Ann model
Pro Research Analysisby 
Searched over 200M research papers for "ann model"
Applications and Enhancements of Artificial Neural Network (ANN) Models
Introduction to ANN Models
Artificial Neural Networks (ANNs) are computational models inspired by the human brain's neural networks. They are capable of identifying complex nonlinear relationships between input and output data sets, making them highly useful in various fields such as engineering, environmental science, and industrial applications .
ANN in Pharmaceutical Process Prediction
Predicting Mean Residence Time in Twin-Screw Granulation
In pharmaceutical manufacturing, ANNs have been effectively used to predict the mean residence time of formulations in twin-screw granulators. By correlating process parameters like feed flow rate, screw speed, and liquid-to-solid ratio, ANNs can build predictive tools that enhance the accuracy of these predictions. The integration of kriging interpolation with ANN models has shown significant improvements in prediction accuracy, with R² values increasing by 4% and 22% in training and validation phases, respectively, and RMSE improving by 40% and 61.5% .
ANN in HVAC System Optimization
Model Predictive Control (MPC) for Residential HVAC Systems
ANNs have also been applied in optimizing HVAC systems through model predictive control (MPC). By developing and calibrating ANN models for a residential house, researchers introduced a new algorithm called Best Network After Multiple Iterations (BNMI), which significantly improved prediction performance. The ANN-based MPC system was able to dynamically adjust temperature set-points, resulting in operating cost reductions of up to 73% depending on the season .
Enhancing ANN Prediction Performance
Addressing Over-Training and Extrapolation Issues
Despite their powerful capabilities, ANNs face limitations such as over-training and poor extrapolation. Over-training occurs when the model is allowed too many training iterations, while poor extrapolation happens when the training data does not cover the entire range of possible inputs. To mitigate these issues, practices such as limiting training iterations and using diverse training data sets have been introduced, significantly improving the prediction performance of ANN models .
Hybrid ANN Models for Time Series Forecasting
Combining Linear and Nonlinear Models
Hybrid models that combine linear models like Exponential Smoothing (ETS) with nonlinear models like ANN have shown superior performance in time series forecasting. By first applying ETS to capture linear patterns and then using ANN to model the residual errors, these hybrid models can effectively handle both linear and nonlinear components of time series data. Comparative studies have demonstrated that hybrid ETS-ANN models outperform traditional ARIMA and standalone ANN models .
ANN in Structural Engineering
Predicting Axial Compression Capacity of SCFST Columns
In structural engineering, ANNs have been used to predict the axial compression capacity (ACC) of square concrete-filled steel tubular (SCFST) columns. By training ANN models with extensive experimental data, researchers developed a new empirical formula that proved to be more accurate and stable than existing design codes. This ANN-based approach allows for the generation of master curves to predict ACC with higher precision .
Advantages of ANN Over Classical Modelling Techniques
Handling Non-Linear Relationships
ANNs have demonstrated significant advantages over classical modeling techniques like Response Surface Methodology (RSM) in handling non-linear relationships. In studies involving tablet compression, ANNs provided better data fitting and prediction capabilities, especially in recognizing important relationships that classical methods overlooked. This makes ANNs a valuable alternative for modeling complex, non-linear data sets .
Optimizing ANN Model Development
Comparing ANN Programs and Training Algorithms
The development of highly predictive ANN models depends on the choice of network architecture and training algorithms. Studies comparing different ANN programs and training algorithms have shown that while various programs can generate comparable models, the optimization of network architecture and training algorithms is crucial. Bayesian regularization, for instance, has been found effective in achieving stable and accurate models .
Conclusion
Artificial Neural Networks (ANNs) are versatile and powerful tools for modeling complex systems across various fields. By addressing limitations such as over-training and poor extrapolation, and by integrating hybrid approaches, the predictive performance of ANN models can be significantly enhanced. Whether in pharmaceutical manufacturing, HVAC system optimization, structural engineering, or time series forecasting, ANNs offer robust solutions for capturing and predicting intricate data patterns.
Sources and full results
Most relevant research papers on this topic
Application of ANN in predicting ACC of SCFST column
The artificial neural network (ANN) model effectively predicts the axial compression capacity of square concrete-filled steel tubular (SCFST) columns, providing a more accurate method than existing formulae.
Advantages of Artificial Neural Networks (ANNs) as alternative modelling technique for data sets showing non-linear relationships using data from a galenical study on a solid dosage form.
Artificial Neural Networks (ANN) methodology provides better data fitting and predicting abilities than Response Surface Methodology (RSM) for data sets with non-linear relationships in tablet compression studies.
DOI
Optimisation of the predictive ability of artificial neural network (ANN) models: a comparison of three ANN programs and four classes of training algorithm.
Different artificial neural network programs can produce comparable predictive models when optimizing their network architecture and training algorithms.
DOI