EVALUATING MECHANICAL PROPERTIES OF UNFIRED CLAY STRUCTURES USING ARTIFICIAL NEURAL NETWORKS (ANN) FOR HERITAGE CONSERVATION

Conservation of historical and vernacular architecture often involves the use of traditional materials such as unfired clay, which require precise mechanical characterization for effective preservation strategies. Experimental analysis for determining the compressive and flexural strengths of these materials can be time-consuming and costly. To address this, the present study aims to streamline the process by leveraging artificial neural networks (ANN). Two ANNs were developed and trained using experimental data from laboratory tests on unfired clay matrices. The trained models provided accurate predictions of mechanical properties, achieving an error rate of less than 1% for test values. These results demonstrate the potential of ANNs as efficient tools for predicting the mechanical behavior of unfired clay, offering significant time and resource savings in the conservation field. This approach enables more effective preservation and restoration of structures that utilize unfired clay, supporting efforts to maintain architectural heritage.