FINITE ELEMENT AND RELIABILITY STUDY OF IN-PLANE BEHAVIOUR OF CONCRETE MASONRY INFILLED RC FRAMES

Abstract

This research provides a finite element and reliability study of the in-plane behaviour of masonry infilled reinforced concrete frames. In the first phase of the research, a reliability analysis was conducted to advance the understanding of the effect of the randomness and uncertainty of masonry materials on the lateral resistance of masonry infill walls from a probabilistic approach. A computationally efficient and robust finite element model was developed in OpenSees and validated using the experimental results obtained in the same research group. Coupled with the Monte-Carlo simulation techniques, the model was used to estimate the failure probability of infilled frames with a spatially varying random field of masonry compressive strength. After 1000 simulations, the resulted failure probability distribution of the infilled frames led to a new recommendation on the masonry resistance reduction factor. Despite the efficiency of the OpenSees model, running simulations of the random fields on the infilled frames on an advanced work station, with 40 Central Processing Units (CPUs) and 128 GBs of memory, took three months to complete. It became evident that a better-performing computing technology was needed for any meaningful reliability analysis of this magnitude. This motivated the second phase of the research. In the second phase, an open-source and modular finite element library for estimating the behaviour of infilled frames was developed. The main feature of the library was to be able to significantly accelerate the numerical simulations for structural applications in general and with a focus on masonry infilled RC frames. The specific algorithms, encoded in C++, were developed for the library to be able to run on either central processing units or graphical processing units (GPUs). The library adopted an advanced smeared crack modelling technique, the Distressed Stress Field Method (DSFM), for modelling of the masonry infilled RC frames. A comparison with the experimental results showed that the DSFM was successfully incorporated in the analysis of masonry infilled RC frames. The performance of the developed library in accelerating the model run speed was demonstrated through comparisons of run speed using a CPU and several commercially available GPUs. It was shown that GPU devices with adequate memory space can lead to significant model run speedup compared with a CPU. The degree of speed-up was highly dependent on the number of elements used in the finite element model and the number of processing core available in the parallel architecture. The greater the number of elements used in the model, the greater rate of acceleration will be achieved.