Artificial Neural Network Architecture Research Articles

A novel reverse and forward directional relaying scheme in six phase overhead transmission lines using adaptive neuro-fuzzy inference system

Recent power system is structurally difficult and is vulnerable to undesirable conditions like transmission faults. In this event of transmission line faults, exact fault zone detection enhances the restoration process, thus improving reliability of the complete power system. In order to solve the above problem, this paper presents an adaptive neuro-fuzzy inference system (ANFIS) based fault zone detector, which combines artificial neural network (ANN) and fuzzy logic technique (FLT) in six phase overhead transmission lines (SPOTL). To overcome the limitation of ANN and fuzzy expert system (FES) architectures and, the selection work has been formulated as an optimization method and solved using ANFIS. The inputs are the zero sequence component currents at the middle bus of the transmission line. The training data are extracted using discrete Fourier transform and collected, and then ANFIS is trained to identify the fault zone. The ANFIS based scheme reach setting has been checked for various types of faults, with a wide range of faults and transmission line parameters. Simulation study ensure that this method has a high reach setting, does not require the design of communication channel. Further, the ANFIS study shows that ANFIS is suitable for all type of faults. The ANFIS significantly outperforms other techniques proposed in the literature in terms of various evaluation metrics.

Multi-Step Ageing Prediction of NMC Lithium-Ion Batteries Based on Temperature Characteristics

The performance of lithium-ion batteries depends strongly on their ageing state; therefore, the monitoring and the prediction of the battery state of health (SoH) is necessary for an optimized and secured functioning of battery systems. This paper evaluates and compares three artificial neural network architectures for multi-step ageing prediction of lithium-ion cells: Recurrent Neural Network (RNN), Gated Recurrent Unit (GRU) and Long short-term memory (LSTM). These models use the features extracted from the cell’s temperature to predict the cell’s capacity. The features are extracted from experimental measurements of the cell’s surface temperature and selected based on Spearman correlation analysis. The prediction results were evaluated and compared considering three different percentages of the training dataset: 60%, 70%, and 80%. Training and testing data were generated experimentally based on accelerated ageing cycling tests. During these experiments, four Nickel Manganese Cobalt/Graphite (NMC) cells were cycled under a controlled temperature environment based on a dynamic current profile extracted from the Worldwide Harmonized Light Vehicles Test Cycles.

Artificial intelligence in suicide prevention: Utilizing deep learning approach for early detection

ABSTRACT Background: Suicide among students is increasing in India and is a matter of grave concern. Early identification of students contemplating suicide would facilitate emergency intervention and may save precious lives. Aim: Our primary objective was to construct an artificial intelligence (AI) model employing an artificial neural network (ANN) architecture to predict students at risk of suicidal tendencies. This initiative was prompted by the necessity to implement a proactive and technologically driven strategy for identifying competitive exam-bound students facing heightened vulnerability. The aim was to facilitate timely interventions aimed at reducing the risk of self-harm. Materials and Methods: An AI model utilizing ANNs is devised for suicide risk prediction among exam-stressed students. A 33-feature input layer is curated based on literature and expert insights, with binary features assigned weighted values. A rigorous hyperparameter optimization approach using the Optuna library to select the most effective neural network model. Ridge regression was used to determine bias or variance in the dataset. Training and testing of the model are conducted using fictional and simulated profiles, respectively, and model performance is assessed through statistical metrics and the Cohen’s Kappa coefficient, benchmarked against expert evaluations. Result: The AI model demonstrates exceptional predictive capabilities for suicide risk assessment among competitive exam students. Quantitative Metrics: The model’s accuracy of 98% aligns predictions with outcomes, distinguishing risk categories. Precision at 100% identifies cases within predicted risks, minimizing false positives. A recall of 97% identifies true risk cases, highlighting sensitivity. F1 Score: The model’s F1 score of 98% balances precision and recall, indicating overall performance. Cohen’s Kappa: With a coefficient of 1.00, the model’s substantial agreement with experts underscores its consistent classifications. Conclusion: The study introduces an AI model utilizing ANNs for suicide risk prediction among stressed students. High precision, recall, and accuracy align with expert evaluations, highlighting its promise for timely risk identification. The model’s efficiency in evaluating large populations swiftly indicates its clinical potential. Refinement and real-world validation remain future considerations.

Enhancing IoT Security through an Artificial Neural Network Approach

This study aims to fortify Internet of Things (IoT) security through the strategic implementation of Artificial Neural Networks (ANNs). With the rapid expansion of IoT devices, traditional security measures have struggled to cope with the dynamic and complex nature of these environments. ANNs, known for their adaptability, are explored as a promising solution to enhance security. The central objective is to significantly improve the accuracy of IoT security measures by optimizing ANN architectures. Using a curated dataset with key environmental parameters, the study evaluates three ANN models—Backpropagation Neural Network (BPNN), Multilayer Perceptron (MLP), and Long Short-Term Memory (LSTM). The evaluation metrics include accuracy, precision, recall, and F1-score across different train-test splits. Results show that LSTM consistently outperforms BPNN and MLP, demonstrating superior accuracy and the ability to capture temporal dependencies within IoT security data. Implications stress the importance of aligning model selection with specific application goals, considering factors like computational efficiency. In conclusion, this research contributes valuable insights into the practical implementation of ANNs for IoT security, guiding future optimization efforts and addressing real-world deployment challenges to safeguard sensitive data and ensure system resilience in the evolving IoT landscape.

Integrated Artificial Neural Network with Trainable Activation Function Enabled by Topological Insulator-Based Spin-Orbit Torque Devices.

Nonvolatile memristors offer a salient platform for artificial neural network (ANN), yet the integration of different function and algorithm blocks into one hardware system remains challenging. Here we demonstrate the brain-like synaptic (SOT-S) and neuronal (SOT-N) functions in the Bi2Te3/CrTe2 heterostructure-based spin-orbit torque (SOT) device. The SOT-S unit exhibits highly linear and symmetrical long-term potentiation/depression process, resulting in a fast-training of the MNIST data set with the classification accuracy above 90%. Meanwhile, the Sigmoid-shape transition curve inherited in the SOT-N cell replaces the software-based activation function block, hence reducing the system complexity. On this basis, we employ a serial-connected, voltage-mode sensing ANN architecture to enhance the vector-matrix multiplication signal strength with low reading error of 0.61% while simplifying the peripheral circuitry. Furthermore, the trainable activation function of SOT-N enables the implementation of the Batch Normalization algorithm and activation operation within one clock cycle, which bring about improved on/off-chip training performance close to the ideal baseline.

Forecasting ETF Performance: A Comparative Study of Deep Learning Models and the Fama-French Three-Factor Model

The global financial landscape has witnessed a significant shift towards Exchange-Traded Funds (ETFs), with their market capitalization surpassing USD 10 trillion in 2023, due to advantages such as low management fees, high liquidity, and broad market exposure. As ETFs become increasingly central to investment strategies, accurately forecasting their performance has become crucial. This study addresses this need by comparing the efficacy of deep learning models against the traditional Fama-French three-factor model in predicting daily ETF returns. The methodology employs eight artificial neural network architectures, including ANN, LSTM, GRU, CNN, and their variants, implemented in Python and applied to data ranging from 2010 to 2020, while also exploring the impact of additional factors on forecast accuracy. Empirical results reveal that LSTM and the Fama-French three-factor model exhibit a superior performance in ETF return prediction. This study contributes to the literature on financial forecasting and offers practical insights into investment decision making. By leveraging advanced artificial intelligence techniques, this study aims to enhance the toolkit available for ETF performance analysis, potentially improving investment strategies in this dynamic market segment.

Computer Modelling of Textures on Images with Human Skin Wound Areas

In this paper, we concentrate on the images of the wounds on the human skin and propose to consider each image as a set of smaller pieces – crops or patches containing different textures. We overview, develop and compare deep learning feature extraction methods to model image crops as 200-dimensional feature vectors using various artificial neural network architectures: convolutional autoencoders, variational convolutional autoencoders, and Siamese convolutional networks trained in the contrastive learning manner. Also, we develop a custom convolutional encoder and decoder, use them in the aforementioned architectures and compare them with the ResNet encoder and decoder alternatives. Finally, we train and evaluate k-nearest neighbors and Multi-Layer Perceptron classifiers on the features extracted with the model above options to discriminate skin, wound, and background image patches. Classification evaluation results on the features, extracted with the Siamese network, show the best test accuracy for all implementations without a significant shift between model versions (accuracy > 93%); variational autoencoders show random results for all options (accuracy around 33%), and convolutional autoencoders reached good results (accuracy > 77%) but with a noticeable difference between the custom and ResNet versions; the latter is better. Custom encoder and decoder implementations are faster and smaller than the ResNet alternatives but may be less stable on larger datasets, which still needs investigation. Possible applications of the feature vectors include an area of interest extraction during wound segmentation or classification and usage as patch embeddings while training vision transformer architectures.

Artificial neural network-based prediction of multiple sclerosis using blood-based metabolomics data

Multiple sclerosis (MS) remains a challenging neurological condition for diagnosis and management and is often detected in late stages, delaying treatment. Artificial intelligence (AI) is emerging as a promising approach to extracting MS information when applied to different patient datasets. Given the critical role of metabolites in MS profiling, metabolomics data may be an ideal platform for the application of AI to predict disease. In the present study, a machine-learning (ML) approach was used for a detailed analysis of metabolite profiles and related pathways in patients with MS and healthy controls (HC). This approach identified unique alterations in biochemical metabolites and their correlation with disease severity parameters. To enhance the efficiency of using metabolic profiles to determine disease severity or the presence of MS, we trained an AI model on a large volume of blood-based metabolomics datasets. We constructed this model using an artificial neural network (ANN) architecture with perceptrons. Data were divided into training, validation, and testing sets to determine model accuracy. After training, accuracy reached 87%, sensitivity was 82.5%, specificity was 89%, and precision was 77.3%. Thus, the developed model seems highly robust, generalizable with a wide scope and can handle large amounts of data, which could potentially assist neurologists. However, a large multicenter cohort study is necessary for further validation of large-scale datasets to allow the integration of AI in clinical settings for accurate diagnosis and improved MS management.

MATHEMATICAL MODEL AND STRUCTURE OF A NEURAL NETWORK FOR DETECTION OF CYBER ATTACKS ON INFORMATION AND COMMUNICATION SYSTEMS

The paper discusses the principles of creating a mathematical model and system architecture by applying the method of artificial intelligence to detect cyberattacks on information and communication systems, where a neural network capable of learning and detecting cyberattacks is used. The proposed approach, based on the application of the developed mathematical model and architecture of artificial neural networks, as a detector of network attacks on information and communication systems, allows to increase the level of detection of network intrusions into computer systems, Web and Internet resources. An algorithm for processing network traffic parameters in real-time systems by structuring a neural network is proposed, which allows to optimize the redundancy of its multi-level structure at the level of inter-element connections.

Application of artificial neural network for peak load forecasting in 150 kV Semarang power system

Accurate load forecasting is essential for reliable and efficient operation of power systems. Traditional forecasting methods often struggle with capturing complex nonlinear patterns in load data. Artificial neural networks (ANNs) have emerged as a promising alternative due to their ability to learn complex relationships from historical data (Syed et al. in IEEEA 9:54992–55008, 2021. https://doi.org/10.1109/ACCESS.2021.3071654). This study investigates the potential of ANNs for short-term peak load forecasting in a 150 kV power system in Semarang, Indonesia. The study examines the impact of different input variables, including historical peak load, minimum load, population, and energy production, on forecasting accuracy. Several ANN architectures are trained and evaluated using mean absolute percentage error (MAPE) and mean squared error (MSE) metrics as reported by Demuth and De Jesús (neural network design). The results indicate that ANNs can achieve high accuracy in predicting peak load, with MAPE values below 10%. The study also demonstrates the importance of carefully selecting input variables and training parameters for optimal model performance. The findings highlight the potential of ANNs for improving load forecasting accuracy in power systems, contributing to enhanced grid reliability and operational efficiency. The findings of this study contribute to a deeper understanding of the application of ANNs in power system load forecasting. They demonstrate the potential of ANNs to achieve high accuracy and provide valuable insights into the factors influencing model performance. The findings are relevant for power system operators, researchers, and policymakers working to improve grid reliability and efficiency as reported by Prabha Kundur and Malik (Power System Stability and Control, McGraw-Hill Education, New York, 2022. https://www.accessengineeringlibrary.com/content/book/9781260473544).

Ionic surfactants critical micelle concentration prediction in water/organic solvent mixtures by artificial neural network

Abstract Critical micellar concentration (CMC) is a key physicochemical property of surfactants used to study their behaviour. This property is affected by factors such as temperature, pressure, pH, the type of organic solvent/water mixture, the chemical structure of the surfactants and the presence of electrolytes. Most of the existing studies in the literature have predicted the CMC under fixed conditions based on the chemical parameters of the surfactant. In this study, a machine learning approach using artificial neural network (ANN) models was used to estimate the CMC of some ionic surfactants. These models considered variables defining both the organic solvent-water mixture (T, molecular weight, molar fraction and log P) and the chemical structure of the surfactant (number of atoms of each element). A database consisting of a total of 258 CMC values for 10 ionic surfactants was collected from the literature. The ANN architecture consisting of an input layer with 12 neurons, an intermediate layer with 25 neurons and one neuron in the output layer is proposed. According to the results, the normalized ANN models provided the best statistical adjustments for the CMC prediction. These ANN models could be a promising method for CMC estimation.

ANN modeling and optimization of friction stir welding performance for AA6061 and AA5083 alloy joints

This research focused on the characteristics of the AA6061/AA5083 alloy joints which are fabricated by the friction stir welding (FSW) technique. The impact of FSW parameters such as tool tilt angle, pin depth, and tool geometry on the mechanical properties and fractographic characteristics were studied. Ultimate tensile strength (UTS), hardness (HBN), and specific wear rate (SWR) were examined for characterization while fractography analyses were done using scanning electron microscopy. Taguchi-Grey relational analysis (GRA) was employed to execute multicriteria optimization and recognize the best grouping of parameters. The optimized values attained from this analysis were 259.2 MPa for UTS, 94.9 HV for HBN, and 0.0819 m³/Nm for SWR. The analysis of variance (ANOVA) results derived from the GRA revealed that tool geometry exerted the most significant influence (39.79%) on output factors, followed by pin depth (24.90%) and tilt angle (24.26%). To enhance the predictive accuracy of the output responses, an artificial neural network (ANN) model was developed utilizing the Levenberg-Marquardt (LM) algorithm. The optimized ANN architecture, configured as (3–10–3), exhibited robust regression analysis outcomes, showcasing correlation coefficients ( R) of 0.99999, 0.99967, and 0.99994 for the training, validation, and test datasets. The overall R-value, computed at 0.99958, affirmed high conformity between experimental and predicted values.

457 Prediction of methane emissions using rumination time and milk mid-infrared spectral data via artificial neural networks

Abstract Cattle methane emissions (ME) account for approximately 6% of global anthropogenic greenhouse gas emissions. Given the challenges in measuring ME directly from individual animals, there is a need for the development of novel indirect methods. Rumination time (RT) and milk mid-infrared spectral data (MIR) show promise for the indirect assessment of ME in dairy cows. Both traits have been used as indicators of reproduction, production, and gas emission traits. Methodologies combining the use of MIR and machine learning algorithms such as artificial neural networks (ANN) for the prediction of ME have been successful; however, the inclusion of RT has not been assessed. This study aimed to evaluate the impact of RT on milk MIR-based models using ANN for the prediction of ME. One-week averages for RT, ME, and MIR from first-lactation Canadian Holstein cows (n = 412) were calculated. Six data sets were evaluated using a multilayer perceptron ANN. All sets included age at calving, season of calving and days in milk as model factors, but varied in using milk MIR data points (1,060 or 235) and including or not including RT. The ANN architecture consisted of one input layer, one hidden layer with one or more neurons, and one output layer. Results showed that sets using both RT and milk MIR data achieved correlations from 0.5 to 0.6 between predicted and observed ME. Notably, the inclusion of RT did not improve the performance of the models. Predictions may be improved through the use of larger data sets, the use of daily records, and inclusion of data across herds and lactations. Optimizing parameters of the ANN could also improve predictions. Further research is needed to fully assess the potential of RT as a predictor of ME in dairy cows.

Long-term power forecasting of photovoltaic plants using artificial neural networks

Increasing the robustness of forecasting energy production models from a photovoltaic plant is crucial in investment decisions, while profitability is closely linked to its location since its energy production depends on the local climate conditions. This study aimed to mitigate uncertainty regarding the installation of photovoltaic systems by addressing the challenge of forecasting electrical energy production over the long term. Artificial Neural Network models were used for this purpose, predicting the power output of a photovoltaic plant based on the ambient temperature, cell temperature, and solar irradiance. Data recorded every minute over one year at an experimental photovoltaic plant revealed a strong correlation between energy production and the input variables. This research compared the performance of multilayer perceptron, feedforward, long short-term memory, and modular artificial neural networks architectures. Validation was carried out using mean squared error, mean absolute error, mean absolute percentage error, root mean squared error, and coefficient of determination as key metrics. Notably, the long short-term memory network demonstrated the highest accuracy in energy forecasting achieving a mean squared error of 0.0089 p.u, a mean absolute error of 0.0527 p.u, a root mean squared error of 0.0944 p.u, and a mean absolute percentage error of 49.0124 %. Nonetheless, the modular model came close to long short-term memory accuracy but with lower computational cost.

Obtaining parallax-free X-ray powder diffraction computed tomography data with a self-supervised neural network

In this study, we introduce a method designed to eliminate parallax artefacts present in X-ray powder diffraction computed tomography data acquired from large samples. These parallax artefacts manifest as artificial peak shifting, broadening and splitting, leading to inaccurate physicochemical information, such as lattice parameters and crystallite sizes. Our approach integrates a 3D artificial neural network architecture with a forward projector that accounts for the experimental geometry and sample thickness. It is a self-supervised tomographic volume reconstruction approach designed to be chemistry-agnostic, eliminating the need for prior knowledge of the sample’s chemical composition. We showcase the efficacy of this method through its application on both simulated and experimental X-ray powder diffraction tomography data, acquired from a phantom sample and an NMC532 cylindrical lithium-ion battery.

Valorization of tomato processing by-products: Predictive modeling and optimization for ultrasound-assisted lycopene extraction

Lycopene is a carotenoid highly valuable to the food, pharmaceutical, dye, and cosmetic industries, present in ripe tomatoes and other fruits with a distinctive red color. The main source of lycopene is tomato crops. This bioactive component can be successfully isolated from tomato processing waste, commonly called tomato pomace, mostly made from tomato skins, seeds, and some residual tomato tissue. The main investigative focus in this work was the application of green engineering principles in each stage of the optimized ultrasound-assisted extraction (UAE) of enzymatically treated tomato skins to obtain functional extracts rich in lycopene. The experimental plan was designed to determine the influence of studied operating parameters: enzymatic reaction time (60, 120, and 180 min), extraction time (0, 5, 10, 15, 30, 60, and 120 min), and temperature (25, 35 and 45 ℃) on lycopene yield. Process optimization was performed based on the yield of lycopene [1018, 1067, and 1120 mg/kg] achieved at optimal operating conditions. An artificial neural network (ANN) model was developed and trained for predictive modeling of the closed extraction system, with operating parameters used as input neurons and experimentally obtained values for lycopene content defined as the output neural layer. Applied ANN architecture provided a high correlation of experimental output with ANN-generated data (R=0.99914) with a model deviation error for the entire data set of RMSE=5.3 mg/kg. The k-Nearest Neighbor algorithm was introduced to predict lycopene yield using experimental key features: operating temperature, extraction time, and time of enzymatic treatment, split into training and testing sets with an 85/15 ratio. The model interpretation was conducted through the SHAP (SHapley Additive exPlanations) methodology.

Artificial Intelligence-Based Improvement of Empirical Methods for Accurate Global Solar Radiation Forecast: Development and Comparative Analysis

Artificial intelligence (AI) technology has expanded its potential in environmental and renewable energy applications, particularly in the use of artificial neural networks (ANNs) as the most widely used technique. To address the shortage of solar measurement in various places worldwide, several solar radiation methods have been developed to forecast global solar radiation (GSR). With this consideration, this study aims to develop temperature-based GSR models using a commonly utilized approach in machine learning techniques, ANNs, to predict GSR using just temperature data. It also compares the performance of these models to the commonly used empirical technique. Additionally, it develops precise GSR models for five new sites and the entire region, which currently lacks AI-based models despite the presence of proposed solar energy plants in the area. The study also examines the impact of varying lengths of validation datasets on solar radiation models’ prediction and accuracy, which has received little attention. Furthermore, it investigates different ANN architectures for GSR estimation and introduces a comprehensive comparative study. The findings indicate that the most advanced models of both methods accurately predict GSR, with coefficient of determination, R2, values ranging from 96% to 98%. Moreover, the local and general formulas of the empirical model exhibit comparable performance at non-coastal sites. Conversely, the local and general ANN-based models perform almost identically, with a high ability to forecast GSR in any location, even during the winter months. Additionally, ANN architectures with fewer neurons in their single hidden layer generally outperform those with more. Furthermore, the efficacy and precision of the models, particularly ANN-based ones, are minimally impacted by the size of the validation data sets. This study also reveals that the performance of the empirical models was significantly influenced by weather conditions such as clouds and rain, especially at coastal sites. In contrast, the ANN-based models were less impacted by such weather variations, with a performance that was approximately 7% better than the empirical ones at coastal sites. The best-developed models, particularly the ANN-based models, are thus highly recommended. They enable the precise and rapid forecast of GSR, which is useful in the design and performance evaluation of various solar applications, with the temperature data continuously and easily recorded for various purposes.

Predicting Sodium-Ion Battery Performance through Surface Chemistry Analysis and Textural Properties of Functionalized Hard Carbons Using AI.

Traditionally, the performance of sodium-ion batteries has been predicted based on a single characteristic of the electrodes and its relationship to specific capacity increase. However, recent studies have shown that this hypothesis is incorrect because their performance depends on multiple physical and chemical variables. Due to the above, the present communication shows machine learning as an innovative strategy to predict the performance of functionalized hard carbon anodes prepared from grapefruit peels. In this sense, a three-layer feed-forward Artificial Neural Network (ANN) was designed. The inputs used to feed the ANN were the physicochemical characteristics of the materials, which consisted of mercury intrusion porosimetry data (SHg and average pore), elemental analysis (C, H, N, S), ID/IG ratio obtained from RAMAN studies, and X-ray photoemission spectroscopy data of the C1s, N1s, and O1s regions. In addition, two more inputs were added: the cycle number and the applied C-rate. The ANN architecture consisted of a first hidden layer with a sigmoid transfer function and a second layer with a log-sigmoid transfer function. Finally, a sigmoid transfer function was used in the output layer. Each layer had 10 neurons. The training algorithm used was Bayesian regularization. The results show that the proposed ANN correctly predicts (R2 > 0.99) the performance of all materials. The proposed strategy provides critical insights into the variables that must be controlled during material synthesis to optimize the process and accelerate progress in developing tailored materials.

When linear inversion fails: Neural-network optimization for sparse-ray travel-time tomography of a volcanic edifice

In this study, we present an artificial neural network (ANN)-based approach for travel-time tomography of a volcanic edifice under sparse-ray coverage. We employ ray tracing to simulate the propagation of seismic waves through the heterogeneous medium of a volcanic edifice, and an inverse modeling algorithm that uses an ANN to estimate the velocity structure from the “observed” travel-time data. The performance of the approach is evaluated through a 2-dimensional numerical study that simulates i) an active source seismic experiment with a few (explosive) sources placed on one side of the edifice and a dense line of receivers placed on the other side, and ii) earthquakes located inside the edifice with receivers placed on both sides of the edifice. The results are compared with those obtained from conventional damped linear inversion. The average Root Mean Square Error (RMSE) between the input and output models is approximately 0.03 km/s for the ANN inversions, whereas it is about 0.4 km/s for the linear inversions, demonstrating that the ANN-based approach outperforms the classical approach, particularly in situations with sparse ray coverage. Our study emphasizes the advantages of employing a relatively simple ANN architecture in conjunction with second-order optimizers to minimize the loss function. Compared to using first-order optimizers, our ANN architecture shows a ∼25% reduction in RMSE. The ANN-based approach is computationally efficient. We observed that even though the ANN is trained based on completely random velocity models, it is still capable of resolving previously unseen anomalous structures within the edifice with about 5% anomalous discrepancies, making it a potentially valuable tool for the detection of low velocity anomalies related to magmatic intrusions or mush.

Imitating the respiratory activity of the brain stem by using artificial neural networks: exploratory study on an animal model of lactic acidosis and proof of concept.

Artificial neural networks (ANNs) are versatile tools capable of learning without prior knowledge. This study aims to evaluate whether ANN can calculate minute volume during spontaneous breathing after being trained using data from an animal model of metabolic acidosis. Data was collected from ten anesthetized, spontaneously breathing pigs divided randomly into two groups, one without dead space and the other with dead space at the beginning of the experiment. Each group underwent two equal sequences of pH lowering with pre-defined targets by continuous infusion of lactic acid. The inputs to ANNs were pH, ΔPaCO2 (variation of the arterial partial pressure of CO2), PaO2, and blood temperature which were sampled from the animal model. The output was the delta minute volume (ΔVM), (the change of minute volume as compared to the minute volume the animal had at the beginning of the experiment). The ANN performance was analyzed using mean squared error (MSE), linear regression, and the Bland-Altman (B-A) method. The animal experiment provided the necessary data to train the ANN. The best architecture of ANN had 17 intermediate neurons; the best performance of the finally trained ANN had a linear regression with R2 of 0.99, an MSE of 0.001 [L/min], a B-A analysis with bias ± standard deviation of 0.006 ± 0.039 [L/min]. ANNs can accurately estimate ΔVM using the same information that arrives at the respiratory centers. This performance makes them a promising component for the future development of closed-loop artificial ventilators.