Developing An Efficient Deep Learning Model For Anomaly Detection For Monitoring The Structural Health Of Buildings

Authors

  • Ujjwal Thakur, Rajan Kumar Singh, Sanjeev Ashatkar

DOI:

https://doi.org/10.17762/msea.v71i4.1777

Abstract

Due to their ability to maintain the structural health and discover deterioration to important civil constructions including buildings, flyovers, and pipelines, contemporary structural health monitoring methods are increasingly used in civil construction. While significant advancements have been made in improving abnormality recognition to maintain public safety, algorithms that could be utilized for long-term supervision on inexpensive technology remain still a community-wide unanswered question. Owing to the unbalanced design of the parts, structural response identification is a difficult task. Consequently, Gabor filters provide a benefit for one-stage detection. To substitute the lowest convolution layers that make up the SSD's architecture, generalized Gabor filters are developed. Using a deep neural network and SSD detectors, a technique for autonomously identifying and removing anomalous information is suggested in this study. Initially, a time series classifying issue is used to simulate the anomalous detecting issue. The initial sequences are processed using data pre-processing and enrichment, comprising information enlargement and down-sampling to create new instances. Information extension techniques such as symmetric switching, noise additions and arbitrarily growing outliers are employed for a limited amount of instances in the information collection, and images that have the same label are contributed without expanding the initial records. The optimum value and lowest mean are simultaneously extracted symmetrical in order to minimise the complexity of the intake samples while keeping the majority of the information's features. CNN is better at handling uneven training dataset when the categorization parameters are hyper variable tuned. The anomalous identification of accelerating statistics on a long-span building is the last demonstration of the efficiency of the suggested technique. The suggested method may successfully identify numerous anomalous behaviour for the anomalous recognition problems that is represented as a standard statistical classifications task.

Downloads

Published

2023-01-21

How to Cite

Ujjwal Thakur, Rajan Kumar Singh, Sanjeev Ashatkar. (2023). Developing An Efficient Deep Learning Model For Anomaly Detection For Monitoring The Structural Health Of Buildings. Mathematical Statistician and Engineering Applications, 71(4), 9724 –. https://doi.org/10.17762/msea.v71i4.1777

Issue

Section

Articles