Detection of Structural Behavior Anomalies in Hybrid Roof Systems
Andrei Chesnokov, Vitalii Mikhailov, Ivan Dolmatov · 2020 2nd International Conference on Control Systems, Mathematical Modeling, Automation and Energy Efficiency (SUMMA) · 2020
Hybrid roofs consist of a frame and flexible polymer membrane. The frame comprises steel beams, hinged struts and high-strength wire-ropes. During a long operational period the roofs are susceptible to adverse effects, which include designed impacts and unexpected influences. The latter ones result in, so-called, anomalous structural behavior, which should be detected as soon as it is possible in order to take corrective measures, preventing severe damage or structural collapse. The problem is aggravated with a huge manifold of eventual anomalies, which is not possible to be directly simulated in the design stage. Autoencoders, being a special type of Artificial Neural Networks, are used to distinguish anomalies from functionally operative behavior of the hybrid roofs. The more the input data deviate from normal structural response, the greater reconstruction error is produced by the Autoencoder. Thus, having received normal parameter set only, the Autoencoder is able to clearly detect anomalous structural response. Such an approach has already been successfully used in cyber- and financial security domains for intrusion and fraud detection, as well as in the field of monitoring of complex engineering systems. Its expansion on the hybrid roof systems contributes to operational reliability enhancement of promising building constructions.