In this paper, a conceptual design of a Delta robot is developed by means of a mechatronic design methodology. A fully integrated conceptual design, clarifying the recurrence of the conceptual design process using black-box/white-box analysis, is presented using the Model Based Systems Engineering (MBSE) paradigm and the SysML language as the formal modeling tool. Multiple designs proposals are then evaluated by the non-linear Choquet integral in order to choose the most appropriate according to a multicriteria requirement. For a preliminary conceptual design, structural parameters for the Delta robot are determined by defining and solving a nonlinear constrained optimization problem, which considers the kinematic model of the robot maximizing its workspace. Both the decision making and the optimization problem are integrated and automated into a common software framework for the design process, by using a standard genetic algorithm and Monte Carlo method to set the optimized conceptual design to be rendered in Computer Aided Design (CAD) software and in a physical prototype, satisfying the technical specifications.
Biogas production estimation has been one of the most important and challenging objectives for anaerobic digestion processes due to the complexity of its dynamics and the lack of high-quality open-access datasets. This study presents a hybrid modeling framework that combines a mechanistic model, based on ordinary differential equations (ODEs), with a machine learning model. Rather than relying exclusively on experimental data, the proposed approach leverages physics-informed synthetic data generation, complemented by a lag-based feature engineering to capture inherent temporal dependencies in the process dynamics available in operational data of a bio-digester. Two configurations were evaluated: a baseline model and an enhanced version incorporating lag features and a simplified temperature profile. This specific computational enhancement provides a robust predictive core that successfully avoids the severe predictive degradation observed in purely mechanistic approaches at high spatial discretizations. While the improved surrogate model achieved high predictive performance (, [L/d]), additional analyses reveal that this resilience is driven by temporal memory and remains sensitive to noise and feature composition. Instead of presenting the model as a final independent physical validation, this work is rigorously framed as a proof-of-concept digital twin core, acknowledging the gap that still exists between simulation-based ODE emulation and unstructured real-world reliability.
Biogas production estimation has been one of the most important and challenging objectives for anaerobic digestion processes due to the complexity of its dynamics and the lack of high-quality open-access datasets. This study presents a hybrid modeling framework that combines a mechanistic model, based on ordinary differential equations (ODE), with a machine learning model. Rather than relying exclusively on experimental data, the proposed approach leverages physics-informed synthetic data generation, complemented by a lag-based feature engineering to capture inherent temporal dependencies in the process dynamics available in operational data of a bio-digester. Two configurations were evaluated: a baseline model and an enhanced version incorporating lag features and simplified temperature profile. While the improved model achieved high predictive performance (R2=0.97885, RMSE=131.80[L/d]), additional analyses reveal that this performance is partly driven by temporal memory and remains sensitive to noise and feature composition. Instead of presenting the model as a final solution, this work frames it as a step toward practical digital twin implementations, acknowledging the gap that still exists between simulation-based accuracy and real-world reliability.
Non-invasive blood glucose level (BGL) estimation using photoplethysmography has been studied as a promising alternative to enzyme-based technologies. However, the reliability of current machine learning models is frequently compromised by overfitting and data leakage. This paper presents a methodological validation framework designed to expose and mitigate these issues by objectively comparing spectral and temporal statistical feature extraction strategies using the MUST dataset. The signals were pre-processed and evaluated through two distinct approaches: extracting 12 Mel-frequency cepstral coefficients (MFCC) for a support vector regressor (SVR), and 22 time-domain statistical features for random forest (RF) and Extreme Gradient Boosting (XGBoost) models. To expose data leakage, models were rigorously tested using both standard 5-fold cross-validation and a strict subject-independent grouped validation. Experimental results confirm that standard cross-validation inflates predictive metrics; the MFCC-SVR model initially achieved a moderate correlation ( , RMSE = 17.00 mg dl ) under this random-split scheme. However, when evaluated through the proposed subject-independent validation framework, the predictive correlation dropped significantly ( , RMSE = 17.44 mg dl ). This degradation demonstrates that the models explain a low percentage of actual BGL variance and tend to output predictions near the population mean rather than tracking individual physiological fluctuations. While the MFCC approach proved comparatively more stable than temporal statistical models (RF: ; XGBoost: ), its weak predictive correlation indicates it is not currently suitable as a standalone tool for clinical decision support. Instead, while 100% of predictions fall within Zones A and B of the Clarke Error Grid Analysis, this outcome reflects the limited blood glucose distribution of the dataset rather than clinical deployment readiness. Ultimately, by exposing the limitations of current models under strict validation, this study establishes the critical need for larger, comprehensive datasets to enable the development of truly robust non-invasive monitoring technologies.