNet).•We present case studies on both simulated and real-world data.•PhysNets are superior in prediction accuracy, training data efficiency and robustness.
In the last decades, 3D city models appear to have been predominantly used for visualisation; however, today they are being increasingly employed in a number of domains and for a large range of tasks ...beyond visualisation. In this paper, we seek to understand and document the state of the art regarding the utilisation of 3D city models across multiple domains based on a comprehensive literature study including hundreds of research papers, technical reports and online resources. A challenge in a study such as ours is that the ways in which 3D city models are used cannot be readily listed due to fuzziness, terminological ambiguity, unclear added-value of 3D geoinformation in some instances, and absence of technical information. To address this challenge, we delineate a hierarchical terminology (spatial operations, use cases, applications), and develop a theoretical reasoning to segment and categorise the diverse uses of 3D city models. Following this framework, we provide a list of identified use cases of 3D city models (with a description of each), and their applications. Our study demonstrates that 3D city models are employed in at least 29 use cases that are a part of more than 100 applications. The classified inventory could be useful for scientists as well as stakeholders in the geospatial industry, such as companies and national mapping agencies, as it may serve as a reference document to better position their operations, design product portfolios, and to better understand the market.
Semantic 3D building models are widely available and used in numerous applications. Such 3D building models display rich semantics but no façade openings, chiefly owing to their aerial acquisition ...techniques. Hence, refining models’ façades using dense, street-level, terrestrial point clouds seems a promising strategy. In this paper, we propose a method of combining visibility analysis and neural networks for enriching 3D models with window and door features. In the method, occupancy voxels are fused with classified point clouds, which provides semantics to voxels. Voxels are also used to identify conflicts between laser observations and 3D models. The semantic voxels and conflicts are combined in a Bayesian network to classify and delineate façade openings, which are reconstructed using a 3D model library. Unaffected building semantics is preserved while the updated one is added, thereby upgrading the building model to LoD3. Moreover, Bayesian network results are back-projected onto point clouds to improve points’ classification accuracy. We tested our method on a municipal CityGML LoD2 repository and the open point cloud datasets: TUM-MLS-2016 and TUM-FAÇADE. Validation results revealed that the method improves the accuracy of point cloud semantic segmentation and upgrades buildings with façade elements. The method can be applied to enhance the accuracy of urban simulations and facilitate the development of semantic segmentation algorithms.
The level of detail (LOD) concept of the OGC standard CityGML 2.0 is intended to differentiate multi-scale representations of semantic 3D city models. The concept is in practice principally used to ...indicate the geometric detail of a model, primarily of buildings. Despite the popularity and the general acceptance of this categorisation, we argue in this paper that from a geometric point of view the five LODs are insufficient and that their specification is ambiguous.
We solve these shortcomings with a better definition of LODs and their refinement. Hereby we present a refined set of 16 LODs focused on the grade of the exterior geometry of buildings, which provide a stricter specification and allow less modelling freedom. This series is a result of an exhaustive research into currently available 3D city models, production workflows, and capabilities of acquisition techniques. Our specification also includes two hybrid models that reflect common acquisition practices. The new LODs are in line with the LODs of CityGML 2.0, and are intended to supplement, rather than replace the geometric part of the current specification. While in our paper we focus on the geometric aspect of the models, our specification is compatible with different levels of semantic granularity. Furthermore, the improved LODs can be considered format-agnostic.
Among other benefits, the refined specification could be useful for companies for a better definition of their product portfolios, and for researchers to specify data requirements when presenting use cases of 3D city models. We support our refined LODs with experiments, proving their uniqueness by showing that each yields a different result in a 3D spatial operation.
Display omitted
•CityGML LODs are an industry standard for conveying the grade of 3D city models.•The 5 LODs are not defined precisely, and they are not sufficient for this purpose.•We present a refined series of 16 LODs that overcomes these issues.
Intra-urban polycentricity has often been described by a balanced distribution of jobs/residences outside the traditional core cities in so-called (sub-)centres. Recently, this purely socioeconomic ...view has changed, so that centres are also increasingly understood as a physical manifestation of spatial development policies. Built-up volumes derived from 3D-building models are therefore frequently used instead of or as complement to employment/population figures when studying intra-urban polycentricity. However, such data are expensive and not available universally and permit only geographically limited investigations to date. To overcome this constraint, we investigate whether globally available and consistent TanDEM-X nDSMs (TDX) provide a valid data base for intra-urban polycentricity research based on built-up volumes. Our study focuses on four urban regions in Germany for which we have obtained official 3D-building models (LoD-1). For each study site, we derive aggregated built-up volumes from the TDX and the LoD-1 data and identify (sub-)centres. We use three centre identification algorithms to account for the diversity of methods and outcomes. We consider the LoD-1 (sub-)centres as reference and the TDX (sub-)centres as the entities to be reviewed. First, we quantify their spatial agreement and compare if polycentricity measures calculated based on both data sets lead to similar results. Second, we explore possible causes for discrepancies between the TDX/LoD-1 (sub-)centres. We find high spatial resemblances between TDX and LoD-1 (sub-)centres. Accordingly, we observe that polycentricity measures display similar trends among the two data sets. Nevertheless, we also show that the agreement between TDX and LoD-1 centres can be affected in uneven terrain, in sparsely built-up areas, and by the algorithms used to identify (sub-)centres. Overall, our results suggest that TDX nDSMs reflect the distribution of built-up structures in sufficient detail so that local-spatial densifications – here equated with (sub-)centres – can be appropriately studied. We therefore conclude that TDX data offer a great potential for the thematic domain of morphological urban analysis at large scale.
•The suitability of TanDEM-X data for large-scale analysis of intra-urban polycentricity is tested.•Centre detection capabilities based on 3D-building models and TanDEM-X data are compared.•Similar polycentricity analysis results are achieved based on TanDEM-X data and 3D-building models.•Factors influencing the agreement between centres from the two data sets are uncovered.•The potential of TanDEM-X data for urban analysis is highlighted with a practical example.
The ubiquity of cameras built in mobile devices has resulted in a renewed interest in image-based localisation in indoor environments where the global navigation satellite system (GNSS) signals are ...not available. Existing approaches for indoor localisation using images either require an initial location or need first to perform a 3D reconstruction of the whole environment using structure-from-motion (SfM) methods, which is challenging and time-consuming for large indoor spaces. In this paper, a visual localisation approach is proposed to eliminate the requirement of image-based reconstruction of the indoor environment by using a 3D indoor model. A deep convolutional neural network (DCNN) is fine-tuned using synthetic images obtained from the 3D indoor model to regress the camera pose. Results of the experiments indicate that the proposed approach can be used for indoor localisation in real-time with an accuracy of approximately 2 m.
•Novel pipeline for automatically reconstruct LOD3 models of free-standing buildings.•New metrics for evaluating the performance of a LOD reconstruction.•Benchmark dataset for evaluating the ...reconstruction of SfM as well as LOD3 models.•Deep learning models for facade and opening segmentation of free-standing masonry buildings.
This paper describes a pipeline for automatically generating level of detail (LOD) models (digital twins), specifically LOD2 and LOD3, from free-standing buildings. Our approach combines structure from motion (SfM) with deep-learning-based segmentation techniques. Given multiple-view images of a building, we compute a three-dimensional (3D) planar abstraction (LOD2 model) of its point cloud using SfM techniques. To obtain LOD3 models, we use deep learning to perform semantic segmentation of the openings in the two-dimensional (2D) images. Unlike existing approaches, we do not rely on complex input, pre-defined 3D shapes or manual intervention. To demonstrate the robustness of our method, we show that it can generate 3D building shapes from a collection of building images with no further input. For evaluating reconstructions, we also propose two novel metrics. The first is a Euclidean–distance-based correlation of the 3D building model with the point cloud. The second involves re-projecting 3D model facades onto source photos to determine dice scores with respect to the ground-truth masks. Finally, we make the code, the image datasets, SfM outputs, and digital twins reported in this work publicly available in github.com/eesd-epfl/LOD3_buildings and doi.org/10.5281/zenodo.6651663. With this work we aim to contribute research in applications such as construction management, city planning, and mechanical analysis, among others.
Current procedures for the rapid inspection of buildings and infrastructure are subjective, time-consuming, and cumbersome to document, necessitating new technologies to automate the process and ...eliminate these shortcomings. Fortunately, recent developments in imaging devices and artificial intelligence, such as computer vision, provide the necessary tools for this, though they are not yet integrated into infrastructure applications. In this paper, we propose an end-to-end pipeline that generates damage-augmented digital twins for buildings at LOD3, including geometrical information as well as data pertaining to damage condition and its characterization. Our framework incorporates multiple-view images to (1) create a level of detail model, (2) segment damage information, and (3) characterize damage. The core of the method is the structure from motion, which is used to reconstruct the building scene, and machine-learning models that segment and characterize damage. In contrast to current practices, our method does not require manual intervention, generates lightweight models, and can be applied to a wide range of assets. The results generated with our pipeline represent a significant step towards an automated infrastructure damage assessment. We intend to expand our work in the future to include real-time applications and applications to other types of infrastructure. Codes and data sets are publicly available (https://github.com/eesd-epfl/DADT_buildings and https://doi.org/10.5281/zenodo.7767478).
•A novel framework to generate Damage Augmented Digital Twins (DADT) of buildings.•State-of-the-art solutions for 3D models, crack segmentation and characterization.•Bench-marking data set for evaluating the automated generation of DADT of buildings.