Bridging Innovation and Application: Advancing Artificial Intelligence in Engineering Systems

1. Introduction

This Special Issue, titled, Artificial Intelligence for Engineering Applications, presents a curated selection of the recent advancements at the intersection of Artificial Intelligence (AI) and engineering. As AI continues to transform traditional engineering domains—from civil and mechanical to electrical and chemical engineering—the need for domain-specific, interpretable, and scalable AI solutions has become increasingly evident. While significant progress has been made in areas such as predictive maintenance, structural health monitoring, intelligent control systems, and the optimization of the design and manufacturing processes, critical gaps remain. These include the lack of standardized benchmarks, limited generalizability of models across diverse engineering contexts, and challenges in integrating AI with physics-based approaches and real-time systems. This Special Issue addresses these challenges by combining cutting-edge research that bridges theoretical development with practical implementation. Contributions span novel algorithms, hybrid modeling techniques, AI-integrated simulation tools, and case studies demonstrating real-world impact. A strong emphasis is placed on AI’s interpretability, robustness, and trustworthiness in safety-critical engineering applications. Moreover, the issue identifies key directions for future research, briefly outlined as follows: developing physics-informed AI models, advancing human-AI collaboration in engineering workflows, enhancing model transparency and explainability, and establishing unified frameworks for validation and deployment in operational settings. This Special Issue aims to catalyze future innovation and collaboration across disciplines by highlighting outstanding achievements and challenges.

2. Harnessing Brainwaves: Deep Learning for EEG-Based Motor Control

Recent developments in brain–computer interface (BCI) technologies have significantly broadened the scope of human–machine interaction by enabling the direct translation of neural activity into control commands for external devices [1]. Among the most promising non-invasive techniques for such applications is electroencephalography (EEG), which captures electrical activity from the scalp and provides a practical, low-cost solution for real-time signal acquisition [2].

The implications of this research are substantial, particularly in designing next-generation BCI systems for individuals with motor impairments and for developing more intuitive and adaptive robotic control mechanisms [3]. Notably, some datasets in the public domain can help train new models [4]. Moreover, the study highlights the potential of deep learning-based EEG decoding as a reliable and scalable solution for real-time neural interfacing, opening new avenues for interdisciplinary research across neuroscience, artificial intelligence, and robotics [5].

Looking toward the future, the field of EEG-based BCI technology is poised for significant advancement across multiple dimensions. The integration of more sophisticated machine learning techniques, including federated learning approaches that can leverage data from multiple institutions while preserving privacy, should enable the development of more robust and generalizable models. Additionally, the incorporation of explainable AI techniques will be crucial in understanding the decision-making processes of these complex models, being particularly important for medical and assistive technology applications where transparency and reliability are paramount.

As the field progresses, addressing the current limitations identified in this study is crucial. This includes expanding the range of motor imagery tasks currently evaluated, developing more comprehensive validation techniques beyond five-fold cross-validation, and creating systems that can effectively handle out-of-distribution data. The future of EEG-based BCI technology promises to deliver more intuitive, reliable, and accessible solutions that have the potential to significantly enhance the quality of life for individuals with disabilities, while opening new possibilities for human–computer interaction across diverse applications [4].

3. Implications for Diabetic Care and Future Research Directions

A key strength of the artificial intelligence model developed lies in its holistic consideration of both advantageous and potentially detrimental dietary components. In addition to maximizing the inclusion of essential nutrients—such as vitamins, minerals, and proteins—the model simultaneously minimizes the presence of risk-associated elements, including sodium, cholesterol, and saturated fats [6]. Furthermore, glycemic load is treated as a primary constraint in the optimization process, ensuring that the recommended diets contribute meaningfully to blood glucose regulation while supporting overall metabolic and cardiovascular health [7].

This comprehensive and data-driven approach to dietary planning demonstrates the potential of AI-assisted tools in improving disease-specific nutrition strategies. It underscores the value of hybrid computational models in addressing complex health-related challenges.

Future developments will likely see the deeper integration of quantum computing principles in nutritional optimization. The enhanced capabilities of quantum-behaved particle swarm optimization (QPSO) and its variants such as GQPSO demonstrate superior performance in minimizing glycemic loads compared to traditional methods. This quantum advantage suggests a promising direction for developing more sophisticated algorithms that can handle the complex, multi-dimensional nature of nutritional optimization problems.

Due to the variability regarding foods’ nutritional values, it is often difficult to estimate them accurately; this complicates data collection and decision making regarding the results. Future research will probably focus on improving data accuracy through advanced sensing technologies, comprehensive food databases, and machine learning models that can predict nutritional values with greater precision, enabling more reliable optimization outcomes.

4. Advanced Cotton Boll Detection: A Breakthrough in Agricultural Technology

In the area of technology applied to agriculture, significant progress has been found in cotton cultivation applications [8]. It suggests models for handling complex field conditions, overlapping bolls, and small objects that make it particularly valuable for precision agriculture. By providing accurate, automated cotton boll counting, this technology can help farmers optimize fertilization, pest control, and harvest timing, ultimately improving crop yield and quality while reducing resource usage [9].

This work has demonstrated relevance in addressing challenges and fostering and creating new lines of research. The challenge of separating overlapped cotton bolls is an important area of research. Future models will need to incorporate advanced 3D imaging capabilities and sophisticated spatial reasoning algorithms to address the occlusion problems that are common in dense cotton fields.

The integration of cotton boll detection with climate monitoring systems should enable the development of adaptive agricultural strategies that can respond to changing environmental conditions. As climate change continues to affect agricultural regions worldwide, the ability to rapidly assess and respond to crop conditions will become increasingly important for maintaining food security and economic stability.

The future success of automated cotton boll detection technologies will depend significantly on the development of international standards and protocols that ensure interoperability across different systems and regions. As countries like China and India continue to dominate global cotton production, the adoption of standardized detection technologies will facilitate international trade and quality assurance processes.

5. Exploring a New Approach: PSO Dynamic Model vs. HOMER for Hybrid PV–Hydrogen Energy Systems

In recent years, energy generation with alternative sources to hydrocarbons has been a topic of considerable interest to the scientific community. Software, such as HOMER, has been of great help in managing and optimizing the use of energy generated from various sources. Consequently, several studies have proposed algorithms focused on improving this energy management, such as particle swarm optimization algorithms [10,11].

The strategy of proposing alternatives to commercial software has positively diversified research. This has demonstrated the possibility of resizing the grid to accommodate different energy generation sources. Therefore, future work has chosen to focus on conducting in-depth studies to address energy demand in different areas using optimization and bio-inspired algorithms [12,13].

The integration of artificial intelligence models also has a significant impact on the development of hybrid optimization models. Trends show that these models help make energy distribution more efficient in locations where demand is substantially exceeded [14].

The current challenges in this area motivate researchers to correctly size energy generation systems that integrate renewable sources, such as hydrogen. This helps meet demand and strengthen the distribution network without generating high costs. Having a power generation system integrating hydrogen also boosts an area’s economy by monetizing the remaining generation through microgrids, facilitating improved energy distribution efficiency. In the long term, it will be possible to see cost reductions for such systems by integrating them across different areas, improving the energy infrastructure of the society.

6. Insights into Advanced Technologies and Machine Learning Solutions for Passenger Counting in Public Transport

The proposed implementation framework addresses the full lifecycle of an APC system—from camera placement and network infrastructure to data collection and model training. This holistic approach acknowledges that successful deployment requires more than selecting the right technology; it demands a careful consideration of the installation, data management, and continuous improvement processes [15].

Future systems will move beyond simple counting to predictive passenger flow modeling. Machine learning algorithms will analyze historical patterns, weather data, events, and social factors to forecast passenger demand. This predictive capability will enable proactive service adjustments, reducing wait times and optimizing resource allocation in smart city environments [16].

7. The Promise of MTEC for Real-World Power System Applications

Several research projects have focused their efforts on developing more efficient electrical systems through maintenance. This involves developing extensive fault simulations to collect historical data to develop predictive model proposals. The ability of artificial intelligence algorithms to perform this task is a significant trend due to these models’ ability to interpret data and predict the failure [17].

Models such as the Multi-Objective Ensemble Classifier (MTEC) offer reliable application due to their ease of integration into fault reduction in electrical systems. Therefore, it is possible to estimate that future work will develop algorithm proposals that can improve upon other prediction models in terms of accuracy and fast response. This includes exploring meta-learning approaches that can automatically select and combine the most appropriate base classifiers for specific fault scenarios.

The evolution toward smart grids will drive the integration of ML-based fault classification with advanced monitoring systems, IoT devices, and distributed energy resources. Future algorithms will need to handle the increased complexity of modern power systems with renewable energy integration and bidirectional power flows.

8. Revolutionizing Pet Emotion Recognition: A Groundbreaking Deep Learning Study

One of the topics addressed by AI has been emotion recognition in humans. However, this has transcended the field to recognize emotions in domestic animals. The results on this topic have directly impacted areas such as veterinary medicine, thus becoming a multidisciplinary practice for animal welfare. With this, humans can improve their connection with animals in the context of coexistence.connections [18].

By developing a model that can accurately interpret pets’ subtle emotional cues, the researchers have created a foundation for automated systems that could transform veterinary care, animal welfare assessment, and human-animal interactions [19,20].

Therefore, it has been identified that in the future, data augmentation techniques can be explored to expand the dataset and improve the performance of emotion classification in animals.

9. Branching-Out Solution Algorithms for Fault Detection in Photovoltaic Systems: A Critical Analysis

According to [21], photovoltaic systems tend to fail primarily due to material degradation, which significantly affects their performance. For this reason, various research projects have focused on detecting and anticipating three types of failures in PV plants generating up to 250 kW. This has led to improved quality and performance of these systems using different deep learning techniques, taking into account the characteristics and target variables. For this purpose, tree models have been used to identify the cause of the fault types through variables. It is also possible to identify that the PV system is operating normally. For this, a classifier ensemble consisting of six different models was created. The result is that the LightGBM, CatBoost, and tree models performed better than the others, while XGBoost, LightGBM, and CatBoost demonstrated rapid convergence during training [22].

While the current framework achieves exceptional results with simulated data, future research directions should focus on validating these approaches with publicly available real-world PV datasets. This transition from simulation to practical implementation will be critical for establishing the framework’s practical applicability and commercial viability [23].

Future research will focus on expanding the framework to handle additional fault types beyond the current three categories (string faults, string-to-ground faults, and string-to-string faults). Integration with predictive maintenance systems and the development of hybrid approaches combining tree-based methods with other machine learning techniques are also key areas of investigation for the future [24].

10. Harnessing Digital Twins and AI Decision Models to Revolutionize Cost Modeling in Off-Site Construction

The integration of machine learning algorithms, neural networks, and adaptive neuro-fuzzy inference systems (ANFIS) promises t revolutionize cost prediction accuracy. Future developments will see AI models achieving outstanding reliability in lifecycle cost forecasting, with real-time adjustments reducing cost variability by up to 10 percent compared to traditional methods.

Digital Twins are projected to evolve to provide continuous, dynamic representations of modular construction projects, bridging physical and digital worlds through IoT sensors, BIM models, and project management tools. This real-time data integration will enable instant cost updates and decision making throughout the construction lifecycle [25].

AI models have also impacted the field of modular construction, aiming to reduce costs when building sustainably designed homes. Some works integrate decision trees to predict dynamic costs, providing a more accurate estimate for the user. This also makes it possible to determine the lifecycle of materials by considering market variability and nonlinearity. lifecycle [26].

However, several challenges remain, including the need for specialized skills to operate and maintain these systems effectively, requirements for substantial investment in technology infrastructure, and the current lack of standardized protocols for seamless integration. Future research opportunities include refining DT integration with AI-driven cost models [27], developing standardized data protocols [28], and validating scalability across diverse project contexts [29].

11. Explore Advanced Control Techniques for PMSM: Neural Network Model and Robust UKF Estimation

For researchers and engineers working in fields ranging from electric vehicles to industrial automation, this paper offers a promising framework that balances theoretical rigor with practical implementation concerns. The state space neural network model, combined with robust UKF estimation, represents a flexible architecture that could be adapted to various motor types and control requirements.

As we move toward more sophisticated electromechanical systems with higher efficiency and precision demands, approaches like the one presented by Velarde-Gomez and Giraldo will likely become increasingly important. Their work exemplifies how advanced estimation techniques can enhance the performance of nonlinear control systems, enabling new applications where traditional methods fall short [30].

Future research must address the computational complexity of real-time neural network training and state estimation. The development of efficient algorithms that can operate within the constraints of industrial control systems while maintaining accuracy and robustness is crucial. The dual UKF approach for state and parameter estimation shows promise, but optimization for real-time implementation remains a key challenge that will drive future algorithmic developments. Furthermore, the development of self-tuning algorithms that can adapt controller gains in real time to maintain optimal performance while respecting system constraints represents a significant opportunity for advancement.

12. Cutting-Edge AI Models to Predict Fouling Resistance in Heat Exchangers

The implications of this research extend beyond heat exchangers. The methodologies developed could be adapted to predict fouling in other industrial equipment, potentially revolutionizing maintenance practices across various sectors [31]. By shifting from systematic to predictive maintenance, industries could significantly reduce downtime, energy consumption, and environmental impact. Computational fluid dynamics (CFD) modeling has been a driving force in the development of predictive models for systematic maintenance. This has enabled researchers to predict phenomena such as the effects of fouling. However, it is increasingly common to observe that model fitting and prediction requires the lowest possible error. [32].

However, it is worth noting that while the models demonstrate excellent performance with the current dataset, their generalizability to different types of heat exchangers or operating conditions remains to be fully explored [33]. Future research should validate these models across diverse industrial settings and incorporate additional sensor inputs, such as pressure or chemical composition data, to further enhance prediction accuracy.

As industries worldwide strive for greater efficiency and sustainability, AI-based predictive maintenance tools like those developed in this study will likely become increasingly essential [34]. The researchers’ work advances our understanding of fouling prediction. It provides a practical framework for implementing AI solutions in industrial settings—a valuable contribution to academic knowledge and industrial practice [35].

As industries continue to prioritize sustainability and operational efficiency, the integration of AI-based fouling prediction systems is likely to become standard practice, transforming how heat exchangers are monitored, maintained, and optimized across various industrial applications.