PSI - Issue 64

Nicola Nisticò et al. / Procedia Structural Integrity 64 (2024) 2230–2237 Nicola Nisticò/ Structural Integrity Procedia 00 (2019) 000–000

2235

6

4. The role of Artificial Intelligence Geometrical and mechanical models stand to benefit significantly from the integration of Artificial Intelligence (AI). The evolution of AI can be traced back to two pivotal milestones: 1) the development of Expert Systems (ES) (Lindsay et al., 1993; Buchanan and Shortliffe, 1984), which are rule-based systems incorporating knowledge defined by human experts. They utilize these rules to infer new data and draw conclusions. While ES contributed to AI development, they are typically considered distinct from it (Kaplan and Haenlein, 2019). Nonetheless, they remain powerful tools, exemplified by IBM's Deep Blue, which defeated world chess champion Garry Kasparov in the 1990s; 2) Machine Learning and Neural Networks (NNs) that when provided with a dataset as inputs, evaluate an output based on a set of tailored relationships for the specific task. Various types of NNs exist, including unsupervised (UNN), supervised (SNN), semi-supervised (SSNN), and reinforced (RSNN). UNNs, pioneered by Hebb (1949), utilize cluster analysis to group data categories and do not require pre-labeled data as inputs. SNNs depend on labeled input-output sets (training data) to infer the input-output function. Semi-supervised strategies blend elements of both UNNs and SNNs, accommodating partially labeled data. Reinforced strategies (Zhang et al., 2020) do not mandate labeled data for either input or output. Convolutional Neural Networks (CNNs), originally proposed by Fukushima (1980), feature hidden convolutional layers loosely connected to other neurons. They excel in computer vision applications (Aloysius and Geetha, 2017), often applied in image classification and real-time object detection, proving valuable for automatically recognizing features such as cracks in existing structures. Finite Element Method (FEM) approaches for engineering applications incorporate intriguing machine learning techniques to emulate human reasoning through Mechanistic Data Science (MDS). Recurrent Neural Networks (Lipton et al., 2015) have been suggested (Ghavamiana and Simone, 2019) for gathering material mechanical properties oriented towards stress-strain relationships. Zhang et al. (2021) proposed Deep Neural Networks (DNNs) for defining shape functions: leveraging the hierarchical structure of DNNs, the approach was termed Hierarchical Deep-learning Neural Network (HiDeNN). Lastly, the method presented by Zhu et al. (2019) deserves mention for modelling uncertainty. Multi-disciplinarity vs Inter-disciplinarity and trans-disciplinarity Both interdisciplinary and transdisciplinary (Darbellay, 2016) approaches are embraced, with a central focus on leveraging technology to yield future social benefits. One key objective is the democratization of cultural heritage, ensuring that Europe's rich history is accessible to future generations. By amalgamating future studies, creativity studies, and design thinking, the project generates innovative scenarios where business, education, and technology solutions converge, resulting in tangible societal gains such as enhanced well-being, social cohesion, inclusion, and a strengthened sense of belonging. Situated within the Technologies for Cultural Heritage domain, the project extends its reach to sub-domains like Cultural Industries, Creative Industries, and Historical-Artistic-Architectural Heritage Management. It aligns closely with international priorities aimed at preserving, managing, and enriching cultural, artistic, and landscape heritage, as well as fostering the development of tourism and cultural enjoyment systems. Transdisciplinarity, as a framework, seamlessly integrates natural, social, and health sciences within a humanities context, transcending traditional disciplinary boundaries. Complex real-world issues are addressed by offering diverse perspectives, formulating comprehensive research questions, establishing consensus definitions and guidelines, and providing extensive support services. Acknowledging that multidisciplinary teamwork presents both advantages and challenges, the primary hurdle for the project lies in attaining the requisite level of transdisciplinarity, especially within the humanities context. In terms of target markets, the project serves to strengthen connections between tourism, cultural heritage, the educational sector, and the creative industry, thereby contributing to the development of regions with significant potential for attraction within the local territory. 5. Conclusions The presented work aims to define, based on the current state of the art, the essential needs for implementing a methodology to develop and deploy innovative tools and techniques for digitizing and digitalizing both visible and non-visible attributes of cultural heritage items. This involves identifying key technological requirements and methodological frameworks necessary for effective digitization, which converts physical attributes into digital form,