Hidden Layer Transfer Function Research Articles

Prediction of Cutting Forces Obtained through Cryo-Treated and Untreated Cutting Tools Using Optimum ANN Determined by Taguchi Design

This experimental and statistical study addresses the prediction of cutting forces by using the optimum Artificial Neural Network employed by Taguchi design. For this purpose, input and output transfer function and training algorithm were selected as control parameters, while Mean Square Error was chosen as output parameters for evaluating optimum ANN structure with S/N ratios. ANN structure was optimized through Taguchi L9 orthogonal design, which occurred 5 set-up for utilizing all training function. According to MSE values of S/N ratios, each set-up was compared with the obtained prediction of making values of cutting forces to the optimal result. For each set, the hidden transfer function, output transfer function and training function used in the optimal ANN structure were determined. The optimal ANN structure for cutting forces obtained in turning experiments were logsig transfer function in hidden layer, Tlm training function and pureline transfer function in output layer, while R square was at 0.999945. It was found that ANN based Taguchi orthogonal design was successful in evaluating the experimental results.

An intelligent model for predicting the dressed weight of pigs using morphometric measurements.

Determining the slaughter weight of pigs is crucial to the profitability of swine production farms. Unfortunately, in developing countries, the basic infrastructure for weight measurement may not always be available, affecting farmers' income. This study presents a machine learning-based approach to determine the dressed weight of pigs using four morphometric dimensions: paunch girth (PG), heart girth (HG), body length and wither height, which can be measured in situ. Different neural network model structures were constructed taking LM, GDX and BR training algorithms, tansigmoid/logsigmoid hidden layer transfer functions and 5-30 hidden layer neurons (HLNs). Results showed that LM training algorithm with logsigmoidal transfer function and 20 HLNs resulted in 99.8% accuracy in determining the pigdressed weight. Further, the number of morphometric parameters as inputs was gradually reduced and it was found that 99% accuracy can still be achieved using just PG and HG, thereby reducing the measurement time.

FORECASTING ELECTRICITY SUPPLIED IN TURKEY USING HOLT-WINTERS’ MULTIPLICATIVE METHOD AND ARTIFICIAL NEURAL NETWORK (ANN) MODELS

Electricity is one of the most essential necessities in today’s world and has an important role for the development of societies and economics. The need for electricity is expanding continuously due to increasing population, urbanization and industrialization. Hence, the purpose of this study was to develop the best model for forecasting electricity supplied in Turkey by applying the multiplicative Holt-Winters method and multilayer feed-forward neural network model. The monthly electricity supplied in Turkey from January 2000 until December 2019 were obtained from monthly electricity statistics report presented by the International Energy Agency (EIA). The data were divided into two sets comprising in-sample data from January 2000 until December 2015 and out-sample data from January 2016 to December 2019. The multiplicative Holt-Winters was used since the electricity supplied in Turkey exhibit trend and seasonal gave the out-sample forecast of 3.6990. The best multilayer feed forward neural network (MFFNN) model with three input lag variable, one hidden node, one output node, sigmoid transfer function in hidden layer and linear transfer function in output layer gave the out-sample forecast of 2.1483. Hence it can be concluded that, the multilayer feed-forward neural network model is more accurate than multiplicative Holt-Winters method to forecast the electricity supplied in Turkey.

Modeling and optimization of small-scale NF/RO seawater desalination using the artificial neural network (ANN)

The performance of seawater hybrid NF/RO desalination plant including permeate conductivity; permeate flow rate and permeate recovery. Under different feed parameters time, inlet temperature, inlet pressure, inlet conductivity and inlet flow rate were modelled by Artificial Neural Network (ANN) back-propagation based on Levenberg– Marquardt training algorithm. The optimal ANN model had a 5-8-3 architecture with a hyperbolic tangent transfer function in hidden layer and linear transfer function at the output layer. The ability of ANN performed model was compared with multiple linear regression (MLR). The results show that MLR is not satisfactory for predicting the performance of NF/RO hybrid desalination process with a correlation coefficient about 0.6. The trained ANN model has presented a good agreement between the prediction and the experimental data during the training with reasonable statistical metrics values (RMSE, MAE and AARD). The coefficient of determination values for the prediction of permeate conductivity, permeate flow rate and recovery by ANN were 0.969, 0.942, and 0.963, respectively. Therefore, the ANN model can successfully predict the performance of NF/RO hybrid seawater desalination plant.

Experimental Investigations and Modeling of Tool Wear in Gun-Drilling Process of 37Cr4 Forge Steel, Using Artificial Neural Network with Taguchi Method

In this study, the tool wear in gun-drilling process of 37Cr4 forge steel was studied. So, the Taguchi method of design of experiments was used to investigate the effects of inputs (tool tip angles (internal and exterior cutting edge angles, free angle of inner and outer cutting edge) and feed rate) on tool wear as output, experimentally. Also, the feedforward backpropagation neural network was used to predict the tool wear in gun-drilling process considering internal and exterior cutting edge angles, free angle of inner and outer cutting edge and feed rate as input parameters of the neural network. In order to select the best performance network with minimum mean squared error in predicting the tool wear in gun-drilling process, the Taguchi method of design of experiments was used. Once again, the input parameters of this design were the number of neurons in hidden layer, type of training function and transfer function of hidden layer of neural network, and mean squared error obtained from neural network run was the output parameter. According to the results, increasing internal and exterior cutting edge angles leads to higher tool wear, while increasing free angle of inner and outer cutting edge decreases tool wear. The tool wear increases by increasing tool feed rate at the first steps, but it decreases by further increase in feed rate. Also, the mean squared error of the neural network with the best performance to predict the tool wear was 0.000433, the most differences of the network and experimental results were 0.096 µm and the regression R value of the network with the best performance was 0.98468.

Modeling callus induction and regeneration in an anther culture of tomato (Lycopersicon esculentum L.) using image processing and artificial neural network method.

Doubled haploids, subsequent to haploid induction, have wide range of applications in basic and applied plant studies. Various parameters can affect the efficiency of haploid induction through an anther culture of tomato. The hybrid system of image processing-artificial neural network (ANN) was used to better understand callus induction and regeneration in an anther culture of tomato. The effect of parameters such as plant genotype, the concentrations of 2,4-dichlorophenoxyacetic acid (2,4-D) and kinetin (Kin) plant growth regulators, the concentration of gum arabic (GA) additive, the cold pretreatment duration, and flower length on callus induction percentage and number of regenerated calli in an anther culture of tomato were studied using multiple linear regression (MLR) and ANN models. The precise flower bud length was measured using an image processing technique. The 4',6-diamidino-2-phenylindole (DAPI) analysis showed that the flowers with 5-6.9mm length had the highest percentage of the mid- to late-uninucleate microspore stage. The best ANN model for both callus induction percentage and number of regenerated calli was a model with one hidden layer, 12-15 neurons in the first hidden layer, Levenberg-Marquardt learning algorithm, and Tan-Sigmoid transfer function in hidden layer, based on the root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R2) statistics. The scatter plot of measured values versus the predicted values showed the superiority of the ANN to MLR model to predict the callus induction percentage in an anther culture of tomato. The sensitivity analysis of MLR and ANN models revealed the plant genotype and 2,4-D concentration as the most important factors affecting both callus induction percentage and number of regenerated calli. Since tomato is a recalcitrant plant to androgenesis-based pathway of haploid induction, therefore the results of the present study can be helpful to develop an efficient haploid induction protocol in tomato through an anther culture pathway.

ANNs in ABC Multi-driver Optimization Based on Thailand Automotive Industry

The purpose of this research was to develop a method for Activity Based Costing (ABC) that provided accurate product production costs. ABC using Single Driver Activity Based Costing (SDABC) can result in distortion of the cost. A more accurate ABC cost calculation based on multiple cost drivers (CDs) in each activity has been devised and proven by considering the various cost drivers using the correlation coefficient or R2. The application of artificial neural networks (ANNs) to choose the CDs is Multiple Drivers Activity Based Costing (MDABC). The ANNs choose the CDs by algorithms including Multilayer Perceptron and Back-propagation. The transfer function for hidden layers is the Log-Sigmoid Function and for the output layer is the Pure Linear transfer function. The results have demonstrated that using MDABC results in more accurate cost calculations than when using SDABC. The study found that both of the extended ABC method, SDABC and MDABC provide more accurate actual cost of production, and both are applicable to products with low turnover or those in a state of loss condition. However, MDABC is better used in situations which include a variety of production activities, while the SDABC method is best used in situations of the factory operations not being very complex. Overall, the resolution, or accuracy, of the calculated production costs is better using the MDABC method, but is more complicated in its use and operation. Computer-based ANNs overcome this problem of complexity.

Refrigerant flow through electronic expansion valve: Experiment and neural network modeling

Electronic expansion valve (EEV) plays a crucial role in controlling refrigerant mass flow rate of refrigeration or heat pump systems for energy savings. However, complexities in two-phase throttling process and geometry make accurate modeling of EEV flow characteristics more difficult. This paper developed an artificial neural network (ANN) model using refrigerant inlet and outlet pressures, inlet subcooling, EEV opening as ANN inputs, refrigerant mass flow rate as ANN output. Both linear and nonlinear transfer functions in hidden layer were used and compared to each other. Experimental data from multiple sources including in-house experiments of one EEV with R410A were used for ANN training and test. In addition, literature correlations were compared with ANN as well. Results showed that the ANN model with nonlinear transfer function worked well in all cases and it is much accurate than the literature correlations. In all cases, nonlinear ANN predicted refrigerant mass flow rates within ±0.4% average relative deviation (A.D.) and 2.7% standard deviation (S.D.), meanwhile it predicted the EEV opening at 0.1% A.D. and 2.1% S.D.

Wind Power Prediction Model Based on Wavelet Neural Network under Missing Data

Wind power prediction is a key problem in optimizing power dispatching. This paper builds a wind power prediction model based on wavelet neural network which substitutes wavelet basis function for the transfer function of hidden layer. A missing data interpolation strategy is also given to improve the applicability of the model. With the wind farm data from southeast coast, the model works and the wind power in the next 30 hours is predicted. In the sense of the mean square errors this paper compared the prediction results of the model and BP neural network model, the results shows the models have a better accuracy.

Comprehensive Assessment Grade of Air Pollutants based on Human Health Risk and ANN Method

China has currently established a single index method to assess ambient air quality based on human health. However, there have no unified standards in comprehensive assessment methods. In order to solve this problem, we tried to explore a comprehensive assessment method based on human health risk by using AQI grading standard. By using sorting algorithm and RBF model of ANN method provided by MATLAB tools, we assessed air quality of six districts in Beijing 2013 to 2014. Firstly, Gauss function was chosen to be transfer function of hidden layer in RBF network. Secondly, extended samples obtained by interpolation method were used to train RBF network. Thirdly, a classified assessment results were output by RBF network. Results showed RBF network get a good effects on classified evaluation, it enhanced the information resolution of pollutants and helps understanding the real effect of pollutants on human health. According to the assessment results, we divided comprehensive assessment grade into five grades, which was lower by one grade than AQI.

Parameter Sensitivity Analysis of Artificial Neural Network for Predicting Water Turbidity

The present study focuses on the discussion over the parameter of Artificial Neural Network (ANN). Sensitivity analysis is applied to assess the effect of the parameters of ANN on the prediction of turbidity of raw water in the water treatment plant. The result shows that transfer function of hidden layer is a critical parameter of ANN. When the transfer function changes, the reliability of prediction of water turbidity is greatly different. Moreover, the estimated water turbidity is less sensitive to training times and learning velocity than the number of neurons in the hidden layer. Therefore, it is important to select an appropriate transfer function and suitable number of neurons in the hidden layer in the process of parameter training and validation. Keywords—Artificial Neural Network (ANN), sensitivity analysis, turbidity.

Machine Learning Elman Technique for Predicting Shelf Life of Burfi

Elman artificial neural network single and multilayer computerized models were developed for predicting the shelf life of burfi stored at 30o C. The experimental data of the product relating to moisture, titratable acidity, free fatty acids, tyrosine, and peroxide value were taken as input variables, and overall acceptability score as output variable for developing the models. Bayesian regularization algorithm was applied as training algorithm for neural network. Transfer function for hidden layers was tangent sigmoid; while for output layer it was pure linear function. Elman model with a combination of 5101 and 5771 performed exceedingly well for predicting the shelf life of burfi.

Drying of black cumin (Nigella sativa) in a microwave assisted drying system and modeling using extreme learning machine

Drying characteristics of Black cumin seeds (BCs) (Nigella sativa) with initial moisture content 58.14% (d.b) was investigated in microwave assisted drying system. The experiments were carried out in two aspects which were the BCs drying with different temperatures (35, 40 and 50 °C) and different microwave power levels (250, 500 and 750 W). The results showed that the drying rates of BCs have high efficiency with drying temperatures at constant microwave power level. Furthermore, in present study, the applications of Extreme Learning Machine (ELM) and Artificial Neural Network (ANN) for predicting the moisture ratio (MR) (output feature for ELM modeling) were investigated. Microwave temperature, microwave power and drying time were input layer features for the modeling. An ELM model by 93 neurons with Sine transfer function in hidden layer was selected. The results revealed that a network with the Sine function made the most accurate prediction for the BCs drying system. For BCs all data set, maximum R2 and minimum RMSE (root mean square error) were found as 0.9987 and 0.0123, respectively.

Dynamic Scientific Method for Predicting Shelf Life of Buffalo Milk Dairy Product

Feedforward multilayer machine learning models were developed to predict the shelf life of burfi stored at 30 o C. Experimental data of the product relating to moisture, titratable acidity, free fatty acids, tyrosine, and peroxide value were input variables, and the overall acceptability score was the output. Bayesian regularization algorithm was used for training the network. The transfer function for hidden layers was tangent sigmoid, and for the output layer it was purelinear function. The network was trained with 100 epochs, and neurons in each hidden layers varied from 3:3 to 20:20. Excellent agreement was found between the actual and predicted values establishing that feedforward multilayer machine learning models are efficient in predicting the shelf life of burfi.

Dynamic Scientific Method for Predicting Shelf Life of Buffalo Milk Dairy Product

Feedforward multilayer machine learning models were developed to predict the shelf life of burfi stored at 30 o C. Experimental data of the product relating to moisture, titratable acidity, free fatty acids, tyrosine, and peroxide value were input variables, and the overall acceptability score was the output. Bayesian regularization algorithm was used for training the network. The transfer function for hidden layers was tangent sigmoid, and for the output layer it was purelinear function. The network was trained with 100 epochs, and neurons in each hidden layers varied from 3:3 to 20:20. Excellent agreement was found between the actual and predicted values establishing that feedforward multilayer machine learning models are efficient in predicting the shelf life of burfi.

The Classification Research of QuickBird Image Based on Variogram Texture and BP Artificial Neural Networks

Using the principal component analysis method of QuickBird image obtained the information of the first and the second principal component was 67.09% and 31.73% respectively,totaling 98.82%.Therefore,the first and the second principal component were selected to extract texture.We used the absolute variogram to extract the texture,the window size is 3×3 and the step size is 1.The land type was divides into four types,it is building,water body,vegetation and bare land.The topological structure of BP artificial neural network is 4-4-1,the hidden layer transfer function is S(logsig) and the output layer transfer function is linear function,the training function is Trainscg.Adopted the neural network to classified QuickBird multispectral image and the principal components texture image,also figure out the preceding kind of P=84.54%,Kappa coefficient K=78.50%;later kind of P=89.46% and Kappa coefficient K=85.29%.The results demonstrated that the image classifications were enhanced after texture was added.

Evolving transfer functions for artificial neural networks

The paper describes a methodology for constructing transfer functions for the hidden layer of a back-propagation network, which is based on evolutionary programming. The method allows the construction of almost any mathematical form. It is tested using four benchmark classification problems from the well-known machine intelligence problems repository maintained by the University of California, Irvine. It was found that functions other than the commonly used sigmoidal function could perform well when used as hidden layer transfer functions. Three of the four problems showed improved test results when these evolved functions were used.