Civil engineering in China has gradually entered a transition period from rapid construction to service maintenance, and the maintenance of in-service infrastructure is becoming the key work. Currently, the safety diagnosis of existing engineering structures based on traditional structural inspection, appraisal and evaluation is difficult to meet the enormous demand for timely safety diagnosis of numerous in-service engineering structures. The evaluation method based on current standards and specifications uses index scoring and comprehensive evaluation and grading considering the importance of indicators, which is challenging to realize the scientific evaluation of the service safety for engineering structures. Therefore, it is necessary to develop new theories and methods for intelligent diagnosis of engineering structures. Intelligent diagnosis includes service information perception, characteristic identification, performance evaluation and evolution prediction. Recently, some achievements have been made in the cross-fusion of structural service safety diagnosis, intelligent sensing technology and artificial intelligence technology. However, an intelligent diagnosis system for the service safety of engineering structures to realize the intelligent evaluation and prediction of the service performance of structures is still facing a series of critical problems. Therefore, based on summarizing the development status of intelligent diagnosis for engineering structures, this paper analyses the major problems of current research and suggested further research ideas and key problems to be solved.

Gaussian mixture model is introduced to analyze the probability distribution of structural effective prestress, and an effective prestress probability estimation method is proposed to solve structural evaluation problems that there is a large number of prestressed tendons in prestressed concrete frame structures, effective prestress presents strong uncertainty and uneven distribution and the sampling detection cannot accurately the overall characteristics of the effective prestress of the structure. The Gaussian mixture model of effective stress of prestressed tendons was firstly established by Monte-Carlo simulation, which was modified by centralization and removal of discrete points to enhance its Gaussian characteristics. The approximate estimation range of the normal distribution was determined based on the Gaussian significance test criterion. Prestressed tendons in this estimation were sampled for testing to estimate the probability distribution of the real-time effective stress of the prestressed tendon assemblage. An engineering example of an unbonded prestressed concrete frame structure was evaluated to verify the effectiveness of this method. The results show that the probability distribution of effective stress of prestressed tendon assemblage in existing structures is composed of several sub-distributions with similar mean and variance, which can be represented by the approximate normal Gaussian mixture model. The inhomogeneity of effective stress distribution caused by design differences can be eliminated by the mathematical method of centralization to improve the accuracy of approximate probability estimation. The upper limit of construction error, main error of tension technology, of existing structures in the construction stage can be estimated by the model combined with the statistical analysis of measured values.

Space structures are the main form of long-span, large-space, and large-area building structures, which are widely used in national infrastructure and urban construction, and their structural safety is of paramount importance. Space structure health monitoring (SSHM) accurately obtains structural responses through sensing technology, reflects the performance state of the structure in real-time, and provides a scientific basis and guidance for the construction, operation, and management decision of space structures. After years of research and engineering practice, wireless-sensor technology based on the internet of things has been developed， which is suitable for large-span space structures. Multi-dimensional data analysis theory has been proposed, and the large-scale applications have been implemented in major projects such as the National Stadium, the National Speed Skating Oval, and the Beijing Daxing International Airport Terminal Building, the research progress has been achieved. By systematically summarizing the current research and applications of SSHM, the characteristics of SSHM are clarified, and the current research situation of the sensing technology is reviewed around the multi-dimensional sensing of load and response, as well as the large-area transmission networks. The monitoring data analysis theories are summarized, including structural load analysis, response analysis, and structural condition assessment. Meanwhile, future research directions are foreseen based on the integration of big data, cloud computing, and artificial intelligence, aiming to further promote the innovation and development of SSHM.

Retrofitting steel structures is not only beneficial for guaranteeing the safety of the structures in services, but also fits the idea of sustainable development. It is of positive significance to scientifically implement the ‘urban renewal’ and promote the achievement of the ‘double carbon’ targets. In terms of steel members, five retrofitting approaches by enlargement of section area, including the welding method, bolt connections or rivet connections, bonded steel plates, bonded fiber-reinforced polymers and composite retrofit, and prestress strengthening method, were summarized in this article. As for the connection and joints, the research advances of retrofitting combining connection with bolts and welds, beam-column joints, bolted spherical joints, welded spherical joints, and tubular joints were illustrated. Meanwhile, the research progress of retrofitting the steel structural system by indirect approaches and repairing techniques for local damage was outlined. On this basis, the representative technical points of GB 51367—2019 ‘Standard for design of strengthening steel structures’ were introduced. In the future, research efforts are suggested to make for developing quantitative design methods of traditional strengthening techniques, retrofitting technique and design theory of steel joints, theory of CFRP retrofitting, application of innovative materials and techniques, and reinforcement mechanism of structural system.

When the field modal test on large and complex engineering structures is performed, it is difficult to quickly implement the signal transmission line between data acquisition equipment and sensors due to the large scale of the structure. Therefore, it is necessary to solve the key problem of accurate time synchronization of distributed acquisition devices in different positions and develop distributed synchronous acquisition and wireless transmission devices that are installed quickly and easily. In this paper, the distributed synchronous acquisition algorithm of structural vibration was proposed based on the ‘Beidou’ satellite timing system. The distributed synchronization acquisition hardware was integrated and the online data acquisition and wireless transmission software were developed to acquire the time synchronization dynamic response of large and complex engineering structures at different spatial positions. Based on the stochastic subspace identification algorithm, the modal parameters of engineering structures in the service state were automatically identified. During the investigation of vibration events of Saige building, the acceleration response of the 69th floor and the bottom of the mast of the Saige building were successfully captured by the distributed synchronous acquisition system when the structure resonance occurred from 12:00 to 13:00 on 20th, May. It is observed that the resonance vibration is mainly dominated by a 2.12 Hz signal. Based on the field modal test performed by this system under ambient vibration excitation, the first 19-order modal parameters of the structure were quickly and accurately identified. It is found that the vibration mode of 2.12 Hz represents the flexural torsional coupling mode of the main structure and the in-plane symmetrical vibration mode of the mast. Based on the field excitation test, the damping ratio of 2.12 Hz decreases suddenly and then gradually increases with the increase of the vibration amplitude. The lower damping ratio is one of the reasons for the Saige building vibration event.

In-depth mining of monitoring data information based on structural health monitoring is an important means to obtain the health status of heritage buildings and ensure their durability and safety. In order to accurately carry out data analysis and structural safety assessment, as well as to distinguish the abnormal hardware, artificial disturbance or environmental mutation two kinds of data anomaly genesis, respectively define two types of data anomalies according to different causes, this paper defines two kinds of data anomalies of heritage building masonry structures. According to the characteristics of long-term static and slow change of monitoring data of heritage building masonry structures, an improved outlier identification algorithm based on density is proposed to improve the quality of monitoring data cleaning. Through compression and segmentation of time series data and transformation of series data, local outliers in data can be accurately picked up. Through qualitative and quantitative analysis, the improved outlier recognition algorithm based on local outlier density has high accuracy and efficiency in identifying local outliers of monitoring data of heritage building masonry structures, which can solve the problem that the existing data outlier identification algorithm is not suitable.

With the gradual transformation of the civil engineering industry of China from construction to operation and maintenance, the demand for service safety assurance of engineering structures has increased sharply, and structural intelligent diagnosis methods that improve quality and efficiency have become a research hotspot. Structural service behavior indicator is the element to characterize the safety level of engineering structure and is the basis of structural maintenance systems and structural health monitoring research. Sensitive indicators for evaluating the service behavior of structures and further realizing the intelligent identification of indicators are the primary tasks of intelligent structural diagnosis. This paper focused on the intelligent identification of sensitive service indicators in various scenarios such as engineering structure operation and maintenance of public buildings, subway tunnels, highway bridges, and highway pavements. This paper sorted out key sensitive indicators, and further summarized the intelligent identification methods of indicators. The results show that the new generation of artificial intelligence technology represented by deep learning has effectively promoted the research and application of perceptual identification of structural service-sensitive indicators, and the digital image method and deep learning algorithm have achieved comprehensive application advantages in the intelligent detection of structural deformation and surface diseases.