In modern world, information technologies are widely used in the different areas of human activity. The main purpose of such systems is to make the process of work easier, faster and high-quality. Recently, artificial neural networks have been increasingly used to solve practical problems. They have a number of advantages over other computing models: adaptability, learning ability, generalizing ability, and others. These systems are computer programs that perform analysis based on certain baseline data, designed to assist specialists in specific areas of knowledge and achieve significant results. Artificial neural networks occupy a special place in the development of new technologies for structural health monitoring.
A neural network is an artificial system designed to recreate a biological model of nervous systems [1]. In the beginning, it was assumed that the neural network should function like a human brain [2]. A considerable number of neurons and their connections are responsible for maintaining the human body’s unique abilities. The brain can process large flows of information almost instantly when it consists of slow-acting cells. However, the artificial networks had little resemblance to it, and their capabilities were very limited. With the further development of neural network technologies, developers are forced to create artificial networks with properties that are not possible in nature.
The basis of each neural network is simple elements called artificial neurons, which imitate the work of brain neurons [3]. Neural network methods can be used independently or serve as an excellent complement to the standard analysis methods. In the modern world, engineers can use neural networks’ capabilities to define construction’s mechanical behavior, its structural defects, their scale, and extract useful data from available varieties by specific criteria.
The advantage of neural networks over other systems is their ability to learn, based on embedded data. For example, in engineering, specialists that training networks can present a significant amount of data as a various parameters and signs of an object and an estimate based on it. During the neural network’s training process, it finds special parameters hidden from human understanding [4]. This demonstrates the nonlinearity of the network – another feature that distinguishes it among analytical systems. As a result of training, the network can improve its performance over time according to certain rules.
When comparing the neural network approach with statistical methods, some differences can be distinguished. Statistical techniques (including regression analysis, learning algorithms, and others) are widely used in structural health monitoring. However, they require a lot of time and effort and are probabilistic in nature [5]. All these problems, including rapid retraining, can be solved through the use of new neural networks. They can also reduce the dimension of the input data, leaving the most significant. Nevertheless, statistics and neural network technologies should not exclude, but complement each other, especially since there is a powerful software of statistical methods.
The use of unverified technologies, errors in monitoring, and other similar factors increase the risk of emergencies, the consequences of which can lead to human casualties. Moreover, accidents and disasters bring not only severe moral and social shocks but also financial losses due to the inoperability of the object and the need to restore it. For this reason, neural networks can become an indispensable tool for structural health monitoring. It is a compelling modeling method that allows reproducing extremely complex dependencies; in particular, they are non-linear. At the same time, neural networks learn from examples and, because of this, are comfortable in use. Nevertheless, they cannot completely replace the existing methods and work of the specialist, but only complement them.
References
A. El-Shahat, Advanced Applications for Artificial Neural Networks. Croatia: IntechOpen, 2018.
D. Graupe, Principles of Artificial Neural Networks: Basic Designs to Deep Learning. Singapore: World Scientific Publishing Company, 2019.
Gurney, Kevin. An Introduction to Neural Networks. United States, CRC Press, 2018.
O. Avci, O. Abdeljaber, S. Kiranyaz and D. Inman, Structural Health Monitoring with Self-Organizing Maps and Artificial Neural Networks. In: M. Mains and B. Dilworth, ed., Topics in Modal Analysis & Testing. United States, Springer, pp.237-246, 2020.
H. Chen, Structural Health Monitoring of Large Civil Engineering Structures. United Kingdom: Wiley, 2018.