3. Producción
Browse
11 results
Search Results
- Some of the metrics are blocked by yourconsent settings
Item type:Publication, Fuzzy data-driven scenario-based robust data envelopment analysis for prediction and optimisation of an electrical discharge machine's parameters(Elsevier Ltd, 2022-05-01)An electrical discharge machine (EDM) has a high impact on production management, with its process having many advantages over conventional machining processes, including the ability of the machine to create very high-quality material that is intricate to inner industrial sections. This study investigates the impact of EDM machining parameters on the volumetric flow rate, electrode corrosion percentage, and surface roughness. These machining parameters are increasingly important for the quality of the final product, leading to higher customer satisfaction and greater market share of the company. Due to dynamic changes in the machine's parameters and production environmental changes, using an uncertain model is inevitable. To investigate the machining data under uncertainty, a mathematical modelling approach based on the fuzzy possibility regression integrated (FPRI) model is developed. One advantage of the proposed model is that it is able to predict the surface roughness, volumetric flow rate, and corrosion percentage of the electrode. An adaptive-network-based fuzzy inference system (ANFIS) is applied to achieve the optimal levels of each output. Since the results and numbers obtained from the neural network are uncertain and their distribution is not clear, a robust data envelopment analysis approach (RDEA) is employed to select the best tuned-level of the parameters. The findings confirm the accuracy and reliability of the proposed method for prediction and optimisation of the EDM's parameters and encourage further tests for other production and supply chain applications. - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Robust automatic retrieval of soot volume fraction, temperature and radiation for axisymmetric flames(Elsevier Ltd, 2024-01-01)This work presents a robust methodology to retrieve local soot properties from line-of-sight integrated measurements without the need to invert a poorly-conditioned matrix arising from the flame geometry and discretization procedureFirst, a forward fit method is presented. Another method, utilizing an Artificial Neural Network informed by the Abel equation (ANNAbel), is then introduced to circumvent the drawbacks of the forward fit method. Both methods are capable to retrieve soot volume fraction, temperature and radiation satisfactorily from experimental data of an ethylene coflow non-premixed flame, without the need for a tuning a regularization parameter. The ANNAbel approach exhibited greater smoothness for retrieved properties, with lower errors when comparing the reconstructed data against the original experimental data. This was also evident when comparing local soot properties in a numerical framework. The ANNAbel approach also showed high resilience to increased levels of noise, contrary to the fitting approach and classical deconvolution methods. Finally, the ANNAbel method was capable to obtain the local properties even with simulated corrupted data, with a level of precision slightly lower than treating the original experimental data. On the contrary, the rest of the methods failed to perform this task. The ANNAbel method is then a promising approach for the robust and accurate determination of local flame properties, which is especially important for obtaining complex soot properties such as size and composition, where involved data treatment is required, and the results are sensitive to noise. - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Experimental thermodynamic investigation and hybrid RSM-ANN prediction of hydrogen-enriched algal biodiesel combustion(Elsevier B.V., 2026-12-01)The growing demand for low-carbon and high-efficiency combustion systems has accelerated research on renewable fuels compatible with existing compression ignition (CI) engines. This study experimentally investigated the combustion, performance and emission characteristics of a single-cylinder diesel engine fuelled with hydrogen-enriched Algal Oil Methyl Ester (AOME) using a hybrid Response Surface Methodology (RSM) and Artificial Neural Network (ANN) approach. Engine load (0 - 100%), injection pressure (200 - 240 bar), and hydrogen flow rate (3 - 9 LPM) were selected as the primary input parameters. A total of 45 experimental runs were conducted to evaluate peak cylinder pressure, heat release rate (HRR), brake thermal efficiency (BTE), brake specific fuel consumption (BSFC) and exhaust emissions. Results indicated that hydrogen enrichment significantly enhanced combustion characteristics, with peak cylinder pressure increasing from 34.75 to 79 bar and HRR rising from 29.72 to 157.62 J/ °CA at full load conditions. Maximum BTE of 44.01% and minimum BSFC of 0.141 kg/kWh were achieved under optimized conditions. Hydrogen addition also reduced CO, HC and smoke emissions by 91%, 90% and 99%, correspondingly, the NOₓ emissions increased at higher loads due to elevated combustion temperatures. The ANN model outperformed RSM, achieving prediction accuracy with R² values exceeding 0.98. Multi-objective optimization produced a desirability value of 0.958. The results prove the potential of hydrogen-enriched AOME dual-fuel operation for sustainable automotive and long-duration CI engine applications aligned with global net-zero and SDG goals.1 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Application of Machine Learning Algorithms to Classify Peruvian Pisco Varieties Using an Electronic Nose(Multidisciplinary Digital Publishing Institute (MDPI), 2023-07-01)Pisco is an alcoholic beverage obtained from grape juice distillation. Considered the flagship drink of Peru, it is produced following strict and specific quality standards. In this work, sensing results for volatile compounds in pisco, obtained with an electronic nose, were analyzed through the application of machine learning algorithms for the differentiation of pisco varieties. This differentiation aids in verifying beverage quality, considering the parameters established in its Designation of Origin”. For signal processing, neural networks, multiclass support vector machines and random forest machine learning algorithms were implemented in MATLAB. In addition, data augmentation was performed using a proposed procedure based on interpolation–extrapolation. All algorithms trained with augmented data showed an increase in performance and more reliable predictions compared to those trained with raw data. From the comparison of these results, it was found that the best performance was achieved with neural networks. - Some of the metrics are blocked by yourconsent settings
Item type:Publication, A Novel Experimental and Machine Learning Model to Remove COD in a Batch Reactor Equipped with Microalgae(Springer Science and Business Media Deutschland GmbH, 2023-07-01)By using microorganisms and the microalgae Chlorella vulgaris in conjunction with sequencing batch reactors (SBRs), the performance of a wastewater treatment facility was studied. For this purpose, the effect of pH, temperature, COD inlet , and air flowrate on COD removal rate and residual was investigated. A single-factorial optimization method is utilized to optimize the amount of COD removal, and the best result is obtained with a pH of 8, CODinlet=600mg/l , and an airflow rate of 55 l/min. Under optimal conditions, the amount of residual COD in the effluent reached 36 mg / l , showing an augmentation in the efficiency of the desired system. Moreover, empirical correlations are proposed for double-factorial optimization of residual COD and COD removal. Also, a multilayer perceptron artificial neural network is proposed to model the process and predict the residual COD concentration. The useful technique of hyperparameter tuning is utilized to obtain the best result for the predictions. All the effective parameters, including the number of hidden layers, neurons, epochs, and batch size, are adjusted. Data from the experiments agreed well with the artificial neural network modeling results. For this modeling, the values of the correlation coefficient (R2) and mean absolute error (MAE) were obtained as 0.98 and 2%, respectively. - Some of the metrics are blocked by yourconsent settings
Item type:Publication, 1 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Bayesian Calibration of a 2d Hydraulic Model Using a Convolutional Neural Network Emulator(RELX Group (Netherlands), 2025-01-01)6 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, AP-Traj2: Transformer-based Trajectory Prediction with Graph-Enhanced Attention Mechanism(Slovenian Society Informatika, 2025-12-15)Trajectory prediction is essential for understanding human mobility patterns, with applications such as itinerary recommendation and urban planning. It involves analyzing sequences of visited locations to forecast the user's next destination. Traditional approaches have often relied on Markov chains or recurrentneural networks (RNNs). More recently, Transformer neural networks have gained attention for sequential prediction tasks due to their superior parallelization and training efficiency. In this study, we propose AP-Traj2 (Attention and Possible directions for TRAJectory prediction 2), a model designed to enhance prediction accuracy by leveraging attention mechanisms and graph-based movement modeling. AP-Traj2 employs self-attention to capture dependencies among visited locations, explores feasible next steps through a graph of possible directions, and incorporates contextual information via location embeddings. Experiments conducted on GPS, CDR, and WiFi datasets demonstrate that AP-Traj2 improves the average matchratio by approximately 50% over state-of-the-art methods. Moreover, it achieves significantly faster training times, with reductions of up to 72% in the best-case scenario. Unlike existing approaches that focus primarily on neural network architecture, this work emphasizes the importance of data preprocessing andfiltering, highlighting their substantial impact on model performance.1 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Deep Neural Network-Assisted Microfluidic pH Sensor(Institute of Electrical and Electronics Engineers Inc., 2025)Water pH measurement is vital as it provides fundamental information about its quality and suitability for agriculture, aquatic ecosystems, industry, and human consumption. Each of these applications may require numerical readings of acidity or alkalinity, preferably using tools that are already ubiquitous, such as cellphones. This work presents a microfluidic lab-on-a-chip system to measure the pH of liquid samples. We used purple cabbage as the colorimetric reagent to produce a 2640-image dataset with pH levels in the range of [2–12] on a polydimethylsiloxane (PDMS) microfluidic recipient. We fed our dataset to our parameterized deep neural network (DNN) to classify our samples and found an accuracy of 99.7%. In addition, we developed a mobile application with an easy-to-use graphic user interface that recognizes the microfluidic device shape, classifies the image’s color, and returns the pH level.1 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Integrating Deep Learning into PnP-ADMM for Ultrasound Attenuation Coefficient Estimation(IEEE Computer Society, 2025)Estimating the attenuation coefficient slope (ACS) is essential for tissue characterization in quantitative ultrasound (QUS). Traditional model-based methods such as the regularized spectral log difference (RSLD) rely on manually tuned priors, while recent end-to-end deep learning approaches struggle to generalize to in vivo data. This work proposes a hybrid method that integrates pre-trained Attention U-Nets within a Plug-and-Play ADMM framework. The fidelity term is replaced by a network conditioned on spectral inputs and the iteration index, while a second network acts as a learned regularizer. The method was evaluated on physical phantoms and in vivo breast and thyroid acquisitions, after being trained entirely on simulations. Results show improved accuracy and generalization over RSLD and end-to-end baselines, suggesting that embedding a pre-trained deep learning model within a physics-based framework enhances robustness and enables more reliable ACS estimates in real-world scenarios.2