ADINA simulation model of a concrete structure can be adapted based on real-world seismic IoT data to produce a more accurate digital twin for simulation so users can get a more accurate and robust design, or expected structure seismic performance. 

A research project entitled “Integration of Physics-Based Building Model and Sensor Data to Develop an Adaptive Digital Twin” was recently supported by Bentley Systems. A poster regarding the project was published at Buildsys 2022 by Barney H. Miao, Yiwen Dong, Zheng Y. Wu, Bulent N. Alemdar, Pei Zhang, Monica D. Kohler and Hae Young Noh who acknowledge the technical support provided by Bentley. A quick overview of project is presented here.

Introduction

Structural health monitoring is an essential tool for providing rapid assessments of damages and thus decisions about the safety of a structure. Traditional structural models were created using physics-based models that are based on the structure's design of a new structure. However, wear and deterioration over time cause the structure and the member properties to change. Thus, a new adaptive digital twin, utilizing physics-based model and real-time sensor data to adapt the model to its current state, is introduced.

The digital twin was developed by finding the relationship between the physics-based models and sensor data through transfer functions. In the project, the physics-based models were developed using linear dynamic analysis software ADINA, from which the acceleration responses can be computed. The sensor data was obtained from Community Seismic Network, which provides the measured responses for specific floors of a structure. Figure 1 shows the overall framework of the digital twin proposed.

Adaptive digital twin development using physics-based model and sensor data

The process adopted consisted of three steps entitled:

1. Development of Physics-Based Structural Model,
2. Sensor Data Acquisition, and
3. Adaptive.

In the following lines, the above three steps are briefly explained.

Development of Physics-Based Structural Model

The physics-based structural model is based on structural and architectural drawings and was developed using ADINA (Figure 2).

Sensor Data Acquisition

Community Seismic Network (CSN) is a low-cost dense seismic strong-motion network, that records waveform data for earthquakes at the ground and upper levels of buildings. The waveforms are accelerations sampled at 50 samples per second with a maximum range of ±2g (Figure 3).

Adaptive Digital Twin Development

The adaptive digital twin was created by identifying the transfer functions for the building. Since it is effective at generating approximations for the initial testing of the approach, linearity in the transfer functions was assumed the parameters of the transfer functions were obtained based on ratios in the frequency domain of the sensor and model data. In addition, the output of the physics-based model from a different test event is scaled, making the model's response more in line with the actual responses of the structure. At the end, the adaptive digital twin was evaluated by using a nine-story instrumented building, Caltech Hall (Figure 4).

Evaluation

The transfer functions were computed using the Cabazon and Trabuco earthquakes and evaluated them using Ridgecrest earthquake. The location estimation and amplitude errors of the dominant frequency peaks decreased by 5.59%-55.97% and 2854.5%-33397.4%, respectively.

Conclusions

Some interesting findings of the work are presented here. Initially, to create an adaptive digital twin, physics-based models and sensor data were integrated. The digital twin helped the authors to increase the location and amplitudes of the dominant frequency peaks by up to 33397.4%, enabling adaptive updating in structural modeling and design. In summary, the authors intend to enhance the digital twin in future works to incorporate loss assessments and damage indications.

Learn more about ADINA

Watch below a quick video overview of the research project: