Heat exchangers are utilized in a vast region of the process industry for heating and cooling. Long-term operation of heat exchangers results in decreased efficiency due to many problems, such as fouling. Therefore, the object of this research paper is to use three artificial intelligence techniques (feedforward neural networks-multilayer perceptron (FNN-MLP), nonlinear autoregressive networks with exogenous inputs (NARX), and support vector machines (SVM-RBF)) for predicting the fouling resistance in the tube and the shell heat exchanger in the preheating circuit of atmospheric distillation. The results summarize the high training as well as the predictive capacity of the "FFNN-MLP" model for predicting the fouling resistance in the heat exchanger with the highest coefficient of correlation (R = 0.99961) and the lowest root-mean-squared error (nRMSE = 1.0031%) for the testing phase, where the FNN-MLP network is superior to that provided using the SVM model (R = 0.9955 and nRMSE = 3.8652%). All the models of artificial networks and machine learning techniques used in the current work can be used to predict the fouling resistance in heat exchanger data with high accuracy. Despite this, the FNN-MLP model is the preferred model compared with the other proposed models, followed by the NARX model.

Heat exchangers play a crucial role in transferring heat between two mediums, directly impacting energy efficiency, product quality, and operational safety in industrial systems. This study presents a novel approach for fouling resistance estimation using two artificial intelligence models, the cascaded forward network (CFN) and the recurrent neural network (RN), with a minimal set of six input parameters. The proposed models utilize temperature and flow sensor data from heat exchangers to predict fouling resistance. The training process is optimized using the Levenberg–Marquardt (LM) algorithm, ensuring rapid convergence and high accuracy. Model performance is assessed based on mean squared error (MSE), regression values (R), and statistical error analysis. The results demonstrate that both models achieve high accuracy in predicting fouling resistance, with the CFN model outperforming the RN model. The CFN model achieves an MSE of 1.54 × 10⁻⁸, significantly lower than the RN model (MSE = 3.05 × 10⁻⁸), resulting in a 49.5% improvement in accuracy. Additionally, statistical analysis, including error histograms and correlation analysis, further confirms the robustness of the proposed models. Compared to traditional methods, the proposed AI-based models reduce computational complexity while maintaining superior accuracy. This study highlights the potential of AI in predictive maintenance and industrial optimization, paving the way for future enhancements in intelligent fouling estimation systems.

In the rapidly developing field of smart cities, accurately predicting weather conditions plays a vital role in various sectors, including industry, tourism, agriculture, social planning, architecture, and economic development. Unfortunately, the instruments used (such as pyranometers, barometers, and thermometers) often suffer from low accuracy, high computational costs, and a lack of robustness. This limitation affects the reliability of weather predictions and their application across these critical areas. This study proposes artificial neural network-multilayer perceptrons (ANN-MLPs). A dataset of 480 data points was used, with 80% allocated for the training phase, 10% for the validation phase, and 10% for the testing phase. The best results were obtained with the structure 6-17-1 (6 inputs, 17 hidden neurons, and 1 output neuron) to predict weather condition data in the Ghardaïa district. Weather conditions parameters include air temperature, relative humidity, wind speed, and cumulative precipitation. Results showed that the most relevant input factors are, in order of importance: earth-sun distance (DT-S) with a relative importance (RI) of 31.10%, factor conversion (d) with an RI of 26.05%, and solar radiation (SR) with an RI of 16.26%. The contribution of the elevation of the sun (HI) has an RI of 13.29%. The optimal configuration includes seventeen neurons in the hidden layer with a logistic sigmoid activation function and a Levenberg–Marquardt learning algorithm, resulting in a root mean square error (RMSE) of 3.3043% and a correlation coefficient (R) of 0.9683. The proposed model can predict both short- and long-term climate factors such as solar radiation, air temperature, and wind energy in areas with similar conditions.

This paper presents advanced techniques for air temperature forecasting based on the relationship between meteorological variables and feedforward neural network models aimed at improving prediction accuracy. An optimal single neural network model was developed, and optimization through dataset partitioning resulted in high predictive preformance, with all correlation coefficients exceeding 0.97 in the test phase and a root mean square error (RMSE) of 1.5587. Additionally, a bootstrap-aggregated neural network (BANN) model comprising 30 networks in a stacked ensemble was implemented, yielding robust results; both validation and testing correlation coefficients were close to the ideal value. Sensitivity analysis further revealed that true solar time, extraterrestrial radiation, and wind speed were the most influential predictors—an outcome that is consistent with their physical relevance to temperature variation. Moreover, the model’s validity was confirmed through applicability domain analysis using the Williams plot, which showed that over 97% of the predictions fell within acceptable error limits. A comparative evaluation against other existing models demonstrated the superior predictive performance of the proposed approach, reinforcing the potential of the BANN model for accurate air temperature forecasting under diverse meteorological conditions.

Accurate prediction of external torque in industrial robotic manipulators remains essential for ensuring precision, operational safety, and system adaptability. The present study proposes an optimized hybrid method that accurately estimates external torque using only joint-level data, addressing the limitations of conventional approaches that rely on additional torque sensors. The performed method is the support vector machine (SVM) and its parameters are optimized using particle swarm optimization (PSO). The proposed SVM-PSO is structured based only the signals of the position sensor that is existing at any industrial manipulator joint. The developed model utilizes three input parameters: the current joint position error, the previous joint position error, and the actual joint velocity. The experimental procedures involved acquiring data under controlled sinusoidal joint motion and some random collisions with the robot links, followed by statistical validation using Williams’ method to confirm the reliability of both training and testing datasets. The model demonstrated excellent performance, achieving a RMSE of 0.0985 and a R of 0.9952 during training. It maintained strong generalization performance during testing, with an RMSE of 0.1669 and a R of 0.986. The minimal discrepancy between estimated and actual external torque values affirms the method’s predictive reliability. When benchmarked against existing state-of-the-art methods, the proposed approach consistently outperformed competitors in both accuracy and robustness. The elimination of external torque sensors, coupled with the exclusive reliance on joint-level data, enables a practical and scalable framework suitable for real-time torque estimation in industrial robotic systems.