Proposed Artificial Neural Network Method Research Articles

Wearable Loops for Dynamic Monitoring of Joint Flexion: A Machine Learning Approach

We present a machine learning driven system to monitor joint flexion angles during dynamic motion, using a wearable loop-based sensor. Our approach uses wearable loops to collect transmission coefficient data and an Artificial Neural Network (ANN) with fine-tuned parameters to increase accuracy of the measured angles. We train and validate the ANN for sagittal plane flexion of a leg phantom emulating slow motion, walking, brisk walking, and jogging. We fabricate the loops on conductive threads and evaluate the effect of fabric drift via measurements in the absence and presence of fabric. In the absence of fabric, our model produced a root mean square error (RMSE) of 5.90°, 6.11°, 5.90°, and 5.44° during slow motion, walking, brisk walking, and jogging. The presence of fabric degraded the RMSE to 8.97°, 7.21°, 9.41°, and 7.79°, respectively. Without the proposed ANN method, errors exceeded 35.07° for all scenarios. Proof-of-concept results on three human subjects further validate this performance. Our approach empowers feasibility of wearable loop sensors for motion capture in dynamic, real-world environments. Increasing speed of motion and the presence of fabric degrade sensor performance due to added noise. Nevertheless, the proposed framework is generalizable and can be expanded upon in the future to improve upon the reported angular resolution.

Optimization of waste tyre steel fiber and rubber added foam concretes using Taguchi method and artificial neural networks

This research presents a study on the waste tyre steel (WTS) and rubber particles (RP) added foam concrete using Taguchi method and Artificial neural networks (ANN) by revealing its potential utilization in construction industry. The fresh and hardened characteristics of foam concretes were experimentally tested. Twenty-five different samples were first prepared according to the Taguchi design and compressive and flexural strengths, dry density and thermal conductivity of the samples were optimized. A Quasi-Newton based ANN algorithm was proposed to obtain prediction equations for the flexural and compressive strength of foam concretes. ANOVA analysis was also conducted to determine the most important factors. The results showed that the combined utilization of WTS and RP decreases the harmful effect of RP on the strength properties of foam concretes. Taguchi optimization results were confirmed through validation tests. With the proposed ANN method, R2 values were obtained as 0.9775 for compressive strength, 0.9723 for flexural strength, 0.9654 for dry density and 0.9527 for the thermal conductivity measurements. Both Taguchi and ANN system were found to be suitable for the design and estimation of the responses.

Smart Control of Springback in Stretch Bending of a Rectangular Tube by an Artificial Neural Network

Abstract Springback is one of the factors that causes decreased product quality in tube bending; 2D and 3D stretch bending can be used to manufacture a complex geometry of a tube with springback reduction. For a nonlinear springback problem in tube bending, an artificial neural network (ANN) is an attractive data-driven approach to achieving springback prediction and control. The main objective of this paper is to control springback and improve geometrical quality with an ANN in 2D and 3D stretch bending of a rectangular tube. In general, an ANN is trained with collected data sets from a large number of experiments, causing expensive costs and time-consuming work. In the present work, the training data sets for the proposed ANN are obtained from both experiments and an analytical springback model. As the analytical model can adopt different bending angles, material properties, and geometries, supplementary data by the analytical model significantly reduced the number of experiments needed for ANN training. Contrary to the typical springback predictions, the proposed ANN synthesizes the machine settings based on the desired dimensions as the inputs. It is shown that springback can be controlled by specifying the bend angles provided by the ANN prediction. The proposed ANN method was validated in 2D and 3D stretch bending of a rectangular tube, and its prediction and control performance are favorably compared to an ANN trained with only experimental data sets.

Evaluation of artificial neural network and adaptive-network-based fuzzy inference system for ovarian and lung cancer prediction

Aims: Every year, a significant number of individuals lose their lives due to cancer or undergo challenging treatments. Indeed, the development of an effective cancer prediction method holds great importance in the field of healthcare. Methods: Machine learning methods have played a significant role in advancing cancer prediction models. In this context, this study focuses on exploring the potential of two machine learning methods: Artificial neural network (ANN) and adaptive-network-based fuzzy inference system (ANFIS) for cancer prediction. In this study, two different types of cancer, ovarian cancer and lung cancer, are taken into consideration. For the prediction of ovarian cancer, three specific biomarkers, namely human epididymis protein 4 (HE4), carbohydrate antigen 125 (CA-125), and carcinoembryonic antigen (CEA), are used to develop a prediction model. For the prediction of lung cancer, six different variables are utilized in the development of both the ANN and ANFIS methods. Results: The findings demonstrated that the proposed methods had an accuracy rate of at least 93.9% in predicting ovarian cancer. With an accuracy rate of at least 89%, the proposed methods predicted lung cancer. Also, the proposed ANN method outperforms the ANFIS method in terms of predictive accuracy for both ovarian cancer and lung cancer. Conclusion: This study suggests that the ANN method provides more reliable and accurate predictions for these specific cancer types based on the chosen variables or biomarkers. This study highlights the potential of machine learning methods, particularly ANN, in improving cancer prediction models and aiding in the early detection and effective management of ovarian and lung cancers.

Effect of vitamin D supplementation on inflammatory markers and total antioxidant capacity in breast cancer women using a machine learning technique.

This study aimed to establish a learning system using an artificial neural network (ANN) to predict the effects of vitamin D supplementation on the serum levels of vitamin D, inflammatory factors, and total antioxidant capacity (TAC) in women with breast cancer. The data set of the current project was created from women with breast cancer who were referred to the Shafa State Hospital of Patients with Cancers in Ahvaz city, Iran. Modeling was implemented using the data set at the serum levels of vitamin D, tumor necrosis factor-α (TNF-α), transforming growth factor β (TGF-β), and TAC, before and after vitamin D3 supplement therapy. A prediction ANN model was designed to detect the effects of vitamin D3 supplementation on the serum level changes of vitamin D, inflammatory factors and TAC. The results showed that the ANN model could predict the effect of vitamin D3 supplementation on the serum level changes of vitamin D, TNF-α, TGF-β1, and TAC with an accuracy average of 85%, 40%, 89.5%, and 88.1%, respectively. According to the findings of the study, the ANN method could accurately predict the effect of vitamin D3 supplementation on the serum levels of vitamin D, TNF-α, TGF-β1, and TAC. The results showed that the proposed ANN method can help specialists to improve the treatment process more confidently in terms of time and accuracy of predicting the influence of vitamin D supplementation on the factors affecting the progression of breast cancer (https://www.irct.ir/ identifier: IRCT2015090623924N1).

Intelligent Model for Avoiding Road Accidents Using Artificial Neural Network

Accidents typically occurred on roads, resulting in significant societal losses. Road accidents are a worldwide issue that result in the loss of precious human lives and property. The purpose of this paper is to create an intelligent system-based on Machine Learning model for avoiding road accidents, as well as a system that effectively reduces road accidents severities. The Artificial Neural Network (ANN) algorithm, along with others such as Logistic Regression (LR), Decision Tree (DT), K-Nearest Neighbor (KNN), Support Vector Machine (SVM), Nave Bayes (NB), Stochastic Gradient Descent (SGD), Random Forest (RF), Gradient Boosting (GB), and AdaBoost, is used to create an intelligence system. Many driving collaborator procedures, installed in a few vehicles, assist drivers in avoiding vehicle crashes by providing early cautioning messages. The intelligence road crash avoidance system model is built on dataset of 29 columns and 1048575 rows. Preprocessing, feature selection, and feature extraction performed with the help of heat map and correlation matrix are used to select features. Linear Discriminant Analysis (LDA) is used for feature extraction. The testing dataset revealed that the proposed ANN method outperforms other algorithms with an accuracy of 0.856. Intelligent systems aid in the prevention of traffic accidents, which aids police officers and researchers in developing new policies.

Fatigue damage prediction of top tensioned riser subjected to vortex-induced vibrations using artificial neural networks

Several marine structures are subjected to critical loads resulting in their fatigue failure. Among these structures, the riser is an essential component that transports the hydrocarbons and fluids from the platform to the subsea well and vice versa. Risers are subjected to vortex-induced vibrations (VIV) caused by current, leading to fatigue damage (Lekkala et al., 2022a; 2022b). Therefore, developing a simplified and valid approach to estimate fatigue behaviour and damage is necessary. Though semi-empirical methods are available for fatigue analysis, their deployments demand immense computational time and costs. In this study, we propose a novel framework based on data-driven analysis utilising artificial neural networks (ANN) to predict the VIV-induced fatigue damage of the top-tensioned riser (TTR). To build a reliable database, OrcaFlex, and SHEAR7, are used for the riser's modal and fatigue damage analysis due to VIV, respectively. A total of 24,686 riser models based on combinations of eight input parameters (outer diameter of riser, wall thickness of riser, water depth, surface, and bottom velocities, top tension, buoyancy modules thickness, and coverage ratio) and corresponding output as fatigue damage results are used for training a multi-layered ANN. The fatigue damage prediction surrogate model has been achieved by tuning the hyperparameters to improve the model's prediction accuracy. The findings of the ANN-based riser fatigue damage approach confirm that the ANN model exhibits a mean absolute percentage error (MAPE) between the predictions and truths from SHEAR7 is 9.29%. Also, the proposed framework can reduce computational time and cost compared to conventional semi-empirical methods. Finally, detailed parametric investigations are carried out using an optimised ANN model to evaluate the accuracy with the semi-empirical tool. These metrics suggest that the proposed ANN method could be used to determine the fatigue damage of risers in real-world field conditions.

Examination of insulation diagnosis in substation by neural network with phase‐resolved partial discharge pattern reconstruction

AbstractSeveral studies for partial discharge (PD) pattern recognition using artificial neural network (ANN) were reported in the early 1990s. Usually, in an actual field such as a substation, data on partial discharge is scarcely available, or even rare. In many cases, the power supply phase required for the PRPD pattern cannot be easily obtained. We propose an ANN method that shifts the phase in which the maximum signal intensity detected with PD sensors is generated and used it as training and input data, even for the few phases resolved PD data available in the field. This ANN method was applied to the PRPD pattern obtained in a practical field. As a result, it was shown that the discrimination rate between PD and noise was improved, and therefore the proposed ANN method was found to be effective.

Performance optimization of heat-exchanger with delta-wing tape inserts using machine learning

Nusselt number (Nu), friction factor (f), and thermal performance factor (η) are effective thermal parameters in determining the robustness of thermal management systems. However, a precise prediction of these parameters is a challenge due to complicated fluid and thermal behaviors of thermal systems, such as heat exchangers. In this study, we propose and develop a machine learning-based approach for predicting the heat transfer performance of a heat exchanger that employs delta-wing tape inserts. An aggregated database, containing 300 data points, is obtained from seven sources that include two working fluids. The wing-width ratio (w/W), pitch ratio (p/W), attack angle (α), Reynolds number (Re), and tube length (L) are in the ranges of 0.31 to 0.83, 0.95 to 1.65, 30° to 70°, 4,000 to 22,000, and 1,200 to 2,500 mm, respectively. Nu, f, and ηare predicted using an artificial neural network (ANN) model based on a universal aggregated database divided into training and test datasets. Optimization is performed, and an ANN model architecture is selected, consisting of input parameters (w/W, Re, α, p/W, and L) and hidden layers (10,10,10) that predict the test data with a mean absolute error (MAE) of 0.002393 for f, 1.821792 for Nu, and 0.00507 for η. The robustness of the developed ANN model is analyzed by precluding the databases from the training datasets together and is utilized to predict these excluded datasets. Furthermore, multiple linear regression analysis is adopted to measure and validate the efficiency of the proposed model. From the study results, it is evident that the proposed ANN method can provide valuable guidance to accurately predict the heat transfer characteristics with the lowest variance. Here, the 5–10–10–10-1 configuration is verified as the optimal configuration for the proposed model as it has the lowest MAE of 0.002393.

Improved accuracy of optical distance sensor based on artificial neural network applied to real-time systems

Optical time-of-flight sensors have potential in the revolution of distance measurement. These sensors can continuously monitor the distance and track the movement of objects. However, the existing sensing methods for such distance optical sensors mainly calculate the flight time, e.g. pulse transmission and receiving time, without considering the environmental effects. Therefore, the measurement accuracy is severely reduced. There are other technologies with higher accuracy in distance measurement. Nonetheless, they are too expensive due to the high accurate power supply. In this paper, we innovatively improve the accuracy of continuous distance measurement using the artificial neural network (ANN) technique. The proposed method can be applied for very cheap optical distance sensors with analog output in a real-time system. Moreover, the proposed method can self-calibrate and be miniaturized for cheap analog sensor applications. The prototype is built with the infrared sensor GP2Y0A02YK0F and an Arduino control board (ESP32_DevC), and the ANN is implemented using the deep learning algorithm. The test results show that the distance measurement accuracy is significantly improved and the measuring range is increased from 15 to 150 cm. In addition, we calculate mean squared error, mean absolute error, mean bias error, and R 2 for further performance evaluation. The experimental results have proven the superiority of the proposed ANN method in optical distance measurement. The proposed method can be applied to many types of sensors.

The Close Relationship between Internal Wave and Ocean Free Surface Wave

A numerical model was used to simulate the propagation of internal waves (IW) along the surface layer. The results show that strong water exchange during IW propagation results in strong free surface flow and produces small but distinct free surface waves. We found a close relationship between the internal and ocean surface waves. Our intuitive reaction is that by training the relationship between the water surface wave height and the internal wave waveform, the internal wave waveform can be reversed from the water surface wave height value. This paper intends to validate our intuition. The artificial neural network (ANN) method was used to train the Fluent simulated results, and then the trained ANN model was used to predict the inner waves below by the free surface wave signal. In addition, two linear internal wave equations (I and II) were derived, one based on the Archimedes principle and the other based on the long wave and Boussinesq approximation. The prediction by equation (II) was superior to the prediction of equation (I), which is independent of depth. The predicted IW of the proposed ANN method was in good agreement with the simulated results, and the predicted quality was much better than the two linear wave formulas. The proposed simple method can help researchers infer the magnitude of IW from the free surface wave signal. In the future, the spatial distribution of IW below the sea surface might be obtained by the proposed method without costly field investigation.

Estimating the State of Health of Lead-Acid Battery Using Feed-Forward Neural Network

A Battery Management System (BMS) can prolong the life of the battery but it depends on the accuracy of the adopted scheme. Different techniques have been developed to enhance the BMS by monitoring the State of Health (SOH) of the battery. In this paper, the detection of battery voltage is analyzed by using the cycle counting method, which is a conventional technique and compared with Artificial Neural Network (ANN), a heuristic method. The advantage of the proposed ANN method is that SOH can be monitored without disconnecting the battery from the load. Also, the sampling data to the ANN are derived from various techniques including Open Circuit Voltage (OCV) method, Ambient temperature measurement, and valley point detection. A feed-forward backpropagation algorithm is used to achieve the purpose of real-time monitoring of the LAB. The results show that the precise estimation of SOH can be obtained by Feed-Forward Neural Network (FFNN) when trained with more sampling data.

Road surface conditions forecasting in rainy weather using artificial neural networks

In general, drivers should be cautious while driving on rainy days because of risk factors such as slippery road conditions or hydroplaning. This study aims to forecast road surface conditions on rainy days using artificial neural networks (ANNs) to provide drivers with useful information that might help prevent traffic accidents arising from poor road conditions. The proposed model forecasts road condition as belonging to one of three categories—hydroplaning, wet, or moist—based on the friction coefficient. The collected data set is divided into two data sets, training and test, and all data are normalized to range from 0 to 1. In addition, the friction coefficient is also normalized. When the normalized friction values are in range of 0.67–1.00, 0.34–0.66, and 0–0.33, the road condition is categorized as hydroplaning, wet, and moist, respectively. The accuracy of the model is verified using statistical parameters such as the Correlation ratio (Cr), Mean Squared Error (MSE), and Root Mean Square Error (RMSE). Furthermore, contingency tables are constituted to evaluate the performance of the forecasts. The Chi-squared test is used to analyze the tables; results showed statistically significant. The observations indicated that the proposed ANN method predicts the road condition with high accuracy for each factor, specifically, 92%, 100%, and 78% for hydroplaning, moist, and wet, respectively. Implementing this model would help reduce traffic accidents on rainy days by providing drivers with road condition information through on-board equipment using V2X technology.

Simultaneous quantitative analysis of 3H and 14C radionuclides in aqueous samples via artificial neural network with a liquid scintillation counter

Liquid scintillation counters are common instruments used in the measurement of pure beta-emitting radionuclides, and while they represent a conventional radiometric technique, they are still competitive for their potential to measure multiple radionuclides simultaneously. In this work, we propose an algorithm based on an artificial neural network (ANN) for the simultaneous analysis of the beta-ray spectra of 3H and 14C in dual beta-labeled samples using a liquid scintillation counter. We achieved percentage deviations below 5.0% using the proposed algorithm in 16 out of 18 cases, with RMSDs below 1.5% in 17 out of 18 cases. The trained ANN also produced activity ratios with high accuracy even while having to deal with highly fluctuating spectra. Results demonstrate that the rapid predictions with a short measurement time from our proposed ANN method are compatible with the calculated ones from previous studies that were obtained with long measurement times.

An Estimation of hourly average solar radiation using artificial neural network in the city of Surabaya

To maximize the effectiveness of solar energy systems installed on the ship, the panels need to be positioned on their optimal inclination angle. To determine this angle, information on hourly average solar radiation is required in addition to position and direction data. However, the amount of hourly solar radiation is not continuously measured in Indonesia. This study proposed a method to estimate the hourly solar radiation for any month. This is based on Artificial Neural Network (ANN) with the month and the hour required for the estimate and the previous daily average solar radiation as the input. The proposed ANN method was validated by comparing the estimation results with the measured hourly average solar radiation in Surabaya from May to July 2020. Its effectiveness was affirmed with the coefficient of determination (R2) of 0.983 or higher for each of the three observed months.

Marshall stability and flow analysis of asphalt concrete under progressive temperature conditions: An application of advance decision-making approach

The behaviour of asphalt concrete mixtures is difficult to understand due to its complex nature under different loading conditions and environmental factors. For prediction, there is a need to find mathematical relations between multiple inputs and outputs using a simple and precise way. Recently, artificial neural networks (ANNs) has been widely used to study the mechanical parameters of asphalt concrete materials and its applications in civil engineering fields. This study presents the application of ANNs method for prediction of Marshall stability of asphalt concrete developed with two different types of aggregates based on mineralogy under four different testing temperatures ranging between 25 °C and 60 °C. The ANNs model established with six input variables including temperature, aggregate type, ultrasonic pulse velocity–time and space volume, unit volume of dry air, and saturated surface dry weight. The proposed model developed using six neurons in hidden layer for the prediction of experimental data. The feasibility of the proposed model checked in terms of root mean square error (RMSE) and coefficient of determination (R2). The R2 values found within range during both training (0.909–0.999) and validation phase (0.886–0.997) depending on estimated parameters. Moreover, the influence of different aggregate type has been investigated under varying temperatures conditions using the proposed ANNs method. The proposed model has shown the potential to understand the mechanical behaviour of sustainable asphalt concretes accurately under various temperature conditions.

A stereological correction method employing an artificial neural network for mineral liberation assessment of ore particles

Mineral liberation and surface exposure, which are important features of processing crushed ore particles in minerals, are generally assessed in terms of cumulative distributions (such as volumetric composition distribution). These distributions are, however, distorted by stereological bias when they are determined by two-dimensional methods such as microscopy or by scanning electron microscopes or energy dispersive X-ray based analyzers. Thus, in the present study, a stereological correction method using an artificial neural network (ANN) was developed. Through preliminary and parametric investigation of the ANN, the network was designed with 12 neurons in each of the input and output layers, and 400 neurons in a hidden layer, using distribution bin-frequencies as the input and output parameters. Particle models exhibiting 17,630 different patterns of internal mineral structure and particle shape were computed and used as training data for the ANN, and a high training level, with a correlation coefficient over 0.999, was obtained. Then, experimental validation was conducted using two- and three-dimensional data obtained by X-ray computed tomography involving artificial bi-phase particles. The method showed very high stereological correction performance in the case of samples that were well approximated by the computed training data, but lesser performance in the case of an unexpected sample distribution. In addition, through comparison with the authors’ previously developed texture method, the advantages and disadvantages of the proposed ANN method are discussed.

Predictions of electricity consumption in a campus building using occupant rates and weather elements with sensitivity analysis: Artificial neural network vs. linear regression

This study compares building electric energy prediction approaches that use a traditional statistical method (linear regression) and artificial neural network (ANN) algorithms. We investigate how significantly occupancy rates and local environmental conditions, such as temperature, humidity ratio, solar radiation, cloud type, and wind speed, impact the actual electric energy consumption of a campus building for both working and non-working days. To analyze the degree of impact of each input data type element, an impact value factor was applied to these data sets. The results illustrate that the ANN modeling was more accurate and stable than the linear regression method in predicting the electricity consumption for working days. By impact factor analysis for working days, occupancy rates were found to strongly dominate the electricity consumption in the building, while temperature and humidity also affected the results. However, there were no accuracy differences between the two models in predicting electricity consumption on non-working days because different data type elements had similar impact on the non-working day results. The two models—linear regression and ANN with a Levenberg–Marquardt Back Propagation (LM-BP) algorithm—were able to meet the long-term and real-time hourly prediction requirements for electricity consumption of an actual building based on occupancy rates and local environmental conditions. The analysis of the input element changes on a macroscopic scale is helpful in predicting how each element influences electric energy consumption in buildings with numerical impact factor. The proposed ANN method with LM-BP algorithm can be used as a reliable approach, compared with the linear regression modeling, for predicting the electricity consumption of a building.

A smart decision tool for the prediction of tunnel crown displacements

Artificial intelligence (AI) provides rigorous solutions to many engineering problems. In geotechnical engineering, AI tools are used mostly in deformation prediction and back analysis. In this paper, deformation monitoring data from measured absolute displacements in a tunnel in Greece, excavated at a complex geological system with fractured and loose cataclastic gouge, are used to train an artificial neural network (ANN) for the prediction of crown displacements along the tunnel. A Multi-layered perceptron neural network has been developed to be used as a quick smart tool for deformation behavior prediction (crown displacements) of the tunnel using the monitoring data measurements as target data and input training data from deformation parameters. The same deformation parameters are applied in finite element models (FEM) that simulate generalized sections of the tunnel from which a set of deformation results is obtained. A detailed description of the developed ANN is given and results are shown. Finally, the deformation results from the ANN tool, the FEM models, and the actual field measurements are provided and the potential of the proposed ANN method as a quick tool for tunnel deformation prediction is discussed.

Neural model of the expulsion fuse link Time–Current Characteristic for computer-aided applications

Due to its simplicity and low cost, one of the most common protection devices used in overhead distribution systems is the expulsion fuse links. Numerous published studies regarding distribution protection schemes ponder the existence of these devices. These studies represent the fuse element through digital models, when it is necessary to perform computational simulations. The purpose of the present paper is to model the Time–Current Characteristics (TCC) of expulsion fuse links through an Artificial Neural Network (ANN). To accomplish this objective, the Least Square Method was used in order to evaluate the results obtained, and the Residual Sum of Square (RSS) was used as a comparative parameter. The objective of using this parameter is to identify the models that best fit in the global scope, taking into account the different nominal ratings of current, speed rates and manufacturers. The proposed method was tested on an IEEE 34 node radial test feeder. The network was built using DIgSILENT PowerFactory™ software. The simulation results show that the proposed ANN method provides a robust and accurate technique for fuse link curve fitting when compared with the traditional double exponential and polynomial functions.