(1) What are the benefits of the application of structural health monitoring to life safety and economic benefits?
(2) How to define the structural damage, the possibility of multiple damages occurring at the same time, and which type is the most worthy of attention?
(3) Under what conditions (different uses, different environments) do we need to monitor the system?
(4) Limitations of data collection during use.
The use of environment limits the completion of monitoring system and monitoring process. This assessment begins to link the process of damage identification with the external characteristics of damage, and of course, it also uses unique damage characteristics to complete the detection. The data acquisition part of structural health monitoring involves the selection of excitation mode, sensor type, number and arrangement, as well as data acquisition, storage and transmission equipment. Economic benefit is an important reference factor for selecting schemes, and sampling period is another factor that cannot be ignored. Because data can be obtained in a changing environment, the ability to normalize these data becomes very important in the process of damage identification. When applied to structural health monitoring, data normalization is to separate the inaccurate values measured by sensors due to environment or operation. The most common method is to normalize the measured response by measuring the input parameters. When the impact of environment or operation is significant, we need to compare the data of similar time periods or corresponding operation cycles. It is necessary to identify the source of data instability and minimize its impact on system monitoring. Generally speaking, not all the influencing factors can be eliminated. Therefore, it is necessary for us to take appropriate measures to ensure the impact of these indelible factors on the monitoring system. Changing environmental factors, test conditions and test discontinuity will aggravate the instability of data.
Data purification is the process of filtering some valuable data to complete transmission, which is opposite to the process of feature extraction. Data purification is largely based on personal experience in data collection. For example, by checking the installation of test equipment, it may be found that the consolidation of a sensor has been loosened, so according to personal experience, this set of data obtained or the data measured by a specific sensor can be deleted during data processing. Data processing techniques, such as filtering and reconstruction, are also good methods for data purification.
In a word, the technology of data acquisition, normalization and purification in the process of structural health monitoring is constantly improving. Further understanding of the feature extraction process and continuous improvement of the data model will contribute to the progress of data acquisition technology. In the field of structural health monitoring, the most concerned is how to distinguish damaged structures from intact structures through data characteristics. Data compression is included in this feature selection process, and the most effective feature for damage identification is still.
The corresponding quantity (such as vibration spectrum or frequency measured in the field) based on the relevant test system is one of the most commonly used features. Another damage identification method is to find the factors that are sensitive to specific damage, that is, the damage of the structural system in a specific environment corresponds to the original definition of a parameter. This damage simulation system is a very effective tool. The application of analytical tools also plays a very important role, such as the finite element model verified by experiments. Analysis tools are usually used to carry out numerical simulation tests and simulate the damage of real structures through computer settings. The cumulative damage test obtained by observing the aging of key components of the structural system under load can also be used to identify some damages. This process includes accelerated damage testing, fatigue testing, corrosion and the accumulation of some types of damage caused by temperature cycling. The organic combination of the above various types of analysis and experimental research or various research methods can deepen the understanding of some damage characteristics. It is the least part of the literature in the field of structural health monitoring to identify whether there is damage in the structure through statistical models. Statistical models focus on the use of algorithms to evaluate the damage state of structures. The algorithms used in statistical models are usually divided into three types: when the data of complete structures and damaged structures can be obtained, pattern recognition algorithms usually use global classification related to reference studies, and global classification and regression analysis belong to the category of reference studies; Non-reference research refers to the lack of data of damaged structures; New detection technology (or quoting more mature technology from other industries) is the basic algorithm applied to non-reference research. All algorithms (analysis and statistics or purification and optimization) promote the improvement of damage identification technology.