The Red Mansion at No. 106 Huangpu Road， Shanghai， was constructed in 1911. It is a three-story brick-wood structure. In its comprehensive protective renovation project， three major challenges were encountered： the simultaneous construction of a complex group of adjacent deep foundation pits， the overall ultra-high jacking by 6.55 m to restore its historical appearance， and the synchronous development of underground space together with the improvement of the building’s seismic performance. To address these challenges， a complete set of key techniques was proposed， integrating underpinning， active deformation control， synchronous high-position jacking， and seismic isolation with story addition. Using an adjustable active underpinning system， the cumulative additional deformation of the Red Mansion during the construction of the surrounding deep foundation pits was controlled within ±5 mm. A relay lifting technique combining "lifting + jacking" was adopted to achieve the ultra-high jacking of 6.55 m， along with the simultaneous high-position rectification of 300 mm. The addition of a seismic isolation layer significantly enhanced the structural seismic performance. After three rounds of whole-process load transfer， the maximum inclination ratio of the building foundation was reduced from the initial 9.12‰ to below 2.5‰， with no new structural cracks generated. This project realized the in-situ preservation， jacking with story addition， and functional upgrading of an outstanding historical building in Shanghai under the condition of simultaneous construction adjacent to a group of deep foundation pits， setting a record for the highest overall jacking height of modern outstanding historical buildings in China.

To address the problem of missing multi-source monitoring data of offshore wind turbines caused by sensor failures or communication interruptions under harsh operating conditions， this paper proposes a novel imputation model based on a multi-head gated residual network. This method achieves collaborative fusion of supervisory control and data acquisition （SCADA） data and structural vibration monitoring data through feature concatenation， employs a gated residual network to extract deep nonlinear coupling features， and uses a multi-head parallel output architecture for the independent reconstruction of these two heterogeneous data types. During the training stage， a dynamic masking mechanism combined with a hybrid loss function is adopted to enhance the model’s adaptability to complex aerodynamic operating conditions. Validated with field data from a 10 MW offshore wind turbine， the proposed model achieved a high coefficient of determination under training conditions， enabling accurate reconstruction of missing multi-source data. In generalization tests for non-training periods， although the coefficient of determination of the model’s predictions fluctuated slightly， the model still effectively captured the overall trends of monitoring signals. Notably， the degradation in generalization performance for vibration data was less pronounced than that for SCADA data， demonstrating its greater stability. The proposed method can significantly improve the completeness and reliability of multi-source monitoring data for wind turbines and holds considerable potential for engineering applications.

Aiming at the problems of low efficiency， strong reliance on manual labor， high risk of high-altitude work， and secondary damage that is easily caused by contact detection in traditional methods for building exterior wall disease detection， this paper proposes an intelligent non-destructive detection method based on machine vision and deep learning. This method enables rapid identification of three types of apparent diseases： spalling， hollowing， and cracking. Using UAV high-precision collection equipment， disease images were collected from typical exterior wall types such as tiles， paint， and cement mortar. A building exterior wall disease image database containing 1018 images of three types of diseases was constructed. Through LabelMe software， disease annotation was performed， forming 1168 spalling labels， 1619 hollowing labels， and 1515 cracking labels. Based on the deep learning YOLO11n model， multiple training schemes were implemented on the training set. This study found that， with 300 training epochs， an image size of 1280 pixels， and data augmentation enabled， a detection performance of mAP50 = 0.753 was achieved. This model relatively accurately identified the three types of apparent diseases： spalling， hollowing， and cracking. Finally， engineering instance applications were carried out in multiple residential communities in the Chengdu area， further proving that the model has good generalization ability and can provide a new technology for non-destructive rapid detection of building exterior wall diseases.

Many existing rural houses face structural safety hazards due to material performance degradation. This study investigated the residual bearing capacity and the strengthening effect of carbon fiber reinforced polymer （CFRP） sheets on old precast prestressed concrete （PC） hollow-core slabs， using specimens obtained from a demolished 45-year-old rural house in Shanghai. The test involved two PC slabs： one served as the control specimen， and the other was strengthened with CFRP sheets. Static loading tests were conducted to compare and analyze their failure modes， deformation capacity， and energy dissipation capacity. Based on the measured data， the strain development pattern of the CFRP sheets was analyzed. The results showed that the 45-year-old PC hollow-core slabs still had a certain bearing capacity but with a low safety margin， exhibiting a typical brittle flexural failure due to under-reinforcement. After CFRP strengthening， the failure mode changed to shear failure with obvious ductile characteristics， because the CFRP sheets bore the main tensile stress in the later loading stage.

Local deformation is a common damage to steel structures， and 3D reconstruction is an important approach to evaluate the bearing capacity of locally deformed angle steels. To accurately evaluate the accuracy of the SfM （Structure from Motion）-MVS （Multi-View Stereo） 3D reconstruction algorithm in establishing point cloud models of locally deformed steel components， this study proposed an accuracy evaluation method for angle steel point cloud models based on spatial Euclidean distance. Four types of locally deformed angle steels with two thicknesses were selected as the research objects， and an accuracy verification test was carried out on the digital image models of locally deformed steel components. The iterative closest point algorithm was used to align the test point cloud with the reference point cloud in space， and the accuracy of the point cloud model was quantified according to the Euclidean distance between corresponding points， so as to verify the accuracy of the angle steel point cloud model from a 3D perspective. The results showed that through the accuracy evaluation of the angle steel point cloud model， in the height and width directions， the points within the allowable range of dimensional deviation in the 3D point cloud models of the four types of locally deformed angle steels accounted for approximately 98% of the total number of point clouds in the models； in the thickness direction， the average error of the 5-mm-thick angle steel point cloud model was approximately 1.09 mm， while the average error of the 10-mm-thick angle steel point cloud model was approximately 1.02 mm； in the local deformation area， the average error of the four angle steel point cloud models was approximately 1.00 mm.

Hyperbolic cooling tower shells are mostly cast-in-place reinforced concrete thin-walled structures. Their high-altitude construction poses significant challenges， and geometric imperfections often occur due to issues in construction layout accuracy. Focusing on a specific engineering case， this study employed terrestrial laser scanning （TLS） technology to capture precise geometric imperfection data of the tower shell. Based on the scanned data， finite element models of the hyperbolic cooling tower， both with and without geometric imperfections， were developed using ABAQUS. The effects of geometric imperfections on the mechanical properties of the cooling tower， as well as the sensitivity of different load effects to these imperfections， were systematically investigated. The results indicated that when the actual imperfection magnitude was introduced based on the measured distribution pattern， the bearing capacity and crack resistance of the tower shell decreased significantly. Moreover， no deterioration in mechanical properties was observed when the imperfection magnitude remained below 150 mm. The effects of dead load and external wind pressure were highly sensitive to geometric imperfections， whereas temperature effects remained almost unaffected. Furthermore， the locations along the meridian lines of the tower shell where the maximum external wind suction occurs were identified as critical regions for safety assessment.