A hybrid AI framework for engineering risk management under data scarcity and high uncertainty

In engineering environments characterized by uncertainty and a lack of structural information, traditional risk management is ineffective. However, it remains unknown how to transform qualitative tacit knowledge into reliable predictive models when stable statistical distributions are inaccessible. This study addresses this gap by developing a hybrid artificial intelligence system that bridges the divide between unstructured experiences and digital decision-making. By integrating Case-Based Reasoning (CBR) with K-medoids clustering and Multiple Correspondence Analysis, we created a framework to classify and reuse qualitative knowledge assets. We evaluated this model across four industrial case studies in volatile construction supply chains involving 212 expert participants. The analysis showed that the model achieved a 72.92% agreement rate with expert judgement in real-world settings. Beyond simple accuracy, the framework maintained robust precision and recall metrics of 0.80, successfully identifying critical risks, such as information security gaps and bureaucratic bottlenecks, that human project managers frequently overlooked. Additionally, validation via the Analytic Hierarchy Process (AHP) confirmed the practical efficiency of the suggested strategies, with a performance rating exceeding 4.25 out of 5. Ultimately, this methodology establishes a generalizable foundation for risk management in data-sparse industries and offers a practical path toward improving project resilience in unstable economies.