Data-model hybrid-driven and artificial intelligence-based monitoring threshold update and short-term response prediction for high-formwork support system

Abstract

In recent years, the frequency of collapse accidents involving ultra-high reinforced concrete formwork support systems has increased, highlighting the need for real-time monitoring and timely early warnings during construction. At present, there are no standardized methods for determining monitoring thresholds for high-formwork support systems, and existing monitoring systems lack trend prediction capabilities, limiting their effectiveness in advance prediction and early warning of structural responses. This paper aims to propose a data-model hybrid-driven framework for determining and updating monitoring thresholds, as well as for short-term response prediction of high-formwork support systems. The proposed framework consists of three key modules. The first module is a visual monitoring system for detecting structural displacement, which collects data, issues real-time warnings, and continuously feeds information to the second module. The second module focuses on determining and updating monitoring thresholds through a data-model hybrid approach and supplies extensive training samples for the third module. The third module concentrates on short-term prediction of structural displacement response and construction load inversion using the CNN-BiLSTM-Adaboost algorithm. This algorithm predicts displacement trends in high-formwork support structures up to 1 h in advance, providing early warnings. Additionally, it quickly and accurately estimates the upper construction load, enabling precise emergency measures and reducing the impact of adverse load effects.