The scientific and research community has used Neural Designer widely.

Many articles use this program to support or even fully develop their research.

In all these instances, Neural Designer is used for different purposes.

You can download a free trial here.

Contents

Healthcare & Life Sciences

  1. PAD: A Pancreatic Cancer Detection based on Extracted Medical Data through Ensemble Methods in Machine Learning.
  2. Machine learning modeling for solubility prediction of recombinant antibody fragments in four different E. coli strains.

Energy & Environment

  1. Application of neural networks for prediction of doubly fed induction generator’s equivalent circuit parameters used in wind generators.
  2. Implication of Artificial Intelligence Approach On Groundwater Level.
  3. Adsorption chiller in a combined heating and cooling system: simulation and optimization by neural networks.
  4. N, S co–doped biocarbon for supercapacitor application: Effect of electrolyte concentration and modeling with an artificial neural network.

Chemistry & Materials Science

  1. Computerized Detection of JWH Synthetic Cannabinoids Class Membership Based on Machine Learning Algorithms and Molecular Descriptors.
  2. Flexural strength estimation of basalt fiber reinforced fly-ash added gypsum-based composites.
  3. Predicting the Properties of High-Performance Epoxy Resin by Machine Learning Using Molecular Dynamics Simulations.
  4. Statistical and Machine Learning-Driven Optimization of Mechanical Properties in Designing Durable HDPE Nanobiocomposites.

Engineering & Manufacturing

  1. Experimental Investigation on the Use of Molecular Sieve 4A as Desiccant in a Heatless Desiccant Air Dryer.
  2. Classification of Pb recovery affecting factors with artificial ıntelligence based algorithms.
  3. Modeling resilient modulus of fine-grained materials using different statistical techniques.
  4. A Novel Integration of Hyper-spectral Imaging and Neural Networks to Process Waste Electrical and Electronic Plastics.

Transportation & Urban Systems

  1. Predicting Travel Times of Bus Transit in Washington, D.C. Using Artificial Neural Networks.

Other Scientific Applications

  1. Key Features of the Performance of Chinese Musicians.
  2. Deep neural networks applications in the study of a geological indicator.
  3. Investigations into succinic acid fermentation.

Application of neural networks for prediction of doubly fed induction generator’s equivalent circuit parameters used in wind generators

Abstract:
Due to its merits, the doubly fed induction generator (DFIG) finds its application in sizeable grid-connected power systems. Its equivalent circuit parameter typically represents DFIG by stator and rotor winding resistance, reactance, and mutual reactance.

In this paper, the equivalent circuit parameters of DFIG are predicted using deep neural networks created and simulated with ‘Neural Designer ‘. This article uses Neural Designer to improve its predictive model with different algorithms available in the program, such as selection order and neuron selection features. This algorithm increases the accuracy of the neural network and optimizes the computational cost.

Source:

  • Ramakrishna(2021). Application of Neural Networks for Prediction of Double Fed Induction Generator’s Equivalent Circuit Parameters used in Wind Generators. IRJCS:: International Research Journal of Computer Science, Volume VIII, 23-31.

Computerized Detection of JWH Synthetic Cannabinoids Class Membership Based on Machine Learning Algorithms and Molecular Descriptors

Abstract:
The detection and discrimination of JWH synthetic cannabinoids and non-JWH cannabinoids from other compounds of forensic interest are extremely important. These compounds have very limited legal uses. Hence, there are very few studies about their pharmacological activity and toxicity. The most essential characteristic of any system screening for these drugs of abuse is its capacity to correctly recognize the class identity of the positive samples, which, according to the requirements for a forensic tool, should not be misclassified under any circumstances.

This paper uses Neural Designer to create a predictive model for detecting these substances. The researcher draws statistical conclusions with the help of model analysis utilities included in Neural Designer.

Source:

Experimental Investigation on use of Molecular Sieve4A as Desiccant in Heatless Desiccant Air Dryer

Abstract:
This research investigates the use of solid desiccant molecular sieve 4A in a heatless air dryer with a 10 CFM capacity to remove moisture from compressed air. Moisture in compressed air, even in small amounts, can cause damage to the final products. As a result, actively removing moisture becomes particularly desirable. This can be achieved by using a heatless desiccant air dryer.

Using Neural Designer, the researcher builds a neural network to forecast the dew point temperature in a heatless air dryer over an extended period. All input parameters, such as flow rate, relative humidity, pressure, etc., have been kept constant, and the model predicts the outlet dew point temperature over time.

The training and analysis of the model are performed using the Model Training and Testing analysis features from Neural Designer, which help to conclude the efficiency of this desiccant with great accuracy.

Source:

Predicting Travel Times of Bus Transit in Washington, D.C. Using Artificial Neural Networks

Abstract:
This study aimed to develop travel time prediction models for transit buses to assist decision-makers in improving service quality and patronage. Six months’ worth of Automatic Vehicle Location and Automatic Passenger Counting data for six Washington Metropolitan Area Transit Authority bus routes operating in Washington, D.C., was used for this study.

In this case, the researcher used a neural network to approximate Travel time for Washington buses, utilizing various models designed and trained in Neural Designer.

The main results of this study were accurate, with a Normalized Selection Squared Error between 0.04 and 0.26 in the different parts of the day.

Source:

Implication of Artificial Intelligence Approach On Groundwater Level

Abstract:
Groundwater is considered one of the most critical water resources for humans and the environment, but it is now declining at an even faster rate. Various models have been developed to analyze the hydrological aspects of water resources. However, due to the complexity of data acquisition and the extensive data requirements for physical models, it is challenging to model water resources problems.

The research aims to simulate and model groundwater levels and assess the impact of climate change on groundwater scenarios using Artificial Intelligence as an alternative approach over physics-based models. In this case, the optimal neural network architecture is obtained after several trials with a different number of layers and neurons using Neural Designer and its features.

In this paper, the impact of climate change on groundwater levels is studied using an artificial neural network approach, concluding that we are in an alarming situation.

Source:

Key Features of the Performance of Chinese Musicians

Abstract:
The article is devoted to the methods of teaching saxophone playing within the framework of two classical approaches to teaching. The purpose of the article is to identify the features of two classical approaches to teaching the saxophone within the French and American traditions from the point of view of the peculiarities of the perception of these two techniques by Chinese students.

Regarding learning to play the saxophone, it is necessary to consider the methodological differences in techniques and methods of sound production. At the same time, in the future, it is necessary to investigate the methodology of teaching saxophone playing from the perspective of a polystylistic approach, when the goal of training musicians is the ability to play all styles at an equal quality level.

In this paper, ‘Neural Designer’ is used for analytics purposes without using complex and non-friendly statistical programs like R. The analytical capacities of Neural Designer allow different types of researchers to perform analyses in a very easy, clear, and understandable way, even using tools like exporting charts or graphs.

Source:

Deep neural networks applications in the study of a geological indicator

Abstract:
The differences between shallow neural networks and deep neural networks are considered. Finally, data from the operational exploration of an open-pit mine are used to train different types of deep neural networks to predict a helpful indicator.

Neural Designer is used in this case to train several neural networks, compare their performance, and choose the best one for a case study using the different parameters provided by the program simply and concisely.

The research concludes that a two-layer neural network provides excellent performance and balance between accuracy and computational cost in this case. Moreover, a neural network with two two-layer networks with four neurons is the optimal solution because of the correlation with higher accuracy, lack of overtraining, and a simplified structure.

Source:

Classification of Pb recovery affecting factors with artificial ıntelligence based algorithms

Abstract:
Lead (Pb) is one of the most widely used non-ferrous materials. Many methods have been developed to produce lead metal from initial raw materials or secondary resources.

Moreover, the Artificial Neural Network structure was proposed with 4 inputs for predicting Pb recovery percentage. Four input parameters —temperature, NaCl concentration, time, and solid to liquid ratio—were selected based on physical considerations and experimental test results.

In this case, Neural Designer provides a complete and straightforward way to perform the training and analysis of the Neural Network and provides a Normalized Squared Selection Error of 0.2374. Moreover, this model provided by Neural Designer allows the researcher to conclude that :

  • There is a strong correlation between the Pb recovery amount and the Solid-liquid ratio and NaCl concentration
  • The study’s outcomes can be assessed by other artificial and mathematical systems for a better understanding of input effects on Pb recovery amounts.

 

This statistical analysis indicates that up to 98 % lead recoveries could be obtained at a low pulp density with considerable NaCl concentration on the laboratory scale. Compared with previous studies, these results yielded consistent findings regarding Pb leach efficiency.

Source:

  • Rusen, A., & Yildizel, S. A. (2018). Classification of Pb recovery effecting factors with artificial intelligence based algorithms. Journal of Engineering Research and Applied Science, 7(1), 811-817.

PAD: A Pancreatic Cancer Detection based on Extracted Medical Data through Ensemble Methods in Machine Learning

Abstract:
Pancreatic adenocarcinoma is one of the most dangerous types of cancer, and its diagnosis can be predicted by several indicators using a neural network.

This study’s data comes from Barts Pancreas Tissue Bank, University College London, University of Liverpool, Spanish National Cancer Research Center, Cambridge University Hospital, and the University of Belgrade. This data set comprises 590 samples, including 183 control samples, 208 samples of benign hepatobiliary disease (such as chronic pancreatitis), and 199 PDAC samples. The data is organized in a CSV file with 590 rows and 14 columns, detailing the various factors that influence PDAC development.

Neural Designer processes the dataset, and the researcher uses the dataset report utilities to pre-analyse the situation. Then, a division is made between Stages I and II, and Stages III and IV. After this, a model is made for each case to be predicted, and a conclusion is made using the tools provided by ‘Neural Designer ‘, such as calculating ROC curves and confusion matrices.

Source:

Flexural strength estimation of basalt fiber reinforced fly-ash added gypsum based composites

Abstract:
Gypsum and fiber-reinforced gypsum-based composites are widely used in many construction projects. Much research has been conducted on these types of composites; however, there are very limited literature studies focused on the estimation of the flexural strength of fiber-reinforced gypsum-based composites. There is no current study on the mechanical strength estimation of these types of composites.

An extensive database of experimental information, comprising 275 results, is utilized to build and train the model. The database includes 5 inputs: fly ash, basalt, superplasticizer contents, water-to-binder ratio, and slump values.

In this case, ‘ Neural Designer ‘ calculates a neural network model with a good correlation between the flexural strength and predicted data.

Source:

Machine learning modeling for solubility prediction of recombinant antibody fragments in four different E. coli strains

Abstract:
This article presents a comparative analysis of ANN and RSM in the statistical optimization of fermentation conditions for E. coli, focusing on the production of soluble recombinant scFv.

In this report, Neural Designer was used for training and optimizing the artificial neural network with input data from densitometry analysis, and the output is the soluble expression of antiEpEX-scFv.

Based on the results, they concluded that three terms: “post-induction time”, “concentration of inducer”, and “different strains” have a linear significant impact, where “post-induction time” is the most critical parameter.

Source:

Predicting the Properties of High-Performance Epoxy Resin by Machine Learning Using Molecular Dynamics Simulations

Abstract:
The performance of epoxy systems varies significantly depending on the composition of the base resin and curing agent. However, there are limitations in exploring numerous formulations of epoxy resins to optimize adhesive properties because of the expense and time-consuming nature of the trial-and-error process. Herein, molecular dynamics (MD) simulations and machine learning (ML) methods were used to overcome these challenges and predict the adhesive properties of epoxy resin.

Datasets for diverse epoxy adhesive formulations were constructed considering the degree of crosslinking, free volume, cohesive energy density, modulus, and glass transition temperature. The researcher used Neural Designer to train the model using these datasets and compare all the models developed with the mean squared error. Also, it’s employed to generate various statistical results, plots, and charts.

Source:

Adsorption chiller in a combined heating and cooling system: simulation and optimization by neural networks

Abstract:
Adsorption chillers, utilizing eco-friendly refrigerants, offer promising capabilities for recovering and utilizing low-grade waste heat, particularly renewable and waste heat at ambient temperatures…

The input variables are THin Inlet temperature of hot water, TMin Inlet temperature of recooling water, TLin Inlet temperature of ice-water, VH Volume flow rate of hot water, VL Volume flow rate of ice-water, VM Volume flow rate of recooling water. The target variables are THout, the Outlet temperature of hot water, and TLout, the Outlet temperature of ice water. Initially, the model undergoes training using Neural Designer. Following this, a statistical error analysis yielded encouraging results with a correlation of 0.91 for TLout and 0.89 for THout.

The model is an easy-to-use and powerful optimization tool for the adsorption chiller, integrated into a combined multigeneration system. Moreover, with model deployment features, this model could be used in different languages and programs like Python or C.

Source:

N,S co–doped biocarbon for supercapacitor application: Effect of electrolytes concentration and modeling with artificial neural network

Abstract:
The world energy crisis stems from excessive consumption and the unsustainable reliance on fossil fuel reserves, which has put undue pressure on these finite resources. As a result, researchers have directed concerted efforts on energy storage devices to curb this hazard. As a result, supercapacitors have attracted interest in recent years as competitive and robust energy storage systems.

In this case, the objective is to predict different electric properties of nitrogen and sulfur co-doped chicken bone-derived biocarbon: Capacitance and energy density. Input variables are Sulfur-doping, Nitrogen-doping, Electrolyte concentration, Electrolyte type, and S/N-co-doping.

The researcher used ‘Neural Designer’ for choosing and selecting the model parameters, i.e., the number of layers and neurons. As a result, this model achieves an R2 of 0.964 and an average desirability function of 0.96.

Source:

Investigations into succinic acid fermentation

Abstract:
Succinic acid (SA) is an essential chemical intermediate from which fine chemicals ( e.g., detergents), additives (for pharmaceuticals, food (taste), plant growth stimulants), as well as other essential intermediates (maleic anhydride, succinimide, 2-pyrrolidinone, dimethyl succinate), can be manufactured. Based on different fermentations, a neural network for modeling and describing how factors influence SA production was established.

Fermentation data were combined into a single MS Excel spreadsheet and exported to a tab-delimited text file, which could be imported into the Neural Designer software. 9 variables, i.e., 7 inputs (time, lactic acid, acetic acid, propionic acid, glycerol, ethanol, total sugar) and 2 outputs (succinic acid, CDW), were applied to 58 fermentation samples. It was also used ‘Neural Designer’ to create and improve the neural network architecture with the order selection algorithm and the algorithm “growing inputs”.

In this case, ‘Neural Designer’ was a key part of the study because it provides a simple and intuitive way to model this problem and draw conclusions.

Source:

Modeling resilient modulus of fine-grained materials using different statistical techniques

Abstract:
The main objective of this study is to use multiple linear regression (MLR), nonlinear regression, and backpropagation artificial neural network algorithms to develop models to predict the resilient modulus of fine-grained materials based on 3709 soil samples collected from the Long-Term Pavement Performance (LTPP) website.

Additionally, the critical input parameters selected for this study are: the confining pressure, nominal maximum axial stress, percent of silt, Liquid Limit, Plasticity Index, percent passing number 200 sieve, maximum dry density, percent of clay, optimum moisture content, and laboratory-determined resilient modulus.

In this case, the best approximation method is the Artificial Neural Networks simulated with the help of Neural Designer software, and it implements algorithms such as the order and input selection methods. Furthermore, the researcher uses the Directional Output utility to analyze the influence of different parameters in the resilient modulus.

Source:

Prediction of the Onset of Diabetes Using Artificial Neural Network and Pima Indians Diabetes Dataset

Abstract:
When a human body cannot respond to insulin appropriately or produce the required amount to regulate glucose, it is suffering from Diabetes.

The prediction model utilized a Three-Layered Artificial Neural Network and the Pima Indians Diabetes dataset. In this Neural network-based prediction model, a logistic-activation function is used for neuron activation, and the Quasi-Newton method is employed as the training algorithm.

Moreover, the features encompassing graphic visualization and statistical analysis, essential components of the paper, are inherently integrated into Neural Designer as default functionalities.

Source:

A Novel Integration of Hyper-spectral Imaging and Neural Networks to Process Waste Electrical and Electronic Plastics

Abstract:
This study introduced a technique combining hyperspectral imaging technology and a neural network-based algorithm to identify and separate different types of e-waste plastics (e-plastics). Although recent technological advancements in computing power enable the handling of big data relatively efficiently, the model must utilize a manageable number of neurons to perform real-time sorting applications for plastic recycling.

Before implementation, the neural network underwent simulation and training in Neural Designer. This software leverages the C++ library OpenNN as its foundation.

Their results have shown that with the five spectral bins, and a neural network with five input neurons, four hidden neurons, and three output neurons, they could classify three different types of plastic at a conveyor belt speed of 1 m/s, with 99% accuracy.

Source:

Statistical and Machine Learning-Driven Optimization of Mechanical Properties in Designing Durable HDPE Nanobiocomposites

Abstract:
The selection of nanofillers and compatibilizing agents, along with their size and concentration, is always considered crucial in designing durable nanobiocomposites with maximized mechanical properties (i.e., fracture strength (FS), yield strength (YS), Young’s modulus (YM), etc.). Therefore, the statistical optimization of the critical design factors has become extremely important to minimize the experimental runs and the cost involved.

In this study, both statistical (i.e., analysis of variance (ANOVA) and response surface methodology (RSM) and machine learning techniques. The researcher used artificial intelligence-based techniques (i.e., artificial neural network (ANN) and genetic algorithm (GA)) to optimize the concentrations of nanofillers and compatibilizing agents of the injection-molded HDPE nanocomposites. ANOVA initially identified that the concentrations of TiO2 and cellulose nanocrystals (CNCs) and their combinations significantly enhanced the durability of the HDPE nanocomposites. Subsequently, the data underwent modeling and prediction through RSM, ANN, and their combination with a genetic algorithm (i.e., RSM-GA and ANN-GA).

In this case, Neural Designer is responsible for creating and modeling the neural network and then using the genetic algorithm to optimize it. The researcher concluded this was the best method to predict the desired features.

Source:

Related posts

Training, validation and testing samples in machine learning What is machine learning? Machine learning tutorial: Model types Use Neural Designer in the cloud with Amazon Web Services (AWS)