Escalating Human–Wildlife Conflict (HWC), particularly involving protected large carnivores such as the wolf, poses a significant challenge in Europe. This problem, exacerbated by ecological pressure, necessitates the development of innovative, non-lethal, and effective prevention methods that overcome the limitations of current passive solutions, such as habituation. This article presents the design and implementation of a prototype for an autonomous, multi-sensory preventive system. Its three-layer architecture is based on a decentralized network of sensory-deterrent nodes that utilize Edge AI for real-time species detection and adaptive selection of deterrent stimuli. During field validation, the prototype’s biological efficacy as a proof-of-concept was confirmed in a crop protection scenario against the European roe deer (Capreolus capreolus). The system’s deployment led to a near-total elimination of damages. The paper also presents key technical performance metrics (e.g., response time, energy consumption) and the accuracy of the implemented AI detection model, verified using both field and historical data. The positive test results demonstrate that the developed platform provides an effective and flexible foundation for preventive systems. Its successful validation on a common herbivore species represents a crucial, measurable step toward the target implementation and further research on the system’s effectiveness in providing protection against large carnivores.

Development of the cellular automata (CA) sphere packing algorithm dedicated to the generation of two- and three-dimensional digital, synthetic microstructure models with heterogenous grain size distribution is presented within the paper. The synthetic microstructure model is generated in four major steps: generation of 2D/3D cellular automata computational domain, generation of circles/spheres with a required size distribution, close-packed filling of the computational domain with generated circles/spheres, growth of the circles/spheres according to the unconstrained CA growth algorithm. As a result, synthetic microstructure models with specific, required grain size distribution described by e.g. uni- or bimodal one are obtained. To reduce computational complexity and decrease execution time, the rotation of the circles/spheres during the packing stage is based on the vector accounting for the distance from computational domain borders and other spheres. The CA grain growth algorithm is also implemented using threads mechanism allowing parallel execution of computations to increase its efficiency. The developed algorithm with the implementation details as well as a set of examples of obtained results are presented within the paper.

The research’s primary goal is to identify the heat source and thermal material model parameters for the numerical simulation of the laser engineered net shaping (LENS). Inconel 718 was selected as a case study for the current investigation. The LENS process’s numerical model was developed within commercial finite element software and was used as a direct problem model during the parameter identification stage. Experimental data were obtained based on a rectangular-shaped sample with thermocouples located under the based material surface. The recorded thermal profiles were used to establish a goal function for the parameter identification stage. As a result, parameters describing the melt pool geometry during the additive manufacturing, as well as thermal coefficients describing interactions between the sample material and surrounding/base material, were determined.