Abstract:With the development of engineering information level and the monitoring technology in the field of shield tunnel, the recorded engineering data contains the internal information of tunneling equipment and its interaction with the external stratum. Machine learning has more application space than traditional modeling statistical analysis methods because of its strong data analysis ability and no requirement on prior theoretical formula and expert knowledge. Improving the efficiency and safety level of shield tunnel construction is helpful to deeply mine the collected information and data and analyze their internal relationship through machine learning method. This paper briefly describes the basic principle of machine learning methods, summarizes and analyzes its application in shield tunnel engineering. In particular, the progress on the equipment status analysis, shield performance prediction, geological parameters analysis, prediction of ground surface deformation and examination of tunnel hazard based on the machine learning method are summarized. Finally, the key problems to be solved so as to realize the intelligent shield tunnel engineering are analyzed and forecasted.

Abstract:Artificial intelligence is the core driver of the next generation of industrial change. It is an important method to comprehensively improve digitalization, automation, informatization, and intelligence in the field of civil engineering. To gain a comprehensive understanding of the development and application of artificial intelligence in civil engineering. The basic research areas of artificial intelligence are analyzed qualitatively. The current research status of artificial intelligence in civil engineering design, manufacturing, and maintenance phases is quantitatively analyzed. The CiteSpace visualization tool is used to dig deeper into the problems, development bottlenecks, and research trends of artificial intelligence in civil engineering, and give corresponding solutions and research ideas. The review of the literature found that a significant amount of artificial intelligence research has been conducted in the field of civil engineering. However the development of intelligence has been uneven at various stages, and there are limitations in practical applications. Therefore, it is necessary to deeply explore the cross-integration of intelligent technologies such as neural networks, big data, and deep learning in the full life cycle of civil engineering. To promote the synergistic development of artificial intelligence research in the field of civil engineering.

Abstract:As a result of the COVID-19 pandemic, a growing number of activities are moving online. With the introduction of the metaverse concept, virtual digital technology has entered the day-to-day life and industrial activities. To date, the applications of metaverse can be found in the education, commercial and industrial sectors. The intelligence level of the construction industry is constantly rising due to the wide and in-depth building information technology. Virtual digital technology is playing a vital role in accelerating the development of the construction industry. This paper introduces the state-of-the-art of the metaverse concept, the interdeperndency between the metaverse and the existing building information technologies. The potential applications of the metaverse in the field of building construction have also been discussed. Findings indicate that the development of a metaverse depends on the integration of existing digital technology such as extended reality, digital twin and blockchain. Metaverse will render significant changes to the production modes in the field of architecture and construction as a new technology, and it may also create new business models.

Abstract:In the context of the transformation and upgrading of the construction industry, aiming at the problems of low management efficiency of large-scale building operation and maintenance process, insufficient accuracy of decision analysis of various operation and maintenance events, and the degree of management intelligence needs to be improved, this study proposes an intelligent operation and maintenance theory system and implementation method based on digital twins. Firstly, the information required for intelligent operation and maintenance of large buildings is summarized, and the problems to be solved in intelligent operation and maintenance are analyzed. The intelligent sensing technology is integrated into digital twins to propose an operation mechanism of intelligent operation and maintenance, and the intelligent operation and maintenance architecture system for large buildings and the multidimensional and multi-scale twinning model is formed. The realization method of building intelligent operation and maintenance is studied. It includes the collection and transmission of all elements of intelligent operation and maintenance information, the construction and operation of intelligent operation and maintenance twins, the management mechanism of intelligent operation and maintenance twin data and the architecture of intelligent operation and maintenance platform. Under the guidance of the intelligent operation and maintenance theory system and implementation method, an intelligent operation and maintenance platform is formed and applied to the operation and maintenance management of a large construction project. Considering the virtual-real interaction and spatio-temporal evolution in the process of operation and maintenance management, the effectiveness of digital twins in improving the intelligence of operation and maintenance management is preliminarily verified.

Abstract:In order to achieve the intelligent placement of dampers in frame structures, the dual-objective optimization algorithm (DOOA) and generative adversarial network (GAN) algorithm are employed for the vertical and horizontal intelligent placement of dampers, respectively, based on the damping design principle and intelligent algorithm. Two seismic design engineering cases of frame structures are applied. In the seismic design of frame structures, dual-objective optimization is adopted for vertical damper placement. Compared with the layer-by-layer approximation method, engineer-designed optimized damper placement schemes, and non-damping design, the vertical arrangement scheme of dampers obtained by the improved optimization algorithm can effectively reduce inter-story drift angles and floor accelerations, and enhance the seismic performance of the original structure. After determining the number of dampers for each floor, the plane installation position of dampers on each floor can be quickly and automatically selected and determined by using the trained generative adversarial network generation model. The comprehensive evaluation index of similarity difference degree between the generated plane layout and the plane layout designed by the engineer is less than the critical value of 0.1, which indicates that the similarity between the two is high, and it is beneficial to improve the torsion resistance of the original structure. The combination of dual-objective optimization and generative adversarial network can meet the seismic performance objectives of frame structures, and enable to achieve of intelligent design of the damper placement scheme, and improve the efficiency of seismic design engineering.

Abstract:The traditional structural design method based on mechanical analysis software has some limitations, such as low efficiency and expert experience reliance. The efficient automatic structural optimization design can be achieved by using intelligent algorithms. However, due to the random search feature, the optimization result and convergence are highly dependent on the parameter settings of the algorithm whose reasonable values need to be determined by the trial-and-error procedure. It results in inefficient optimization and substantial computational cost. Therefore, this paper introduces the multi-population collaboration and information sharing mechanism to improve such problems and its applicability in the structural optimization design is studied. The finite element model of a steel frame is built by MSC.Marc and the equivalent horizontal load from earthquake obtained by base shear method is exerted on the structure. The automatic optimization process is established based on finite element software and the intelligent algorithm with the aim of mininizing the total material cost of the structure. Multiple structural constraints are considered including the inter-story drift ratio, the stress ratio, and the stability and width-thickness ratio of the component. Several strategies are used to improve the performance of the genetic algorithm, such as the fitness scaling, the direction-based crossover operator, the non-uniform mutation operator, the adaptive probability, the elite strategy, the duplicate substitution mechanism, and the constraint-based strategy. Then the multi-population mechanism is introduced to such an algorithm. The results of different algorithms are compared with each other, which shows that the multi-population genetic algorithm can improve the dependence of optimization results on algorithm parameters and the efficiency of structural optimization design.

Abstract:Intelligent structural design in the scheme phase is an essential component of intelligent construction. Existing studies have proposed the deep neural network-based framework of intelligent generative structural design, intelligent design algorithms, and design performance evaluation methods for shear wall structures, which have developed intelligent structural design methods from data-driven to physics-enhanced data-driven. However, little detailed design performance comparison of data-driven and physics-enhanced methods under different design conditions is conducted. Furthermore, the relationship between the computer vision-based and mechanical analysis-based evaluation methods are still unclear, resulting in difficulties in effectively guaranteeing the rationality of the computer vision-based evaluation methods. Hence, in this study, the comparative analysis of data-driven and physics-enhanced intelligent design methods is conducted by algorithm comparison and case studies; and the consistent relationship between computer vision-based and mechanical analysis-based evaluation methods is validated. The comparison results reveal that data-driven methods are more prone to be limited by the quality and quantity of training data. In contrast, the physics-enhanced data-driven design method is more robust under different design conditions and is little affected by the data-caused limitation. Moreover, the rationality threshold of the computer vision-based evaluation index (SCV) is 0.5, corresponding to a difference in the mechanical performance of approximately 10%.

Abstract:The unqualified size of prefabricated component in the production process will lead to the failure of the installation on the construction site, and affect the construction period. In order to promote the process of intelligent production of prefabricated components. Based on a convolutional neural network, the prefabricated laminated board is used as an example to study the intelligent detection method of the production process. Design and install an image acquisition system on the production line, establish a prefabricated laminated board detection data set, and use the YOLOv5 algorithm to detect the concrete plate, the embedded PVC junction box and the overhanging steel bar. The fixed magnetic box is used as the benchmark to analyze the detection error of the dimension of the concrete plate and the coordinate of the embedded PVC junction box, and maintains a high recognition accuracy with a smaller parameter scale of the training data set. The result shows that the method can effectively detect the number and dimension of the concrete plate, the number and coordinate of the embedded PVC junction box, and detect the overhanging steel bar of unqualified bending direction. The method can reduce labor costs, improve detection accuracy, speed up detection process, and improve the delivery quality of prefabricated laminated board.

Abstract:Dimensional quality assessment (DQA) of precast laminated panels (PLPs) is required to ensure the installation of the components on the construction site. However, existing methods cannot meet the precision and comprehensiveness of PLPs, DQA. In this paper, we propose an automated multi-dimensional quality assessment method for PLPs based on point cloud data (PCD). After pre-processing the original PCD, the scanned PLPs, PCD is automatically extracted by the machine learning algorithms, and down-mapped in different directions to generate two-dimensional (2D) images. DQA of PLPs can be realized automatically by the image feature detection algorithms. The validation experiment is conducted on three PLPs. The result shows that the proposed method meets the dimensional quality assessment accuracy of precast laminated panels.

Abstract:The time span of the civil engineering structural damage state usually accounts for a small part of the total life cycle. In order to solve the problem that traditional data-driven structural damage identification methods lack enough damage state data for training, a structural damage identification method based on digital twins and deep learning is proposed in this paper for practical application. Firstly, the digital twin is constructed by using the numerical simulation model and online monitoring data to obtain the “big data” of the structural dynamic response under different damage conditions. Secondly, to get rid of the dependence on the external excitation, the empirical mode decomposition method and transmissibility function are used to preprocess the obtained data. Then, the damage identification is realized by using deep learning. To verify the effectiveness of this method, untrained monitoring data of structures are analyzed. The results show that the method has good generalization ability and can identify the structural damage condition effectively. The problem of data hunger is solved by digital twin technology, and the deep neural network trained by the intrinsic mode vibration transmissibility function data sets can still maintain a high accuracy of damage identification without any seismic information. The combination of the two methods can make structural health monitoring more active, reliable and efficient.

Abstract:The aerodynamic shape of the bluff body section is very important to the aerodynamic performance. However, it takes a lot of time to obtain the aerodynamic performance of the bluff body section using traditional wind tunnel tests and CFD simulation calculations, which greatly affects the aerodynamic performance evaluation efficiency of the bluff body section , s aerodynamic shape. This paper proposes to use the deep learning technology of convolutional neural networks to realize the rapid prediction of aerodynamic performance. After the deep learning model is trained, the shape information and the shape-related flow field information can be input to output the drag coefficients under different geometric shapes, then the aerodynamic performance of the bluff body section. However to find the best deep learning model, this paper optimizes the depth and width of the convolutional neural network structure through comprehensive judgment error and time performance. The output resistance coefficient of the deep learning model is compared with the CFD calculation results. It is found that the error meets the expected requirements, and the prediction time based on the deep learning network is an order of magnitude improvement compared with the calculation time required by the traditional method. It can be used as the bluff body section aerodynamic shape optimization in the future.

Abstract:Compared with the traditional cast-in-place frame structure, the construction of the prefabricated steel joints concrete frame structure is faster, but there is a sudden change of stress at the joints of the new frame steel joints, so it is necessary to adopt the digital twin method to monitor the construction process of the actual engineering structure to obtain the influence of steel joints on the structural performance. Based on building information modeling (BIM), finite element, and sensor technologies, an intelligent building framework of the new type of prefabricated concrete frame structure with a steel connector based on digital twin was established in this study. An intelligent construction method based on digital twin was proposed in terms of physical data collection, virtual model establishment and model information interaction. In the engineering intelligent construction stage, real-time monitoring of data of the new prefabricated concrete frame with steel connector was realized via sensor technology, and sensor data was compared with BIM and finite element data. The frame structure stress in the physical model would be further adjusted and modified. Finally, the digital twin model of the prefabricated concrete frame with a steel connector was established and applied. The results show that the digital dual model can effectively monitor prefabricated concrete frame with steel connectors in real-time, predict the hazard position based on sensor network and spatiotemporal parameter analysis, reduce the resource consumption, and provide sufficient data information for subsequent applications.

Abstract:In the traditional structural design process, rural buildings require a lot of manual calculations and repeated modeling. However, due to the constraint of construction cost, they cannot be designed and checked professionally as urban types, and the safety and economy are difficult to meet the requirements. In this study, an intelligent design approach for structural design of rural light steel frame structure was proposed, including the intelligent modeling and optimization. Based on the automatic layer classification method (ALCM), optical character recognition technology (OCR) and adaptive block algorithm, BIM intelligent modeling method was proposed where layer recognition, the extraction of axis text data and wall contour were included, and generated structural models satisfied the requirements of engineering practice. Based on the proposed two-stage simulated annealing algorithm, the intelligent optimization method was proposed and verified by case histories. Results showed that the proposed intelligent design method was feasible. Compared with the traditional method, its time use could be shortened by more than 70%, and the material consumption and structural design parameters were similar.

Abstract:This paper aims at solving the contradictive design problem of the modular steel frame structure in cold regions considering both energy- and cost- saving. A synchronous optimization study with energy consumption and cost objectives is hence carried out for target modular steel frame structures. Parametric modeling of modular steel frame structures is studied according to their characteristics. An automatic BIM modeling method is developed for modular steel frame structures. The building energy consumption is modeled using various machine learning algorithms based on the database constructed from the Energyplus software. The proposed XGBoost model provides efficient and accurate predictions for the building energy consumption. The energy consumption model as well as the cost formula serve as the objective functions in NSGA-Ⅱ algorithm to build the design optimization program. During optimization, structural bearing capacity must be satisfied. Pareto solution set is then achieved by the developed program and analyzed. By solving the multi-objective design problem of modular steel frame structures with advanced computing techniques, this study contributes to the intelligent upgrade of the modular steel frame structure industry, and realizes its rapid and efficient design.

Abstract:Construction progress is a key part of project management. Traditional progress management mostly relies on manual inspection, which is time-consuming and can not guarantee the progress evaluation. In order to automatically and efficiently monitor construction progress, this article proposed an intelligent framework for indoor construction progress evaluation. This framework can automatically extract the tile laying progress of indoor wall and ground based on the improved Mask R-CNN, and map the progress results to the BIM for visualization using camera tracking algorithm. The framework was successfully applied in a building project in Shanghai with a high precision of image segmentation, verifying the feasibility of the presented framework. This study provides theoretical and practical reference for the automated progress tracking and unmanned construction progress supervision of indoor continuous space.

Abstract:With the increasing complexity of construction project management, more and more automatic and intelligent construction schedule management methods are concerned by the traditional management. However, the existing mainstream methods are limited by high cost and complex use, which are difficult to apply to intricate construction schedule management scenarios. By comparing the characteristics of various kinds of 3D reconstruction technology, this study built a collaborative management system of construction schedule based on deep learning 3D Reconstruction Technology (DLR-P). By collecting the real-time image information of the construction site, the system completes the reconstruction from 2D information to 3D, and realizes the automatic control of the construction progress combined with BIM dynamic model technology. In view of the system, this study conducted a case study in the construction site of a project in Banan District of Chongqing, and analyzed the data in the process of system operation. The results show that the average 3D reconstruction time of construction schedule collaborative management system (DLR-P) based on deep learning is 61 seconds, which can meet the basic schedule management requirements, realize the automatic management of construction schedule, and effectively improve the efficiency. Compared with the existing mode, it has great advantages in the operation cost and convenience.

Abstract:At present, finite element modeling or laboratory testing methods are generally used in the research of the overall stability of high-strength steel members. However, the prediction method based on machine learning (ML) has greatly improved the accuracy and convenience of component performance prediction. To accurately predict the overall stability of welded constant section box columns made of high strength steel, ML method together with a database based on the fiber model is proposed in this paper. Firstly, the input and output parameters of the model are determined, and the database is provided. Then, three different ML models and empirical models in the existing specifications are selected for prediction, and the performance is compared according to the evaluation index. Finally, the rationality of ML models is analyzed according to interpretable algorithms. The results show that the prediction results of most ML models are in good agreement with the experimental results, which are slightly higher than the empirical models, and the Gaussian process regression model has the best prediction performance for the overall stability of high-strength steel members; the influential trend of various parameters on the overall stability of components meets the expectation, which verifies the rationality and reliability of the ML model; the regularized slenderness ratio has the greatest influence on the prediction results, while the initial defects have the least.

Abstract:In order to further investigate the compressive performance of composite columns with 3D printed concrete permanent formwork, an interface based finite element (FE) model was established to analyze the load-displacement response and failure mode of composite columns and the same size cast-in-place columns under axial compression on the basis of the experiment. The parameter sensitivity analysis of composite columns with 3D printed concrete permanent formwork was carried out with consideration of the interface bonding property, compressive strength of cast-in-place concrete, thickness of printing template, and load eccentricity. Results showed that the ultimate axial compression bearing capacity of composite columns increased with the development of shear strength, stiffness of interface and compressive strength of cast-in-place concrete. As the compressive strength of the printing material is higher than that of the cast-in-place concrete, the compressive ultimate bearing capacity of the composite column appeared approximately linear growth with thickness of printing template, and negative linear correlation with the load eccentricity. In addition, the influence of eccentricity on the reduction of the ultimate bearing capacity of composite columns is greater than that of cast-in-place columns.

Abstract:In order to solve the problem of low utilization rate of formwork in the production process of precast concrete components, this paper proposes an efficient component layout sequencing and positioning optimization method. Firstly, the mathematical model based on the minimum horizontal line layout algorithm is established with the objective of minimizing the total length of the mould table occupied by the components, considering the actual constraints of the mould table, mould, component layout size and operation space in the production process of the components. Secondly, for the purpose of determining the sequence of component layout, this study uses the improved grey wolf algorithm and the components to be arranged are coded through component type, size information, and constraint conditions. Via setting the coefficient and free coefficient, which can control the convergence speed and convergence range of the grey wolf algorithm, the optimal layout scheme is obtained via multiple optimization iterations. Through empirical analysis, this method can achieve better optimization effect in a short time, and the utilization rate of the mold is significantly improved, which verifies the effectiveness and feasibility of the method.

Abstract:Mainstream deep learning algorithm for crack segmentation consumes a lot of computing resources while the traditional image processing methods are of low detection accuracy and lost crack features. In order to realize the real-time detection of concrete cracks and the segmentation of cracks at the pixel level, a crack semantic segmentation model based on lightweight convolutional neural network M-Unet is proposed. Firstly, the MobileNet_V2 lightweight network is improved, its network structure is trimmed and the activation function is optimized, and then the encoder part with huge parameters of U-Net is replaced by the improved MobileNet_V2 to realize the lightweight of the model and improve the segmentation effect of cracks. The SegCracks data set containing 5 160 crack images is constructed to verify the proposed method. The experimental results show that the crack segmentation effect of the optimized M-Unet is better than the mainstream segmentation networks of U-Net, FCN8 and SegNet and the traditional image processing techniques, the obtained IoU_Score is 96.10%, F1_Score is 97.99%. Compared with the original U-Net, the weight file size M-Unet is reduced by 7 %, the iteration time and prediction time are reduced by 63.3% and 68.6% respectively, and the IoU_Score and F1_Score are increased by 5.79 % and 3.14 % respectively. The cross validation results on different open source data sets are good, which shows that the proposed network has the advantages of high accuracy, good robustness and strong generalization ability.

Abstract:Aiming at the problems of low efficiency, unsatisfactory detection effect and poor safety of manual detection methods for building exterior wall cracks, a crack detection method based on aerial images and computer vision was proposed. Firstly, the Unmanned Aerial Vehicle (UAV) was used to collect the crack images through aerial photography around the buildings, and a crack dataset was constructed. Secondly, the U-Net was optimized to solve the problems of discontinuous segmentation of slender cracks as well as the missed and false detection under complex backgrounds. The encoder was replaced with pre-trained ResNet50 to improve the feature expression ability of the model. An improved Atrous Spatial Pyramid Pooling (ASPP) module was added to obtain multi-scale context information. The improved loss function was used to deal with the problem of extremely uneven distribution of positive and negative samples in crack images. Experiments show that the improved U-Net model solved the problems existing in the original model; the IoU and F1_score were increased by 3.53% and 4.18%, respectively. Compared with the classical segmentation model, the improved model has the best crack segmentation performance. Compared with manual detection methods, it can efficiently, accurately, and safely detect building exterior wall cracks.

Abstract:Large number of cracks, as the main disease, exist in the concrete bridge, and some cracks will be secondary dehisced after maintenance, and the crack repair traces are easily confused with concrete spalling and other defects when identifying disease intelligently, as a result of which identifying the crack repair traces accurately is not only the basis for identification of secondary cracks but also important for identification of the overall disease of concrete bridges. To obtain crack repair traces with continuous edges clearly, Poisson-noise is firstly added to the image of crack repair traces, then bilateral-filtering was adopted to smooth the added and the original noise, the Otsu algorithm was also used to segment the image of crack repair traces. The filtering effect is evaluated using the peak signal-to-noise ratio (PSNR) and structural similarity (SSIM), and the segmentation effect is evaluated using the running time and maximum continuous memory block (LCFB) use. The results show that the highest PSNR value of the crack repair trace images processed by the Poisson-noise and bilateral-filtering algorithm is about 35.090 1 dB, and the SSIM value reach about 0.880 1, which shows that adding Poisson-noise improves image quality and optimizes the bilateral filtering effect. The running time of image segmentation by the Otsu algorithm is about 25%-50% shorter than other methods, and meanwhile the LCFB is about 0.25% higher. The processed crack repair trace images achieve the desired effect, which verifies the effectiveness and feasibility of the method proposed.

Abstract:In the construction industry, information sharing by the building information modeling (BIM) generally relies on the industry foundation classes (IFC) schema, but the latter,s unsatisfactory adaptability and inextensibility restrain the former. To overcome the limitations, the Semantic Web was introduced based on the IFC schema to realize heterogeneous data integration and sharing and further achieve semantic-level information delivery. Firstly, the semantic modeling method was introduced through algorithm analysis and model transformation. Secondly, this method was used to create a semantic model for a two-story steel frame building. Finally, the schema of this model was verified for accuracy of the building information delivery and the transferability of the building semantics. The practicability of the semantic modeling method with IfcOWL ontology was supported by the modeling case. The key factors that restrain this semantic modeling method from empowering building information delivery was explored by analyzing the schema of the element of the semantic model. And introductory ideas about redundant information avoidance, domain ontology development, and lightweight semantic modeling were proposed to fill the gap. The SPARQL query case shows that the parsed schema is effective for avoiding redundant information. Consequently, this method has advantages in sharing and integrating multi-source heterogeneous building information and can effectively improve the intelligent level of building information management.

Abstract:At present, the code compliance checking based on BIM model are done manually with heavy workload and low information intelligence. It is of great significance to carry out the research on automatic compliance checking. As an important step, standard translation involves many fields such as philosophy, mathematics, mathematical logic, computer science, artificial intelligence, natural language processing and semantics. It is not only a research issue of interdisciplinary integration, but also a prerequisite for realizing the automation of building design and compliance checking. In the field of structural design, code representation methods which can effectively support knowledge representation, reasoning, and automation of compliance checking needs to be developed. Based on the first-order predicate logic, this paper proposes a method of expression and reasoning of structural design rules by translating the provisions in the “code for design of concrete structures”. Through the definition of “predicate” and “function”, the design provisions, table and formula in design specification are translated. It effectively solves the problem of insufficient and inaccurate representation of structural design clauses by using traditional first order predicates, and thus provides a reference for the computer expression of design specifications.