Search

Tensile properties of multifilament polyester woven fabrics are of great importance for their end uses such as parachutes, sails, tents, sleeping bags, filters and surgical textiles. The filament fineness, weave type and weave density have a great influence on the tensile properties of these fabrics. In this study, artificial intelligence (AI) models such as artificial neural networks (ANN), fuzzy logic (FL) and genetic algorithms (GA) were developed to forecast breaking strength and breaking elongation values of multifilament polyester woven fabrics. The fabric samples used in the study have three different microfilament finenesses and two different conventional filament finenesses with plain, twill and satin weave types. By applying four different weft density values, totally 60 woven fabric samples were obtained in the experimental design. The regression coefficient values ( R 2 ) between actual and predicted results were obtained as 0.80, 0.90 and 0.92 with ANN, FL and ANN–GA hybrid methods, respectively. The mean absolute percentage error (MAPE) was lower than 6% for all AI techniques used in this study. As a conclusion, it was proved that the breaking strength and breaking elongation properties of multifilament polyester woven fabrics can be forecasted with high accuracy rates by AI techniques. [ABSTRACT FROM AUTHOR]

A lead-free Ni–B–W (ENB-W) coating deposited by electroless method was investigated in present work. Heat-treated ENB-W coating (450 °C for 3 h) were exposed to tribological tests within a range of parameters, including load (10–50 N), speed (0.5–1.5 m/s), and distance (400–1000 m). Prior to tests, coated specimens before and after heat-treated condition were characterized. The typical globular morphology and crystalline nature was seen. The hardness, scratch hardness and first critical load of failure improved in heat treated condition. To optimize wear rate and coefficient of friction (COF), an approach has been introduced, i.e., integration of an artificial neural network (ANN) with a genetic algorithm (GA). The ANN model yielded a mean absolute percentage error of 8.4421% for predicting wear rate and 2.4138% for predicting COF. Pareto front analysis identified the optimal operating conditions as load of 34.2099 N, speed of 0.5006 m/s, and distance of 726.7243 m, resulting in both a minimum wear rate of 0.1003 × 10−8 g N−1 m−1 and a COF of 0.2099. In addition to optimization, the study also involved the characterization of the wear mechanisms of the coatings to gain a deeper understanding of their tribological behaviour. Different wear mechanisms were seen at different range of the parameters. The wear rate and COF did not vary significantly with sliding distance except at 10 N load. In fact, the wear rate was lower at 30 N and 50 N compared to 10 N. [ABSTRACT FROM AUTHOR]

The increasing demands for precision and efficiency in machining processes, driven by advancements in aerospace, biomedical, and automotive industries, have led to a growing need for highly accurate machined parts. Stainless steel 316, a commonly used material in these sectors, presents specific challenges due to its intended applications, particularly in biomedical and aerospace fields. Electric discharge machining (EDM) is a favored method for working with this alloy. However, EDM has inherent challenges, such as dimensional overcuts, which have limited its applicability. To address these issues, the potential of three different electrode materials, namely, copper (Cu), brass, and aluminum (Al), has been extensively explored. Additionally, the choice of the optimal dielectric is crucial as it directly affects the heat input to the electrode, influencing the melting and vaporization of the tool wear. In this context, the inclusion of graphene as an additive has been explored to minimize radial overcuts. It is worth noting that these concerns have not been comprehensively addressed before. The experimental design by Taguchi has been employed for the research, and the results indicate that the performance of the Cu electrode surpasses that of other dielectrics. The artificial intelligence-based artificial neural network (ANN) was constructed to predict the values of OC. It was found that in the given dataset, the R2 has a value greater than 0.99 for all possible training and validation. Based on the ANN, a multi-objective optimization through genetic algorithm (MOGA) was performed. It was found that the dimensional accuracy achieved by Cu electrode was 70.51% better than the highest OC value given by same electrode. Moreover, the OC suggested by the MOGA for brass and Al electrodes was 40.21% and 34.37% better compared to the highest values of radial overcut obtained during the actual experimentation. [ABSTRACT FROM AUTHOR]

Wastewater surveillance has recently emerged as a valuable tool for environmental and public health monitoring. By analyzing the constituents and biomarkers present in wastewater, stakeholders can gather critical information regarding contamination events and disease outbreaks. However, little attention has been given to the crucial question of where to collect water quality samples or place water quality sensors to maximize the usefulness of wastewater surveillance data. To address this gap, this study introduces a novel framework for sensor placement (SP) optimization in sewer networks. The objective of the optimization is to maximize both the observability and reliability of source identification (SI) under different scenarios. To achieve this objective, a machine learning–based SI model was integrated within the SP optimization framework. The SI model features a multilayer perceptron neural network model that was trained to forecast concentrations at various sensor locations, which were then propagated into a genetic algorithm that finds the optimal sensor network design that maximizes SI performance. The capabilities of the SP framework were demonstrated in a case study featuring a real-life, midsize sewer network. The SP framework was applied to multiple scenarios, including optimal design of a sensor network comprising one or more sensors, as well as optimal extension of existing sensor networks. The results showed that a clear trade-off exists between the sensor network's observability and reliability, highlighting the importance of considering both metrics for SP optimization. Overall, this study offers a practical approach for SP optimization to improve environmental and public health monitoring in a variety of contexts. [ABSTRACT FROM AUTHOR]

Deep Neural Networks (DNNs) are critical for modern intelligent processing but cause significant latency and energy consumption issues on mobile devices due to their high computational demands. Moreover, different tasks have different accuracy demands for DNN inference. To balance latency and accuracy across various tasks, we introduce PreDNNOff, a method that offloads DNNs at a layer granularity within the Mobile Edge Computing (MEC) environment. PreDNNOff utilizes a binary stochastic programming model and Genetic Algorithms (GAs) to optimize the expected latency for multiple exit points based on the distribution of task inference accuracy and layer latency regression models. Compared to the existing method Edgent, PreDNNOff has achieved a reduction of about 10% in the expected total latency, and due to the consideration of different tasks' varying requirements for accuracy, it has a broader applicability. [ABSTRACT FROM AUTHOR]

The existing nodes of IoT networks are very small in size, deployed for long periods, and have very limited resources, which means that an IoT network must be very energy efficient to survive for a long time. Therefore, finding optimal routing techniques that lead to better data sharing without wasting energy can lead to more energy savings. This is an optimization problem, which means that we need to use optimization algorithms to find the optimal path in an IoT network. Some of the optimization algorithms are called meta-heuristic algorithms, these algorithms are inspired by nature, such as Artificial Neural Networks (ANN), which are Gradient methods to find the most suitable solution for a given problem. Our next algorithm is Particle Swarm Optimization (PSO). If the search combination of both algorithms is used in parallel, the search power will increase and better answers will be found in less time. For this reason, we suggest using a combination of the above algorithms. This idea is a combination of two optimization algorithms, PSO (Particle Swarm Optimization) and ANN (Neural Network) to optimize routing and reduce latency when sending information between IoT nodes in an IoT system. The proposed protocol is focused on optimizing energy consumption and execution time with the help of the GA-PSO algorithm based on routing-based clustering. Finally, to evaluate the proposed protocol, it was simulated using C++ software and compared with the method presented in the reference article based on the enhanced Ant Colony Algorithm, and the results show the efficiency of the proposed method in terms of energy consumption and execution time. The results show that in the presented algorithm, the execution time has been reduced to almost a quarter of the execution time in the algorithm of the reference article. Also, the results showed that our proposed method consumed 20 kJ less energy. [ABSTRACT FROM AUTHOR]

The blades of an axial fan are optimized using artificial neural networks and genetic algorithms in this paper. In first, a parametric axial fan blade model is established with constraints imposed on several parameters. The chord length, maximum camber, maximum camber position, blade thickness, and airfoil stagger angle are considered as an optimization parameter of axial fan. The static pressure efficiency and static pressure of axial fan are regarded as the optimization objectives. An optimization calculation of an axial fan blade is carried out based on the combination of artificial neural network and genetic algorithm. The objective aim of optimization is to improve the static pressure efficiency, the static pressure of axial fan and to reduce the flow loss of axial fan. Numerical results of axial fan demonstrate that the pressure distribution gradient and turbulent kinetic energy contour maps of the optimized axial fan are effectively suppressed within the impeller region compared with that of original axial fan. Furthermore, the internal flow stability of the optimized axial fan also is significantly improved by studying the pressure fluctuation and the Fast Fourier Transform (FFT) of pressure fluctuation. Experimental results of axial fan aerodynamic performance further demonstrate that the static pressure of the optimized axial fan rises as much as 90.93 Pa and the improved static pressure efficiency is effectively improved as much as 7.43% at the design flow rates compared with that of the original axial fan. The application of optimized axial flow fans is of great significance in energy-saving of energy equipment. [ABSTRACT FROM AUTHOR]

The neural network technology combining genetic algorithm is utilized to predict the sulfur content and optimize the desulfurization operation at the end of the refining process. Three types of prediction models are developed to achieve the optimal model. The prediction accuracy can be improved by the application of the deep neural network while the root means square error (RMSE) value of the optimal prediction model and the mean absolute error (MAE) value are less than 5 ppm. Moreover, the proportion of heats with prediction errors less than 5 ppm reaches 82%. Effects of dissolved oxygen contents, initial sulfur contents, carbon contents, and the amount of desulfurizer addition on the desulfurization process are considered. The optimal amount of slag addition with various initial sulfur contents is calculated. With the increase of initial sulfur content in the molten steel, the optimal amount of slag‐modified agent addition increases from about 500–750 kg. [ABSTRACT FROM AUTHOR]

Optimizing extraction conditions can help maximize the efficiency and yield of the extraction process while minimizing negative impacts on the environment and human health. For the purpose of the current study, an artificial neural network (ANN) combined with a genetic algorithm (GA) was utilized for that the extraction conditions of Hypericum spectabile were optimized. In this particular investigation, the main objective was to get the highest possible levels of total antioxidant status (TAS) for the extracts that were obtained. In addition to this, conditions of the extract that exhibited the maximum activity have been determined and the biological activity of the extract that was obtained under these conditions was analyzed. TAS values were obtained from extracts obtained using extraction temperatures of 30–60 °C, extraction times of 4–10 h, and extract concentrations of 0.25-2 mg/mL. The best model selected from the established ANN models had a mean absolute percentage error (MAPE) value of 0.643%, a mean squared error (MSE) value of 0.004, and a correlation coefficient (R) value of 0.996, respectively. The genetic algorithm proposed optimal extraction conditions of an extraction temperature of 59.391 °C, an extraction time of 8.841 h, and an extraction concentration of 1.951 mg/mL. It was concluded that the integration of ANN-GA can successfully be used to optimize extraction parameters of Hypericum spectabile. The total antioxidant value of the extract obtained under optimum conditions was determined as 9.306 ± 0.080 mmol/L, total oxidant value as 13.065 ± 0.112 µmol/L, oxidative stress index as 0.140 ± 0.001. Total phenolic content (TPC) was 109.34 ± 1.29 mg/g, total flavonoid content (TFC) was measured as 148.34 ± 1.48 mg/g. Anti-AChE value was determined as 30.68 ± 0.77 µg/mL, anti-BChE value was determined as 41.30 ± 0.48 µg/mL. It was also observed that the extract exhibited strong antiproliferative activities depending on the increase in concentration. As a result of LC-MS/MS analysis of the extract produced under optimum conditions in terms of phenolic content. The presence of fumaric, gallic, protocatechuic, 4-hydroxybenzoic, caffeic, 2-hydoxycinamic acids, quercetin and kaempferol was detected. As a result, it was determined that the H. spectabile extract produced under optimum conditions had significant effects in terms of biological activity. [ABSTRACT FROM AUTHOR]

In the present investigation, process parameters were optimized in order to enhance the reduction of calcium oxide (CaO) from sugarcane molasses using acid centrifugation treatment. To predict the effects of process factors on CaO reduction efficiency, a response surface approach with a central composite design was selected. The polynomial quadratic equation was used to predict CaO removal efficiency, and the analysis of variance (ANOVA) test was utilized to assess the relevance of process factors. The appropriateness of the developed model was determined by regression analysis, which yielded a higher R‐squared value of 0.99334 ± 0.01. At the optimum process parameters of 100°C temperature, 50°Bx, and 3.50 pH, the CaO clarification efficacy of 66.17 wt.% was achieved. The experimental results indicated that for acidic centrifugation treatment, the experimentally observed CaO reduction of 65.94 wt% is in close agreement with the model equation's predicted maximum CaO reduction of 66.17 wt% with a t‐test value of 0.497726. Under such conditions, 0.982 wt.% CaO sugarcane molasses was obtained, which is low when compared to the world average of 1.5% CaO content of sugarcane molasses. Furthermore, the implementation of an artificial neural network (ANN) provided a better prediction model for CaO reduction, with a substantial R‐squared value of 0.99866. However, the genetic algorithm (GA) optimization resulted in an actual CaO reduction of 66.21 wt.% with a t‐test value of 0.497726. [ABSTRACT FROM AUTHOR]

This paper proposes a solution to enhance and compare different neural network (NN)-based side-slip angle estimators. The feed-forward neural networks (FFNNs), recurrent neural networks, long short-term memory units (LSTMs), and gated recurrent units are investigated. However, there is a lack in the selection criteria of the architectures' hyper-parameters. Therefore, the genetic algorithm is integrated with the NN-based estimators to find the optimal hyper-parameters for the studied architectures. The tuned hyper-parameters in this work include the number of neurons, number of layers, activation function, optimizer type, and learning rate. The objective function of the optimization problem is minimizing the root-mean-square error (RMSE) on multiple testing data. The optimal models are further included in the design of a hybrid NN estimator with Kalman filter. In the hybrid estimators, the optimal NN estimators are used as virtual sensors to correct the prediction of the side-slip angle resulting from the mathematical lateral vehicle model. Eventually, the performance of the best selected model is evaluated in terms of different metrics; mean RMSE, mean error variance, mean training time, and mean estimation time. LSTMs are found to achieve the lowest mean RMSE while being tested on highly generalized data yielding the highest training and estimation time. However, FFNNs achieve the lowest RMSE while being tested on low generalized data and the lowest training and estimation time. Meanwhile, it is observed that the hybrid estimators achieved lower RMSE with great enhancement compared to the non-hybridized ones proving the effectiveness of the proposed approach and increasing the side-slip estimation generalization ability in unknown environments with high uncertainties, which are not covered by the training dataset for the NNs estimators. [ABSTRACT FROM AUTHOR]

This work presents an experimental analysis related to 3D-printed carbon-fiber-reinforced-polymer (CFRP) machining. A polyethylene-terephthalate-glycol (PETG)-based composite, reinforced with 20% carbon fibers, was selected as the test material. The aim of the study was to evaluate the influence of cutting conditions used in light operations on the generated surface quality of the 3D-printed specimens. For this purpose, nine specimens were fabricated and machined under a wide range of cutting parameters, including cutting speed, feed, and depth of cut. The generated surface roughness was measured with a mechanical gauge and the acquired data were used to develop a shallow artificial neural network (ANN) for prediction purposes, showing that a 3-6-1 structure is the best solution. Following this, a genetic algorithm (GA) was utilized to minimize the response, revealing that the optimal combination is 205 m/min speed, 0.0578 mm/rev feed, and 0.523 mm depth of cut, contributing to the fabrication of low friction parts and shafts with a high quality surface, as well as to the reduction of resource waste. A validation study supported the accuracy of the developed model, by exhibiting errors below 10%. Finally, a set of enhanced images were taken to assess the machined surfaces. It was found that 1.50 mm depth of cut is responsible for the generation of defects across the circumference of the specimens. Especially, combined with 150 m/min cutting speed and 0.11 mm/rev feed, more flaws are produced. [ABSTRACT FROM AUTHOR]

In this study, four Artificial intelligence (AI) - based machine learning models were developed to estimate the Residual compressive strength (RCS) value of concrete supported with nano additives of Nanocarbon tubes (NCTs) and Nano alumina (NAl), after exposure to elevated temperatures ranging from 200 to 800 degrees. These models were developed via adapting meta- heuristic models including the Water cycle algorithm (WCA), Genetic algorithm (GA), and classical AI models of Artificial neural networks (ANNs), Fuzzy logic models (FLM), in addition to the statistical method of Multiple linear regression (MLR). 156 post heating experimental results available as a literature data (represents four input parameters of temperature change, heat exposure duration, nanomaterial type, and replacement proportion) are used to achieve the study's objective. Results of the developed models demonstrated that ANN and FLM have strong potential in predicting RCS. However, it is often infeasible to generate practical equations that relate input and output variables from these models. Upon analysing the results of the WCA and GA, it was found that WCA yielded the most accurate predictions based on all performance indicators. Furthermore, RCS prediction equations with superior accuracy were derived utilizing the meta-heuristic AI models of WCA and GA, with Mean absolute errors (MAEs) of 3.09 kg/cm² and 3.53 kg/cm² for the training, 1.91 kg/cm² and 2.72 kg/cm² for the validation, and 1.91 kg/cm² and 2.72 kg/cm² for the testing data sets, respectively. Additionally, sensitivity analysis via neural networks weights and SHAP investigation were performed to reveals the impact and relationship of the input variables with the output variables. Both techniques reveal that temperature degree and time of exposure had the highest positive impact on RCS value, followed by NAl and NCTs, in order. [ABSTRACT FROM AUTHOR]

Dynamic vibration absorbers (DVAs) have been widely employed in vibration suppression applications for decades. While DVAs offer an effective solution, they are limited by the need for a high mass ratio between the DVA and the primary system to achieve significant vibration attenuation. To overcome this, researchers have introduced lever mechanisms, allowing for enhanced vibration suppression without increasing the mass ratio. However, levers, commonly used as amplification mechanisms, suffer from high inertia and limited amplification, particularly in larger applications. Another limitation is when DVAs are employed for energy harvesting as a secondary objective, they exhibit high sensitivity to system parameter variations, requiring extensive optimization. Various optimization techniques have been applied to DVAs for multiobjective optimization, including fixed-point theory, which is complex and requires intensive mathematical derivation, and simple metaheuristic techniques such as genetic algorithms (GA). This study proposes four novel DVAs using a hydraulic amplifier (HA) to address the limitations of traditional lever mechanisms and a mechanical inerter to improve the vibration damping. Also, multi-objective optimization was performed using particle swarm optimization (PSO) which is considered innovative in this application and compared with commonly used genetic algorithms (GA). The governing equations were derived using Newton's second law and solved numerically with the Runge-Kutta method. An AI-based approach was utilized for HA design. The results show that integrating HA and mechanical inerters significantly enhances vibration attenuation and broadens the frequency response. Additionally, the location of the mechanical inerter is critical in reducing vibration amplitude. Also, the multi-objective PSO outperforms GA in solution diversity and quality. The proposed integration of HA in DVAs offers potential applications across various engineering fields. [ABSTRACT FROM AUTHOR]

In this paper, the Artificial Neural Network (ANN) was utilized to predict the peak discharge of dam failures, which was based on the combined Genetic Algorithm (GA) and Back Propagation (BP) neural network. The dataset comprises 40 samples from self-conducted experiments and available literature. To compare the efficiency of the suggested approach, three evaluation metrics, including the coefficient of determination (R2), the root mean square error (RMSE) and the mean absolute error (MAE), were analyzed for both the BP neural network and the GA-BP neural network. The findings suggest that (1) The prediction accuracy of the GA-BP was better than that of the BP; and (2) Compared to BP, GA-BP demonstrated a 9.07% average improvement in R2, a 57.36% average reduction in MAE, and a 57.53% average reduction in RMSE. In addition, the results of GA-BP and semi-empirical formulas were compared and the effect of three parameters on the peak discharge was analyzed. The results showed that the GA-BP model could effectively predict the peak discharge of dam failures. [ABSTRACT FROM AUTHOR]

The peak shear strength of a rock joint is an important indicator in rock engineering, such as mining and sloping. Therefore, direct shear tests were conducted using an RDS-200 rock direct shear apparatus, and the related data such as normal stress, roughness, size, normal loading rate, basic friction angle, and JCS were collected. A peak shear strength prediction model for rock joints was established, by which a predicted rock joint peak shear strength can be obtained by inputting the influencing factors. Firstly, the study used the correlation analysis method to find out the correlation coefficient between the above factors and rock joint peak shear strength to provide a reference for factor selection of the peak shear strength prediction model. Then, the JRC-JCS model and four established GA-BP neural network models were studied to identify the most valuable rock joint peak shear strength prediction method. The GA-BP neural network models used a genetic algorithm to optimize the BP neural network with different input factors to predict rock joint peak shear strength, after dividing the selected data into 80% training set and 20% test set. The results show that the error of the JRC-JCS model is a little bigger, with a value of 11.2%, while the errors of the established GA-BP neural network models are smaller than 6%, which indicates that the four established GA-BP neural network models can well fit the relationship between the peak shear strength and selected input factors. Additionally, increasing the factor number of the input layer can effectively improve the prediction accuracy of the GA-BP neural network models, and the prediction accuracy of the GA-BP neural network models will be higher if factors that have higher correlation with the output results are used as input factors. [ABSTRACT FROM AUTHOR]

The fault-induced delayed voltage recovery (FIDVR) and short-term voltage instability are increasing, especially due to the widespread implementation of residential air conditioners (RACs) in modern power systems. Single-phase induction motors in RACs have a high potential to stall in less than two to three cycles following a voltage dip in transmission or distribution systems. Using Shunt-FACTS devices, such as SVC and STATCOM, is a suitable solution for mitigating FIDVR events. In this paper, the Bayesian regularized artificial neural networks technique is employed to solve multidimensional mapping problems, taking into account the reactive powers injected into Busses. Following this, a multi-objective dynamic VAR programming is proposed to identify the optimal size of STATCOM for short-term voltage instability using trajectory sensitivities and heuristic optimization. This method is subject to complying with the criteria for dynamic and transient performance during FIDVR events. Dynamic VAR planning is carried out with assistance of the non-dominated sorting genetic algorithm II (NSGA-ӀӀ). The proposed multi-objective approach has been tested on the IEEE 39-bus system, taking into account time-varying practical load models. The results illustrate the effectiveness of the proposed approach in solving reactive power optimization problems while moderating the consequences of FIDVR. [ABSTRACT FROM AUTHOR]

The current work proposes the incorporation of an artificial neural network to solve the Gross–Pitaevskii equation (GPE) efficiently, using a few realistic external potentials. With the assistance of neural networks, a model is formed that is capable of solving this equation. The adaptation of the parameters for the constructed model is performed using some evolutionary techniques, such as genetic algorithms and particle swarm optimization. The proposed model is used to solve the GPE for the linear case ( γ = 0 ) and the nonlinear case ( γ ≠ 0 ), where γ is the nonlinearity parameter in GPE. The results are close to the reported results regarding the behavior and the amplitudes of the wavefunctions. [ABSTRACT FROM AUTHOR]

The presented work outlines a new genetic crossover operator, which can be used to solve problems by the Grammatical Evolution technique. This new operator intensively applies the one-point crossover procedure to randomly selected chromosomes with the aim of drastically reducing their fitness value. The new operator is applied to chromosomes selected randomly from the genetic population. This new operator was applied to two techniques from the recent literature that exploit Grammatical Evolution: artificial neural network construction and rule construction. In both case studies, an extensive set of classification problems and data-fitting problems were incorporated to estimate the effectiveness of the proposed genetic operator. The proposed operator significantly reduced both the classification error on the classification datasets and the feature learning error on the fitting datasets compared to other machine learning techniques and also to the original models before applying the new operator. [ABSTRACT FROM AUTHOR]

Machine learning-based Surrogate Reservoir Models (SRMs) can replace/augment multi-physics numerical simulations by replicating the reservoir simulation results with reduced computational effort while maintaining accuracy compared with numerical simulations. This research will demonstrate SRMs' potential in long-term simulations and optimization of geological carbon storage in a real-world geological setting and address challenges in big data curation and model training. The present study focuses on CO2 storage in the Smeaheia saline aquifer. Two SRMs were created using Deep Neural Networks (DNNs) to predict CO2 saturation and pressure over all grid blocks for 50 years. 18 million samples and 31 features, including reservoir static and dynamic properties, build the input data. Models comprise 3–5 hidden layers with 128–512 units apiece. SRMs showed a runtime improvement of 300 times and an accuracy of 99% compared to the 3D numerical simulator. The genetic algorithm was then employed to determine the optimal rate and duration of CO2 injection, which maximizes the volume of injected CO2 while ensuring storage operations' safety through constraints. The optimization continued for the reproduction of 100 generations, each containing 100 individuals, without any hyperparameter tuning. Finally, the optimization results confirm the significant potential of Smeaheia for storing 170 Mt CO2. [ABSTRACT FROM AUTHOR]

Although the double-double (DD) laminates proposed by Tsai provide a promising option for achieving better structural performance with lower manufacturing and maintenance costs, the buckling performance of perforated DD laminates still remains clear. In this study, optimal ply angles, rotation angles, and the corresponding maximum buckling loads are determined for DD laminates with various cutouts, which are used for comparisons to evaluate the effects of cutout size and shape on the buckling behaviour of perforated DD laminates. Apart from conventional circular and elliptical cutouts, the use of a combined-shape cutout for DD laminates is also investigated. As a large number of optimisations are required to obtain the maximum buckling loads for different cases in this study, an efficient optimisation method for perforated DD laminates is proposed based on an artificial neural network (ANN) and a genetic algorithm (GA). Unlike conventional quadaxial (QUAD) laminates, the repetition of a four-ply sublaminate in DD laminates makes their layup to be represented by only two ply angles; hence, the application of ANN models for predicting the buckling behaviour of various perforated DD laminates is studied in this paper. The superior performance of the ANN models is demonstrated by comparisons with other machine learning models. Instead of using the time-consuming FEA, the developed ANN model is utilised within a GA to obtain the maximum buckling load of perforated DD laminates. Compared to the circular cutout, the use of elliptical and combined-shape cutouts leads to more noticeable changes in the optimal ply angles as the cutout size increases. Based on the obtained results, the use of the combined-shape cutout is recommended for DD laminates. [ABSTRACT FROM AUTHOR]

Hardware‐based sensing frameworks such as cooperative fuel research engines are conventionally used to monitor research octane number (RON) in the petroleum refining industry. Machine learning techniques are employed to predict the RON of integrated naphtha reforming and isomerisation processes. A dynamic Aspen HYSYS model was used to generate data by introducing artificial uncertainties in the range of ±5% in process conditions, such as temperature, flow rates, etc. The generated data was used to train support vector machines (SVM), Gaussian process regression (GPR), artificial neural networks (ANN), regression trees (RT), and ensemble trees (ET). Hyperparameter tuning was performed to enhance the prediction capabilities of GPR, ANN, SVM, ET and RT models. Performance analysis of the models indicates that GPR, ANN, and SVM with R2 values of 0.99, 0.978, and 0.979 and RMSE values of 0.108, 0.262, and 0.258, respectively performed better than the remaining models and had the prediction capability to capture the RON dependence on predictor variables. ET and RT had an R2 value of 0.94 and 0.89, respectively. The GPR model was used as a surrogate model for fitness function evaluations in two optimisation frameworks based on genetic algorithm and particle swarm method. Optimal parameter values found by the optimisation methodology increased the RON value by 3.52%. The proposed methodology of surrogate‐based optimisation will provide a platform for plant‐level implementation to realise the concept of industry 4.0 in the refinery. [ABSTRACT FROM AUTHOR]

Spherical particles stand out as high‐value products with superior macroscopic properties and enhanced downstream processing efficiency. In this study, an integrated digital design strategy, combining artificial neural networks (ANN) and genetic algorithms (GA) has been employed to optimize the spherical agglomeration (SA) process. Initially, a dataset of benzoic acid SA processes was created, which was subsequently employed for training and testing the ANN model. An environmental impact sustainability index (STI) was constructed to assess the environmental effects associated with each operational variable in the SA process. To attain multi‐objective optimization, a GA was employed in combination with the ANN model. In addition, a Score function was formulated to generate Pareto fronts, tailored to meet the specific needs of real scenarios, considering variations in the assigned weights. Furthermore, the model was adapted for aspirin SA process, enhancing predictive abilities with only 20% of original data on operating conditions. [ABSTRACT FROM AUTHOR]

The duration of a residential building project will directly influence its successful implementation; hence, it is essential to estimate a reasonable timeframe. In this study, a genetic algorithm (GA) was employed to optimize and refine the weights and thresholds of a back propagation (BP) neural network, thereby creating a GA‐BP neural network model. A dataset comprising 111 instances of residential building durations was gathered, segmented into 90 training sets and 21 test sets. The model was validated and assessed through root mean square error (RMSE), correlation coefficient (R), and average error rate, demonstrating that the GA‐BP neural network model is effective in predicting the duration of residential buildings. To enhance the predictive accuracy of the GA‐BP neural network model, this research utilized an artificial bee colony (ABC)‐improved K‐means clustering algorithm to categorize 111 experimental datasets and 33 new datasets. The results indicated that the ABC‐K‐means‐GA‐BP model exhibited robust generalization capabilities and high predictive accuracy, with the fitness function showing optimal performance after 10, 15, and 35 generations, and the best validation performances recorded as 0.0019156, 0.00035905, and 0.0036914. This validates that the proposed ABC‐K‐means‐GA‐BP neural network model significantly aids in forecasting the construction period of residential buildings, which holds substantial practical value for enhancing construction efficiency. [ABSTRACT FROM AUTHOR]

The water supply pipeline is regarded as the "lifeline" of the city. In recent years, pipeline accidents caused by aging and other factors are common and have caused large economic losses. Therefore, in order to avoid large economic losses, it is necessary to analyze the failure prediction of pipelines so that the pipelines that are going to fail can be replaced in a timely manner. In this paper, we propose a method for predicting the failure pressure of pipelines, i.e., a genetic algorithm was used to optimize the weights and thresholds of a BP neural network. The first step was to determine the topology of the neural network and the number of input and output variables. The second step was to optimize the weights and thresholds initially set for the back propagation neural network using a genetic algorithm. Finally, the optimized back-propagation neural network was used to simulate and predict pipeline failures. It was proved by examples that compared with the separate back propagation neural network model and the optimized and trained genetic algorithm-back propagation neural network, the model performed better in simulation prediction, and the prediction accuracy could reach up to 91%, whereas the unoptimized back propagation neural network model could only reach 85%. It is feasible to apply this model for fault prediction of pipelines. [ABSTRACT FROM AUTHOR]

The performance of a deep neural network is affected by its configuration as well as its training process. Determining the configuration of a DNN and training its parameters are challenging tasks due to high-dimensional problems. Therefore, there is a need for methods that can optimize the configuration and parameters of a DNN. Most of the existing DNN optimization research concerns the optimization of DNN parameters, and there are only a few studies discussing the optimization of DNN configuration. In this paper, enhanced harmony search is proposed to optimize the configuration of a fully connected neural network. The proposed harmony search enhancement is conducted by introducing various types of harmony memory consideration rate and various types of harmony memory selection. Four types of harmony memory consideration rate are proposed in this research: constant rate, linear increase rate, linear decrease rate, and sigmoid rate. Two types of harmony memory selection are proposed in this research: rank-based selection and random selection. The combination of types of harmony memory consideration rate and types of selection generates eight harmony search scenarios. The performance of the proposed method is compared to random search and genetic algorithm using 12 datasets of classification problems. The experiment results show that the proposed harmony search outperforms random search in 8 out of 12 problems and approximately has the same performance in 4 problems. Harmony search also outperforms genetic algorithm in five problems, approximately has the same performance in six problems, and has worse performance in one problem. In addition, combining various types of harmony memory consideration rate and rank-based selection increases the performance of the ordinary harmony search. The combination of harmony memory consideration with linear increase rate and rank-based selection performs the best among all combinations. It is better than the ordinary harmony search in seven problems, approximately equal in three problems, and worse in two problems. The results show that the proposed method has some advantages in solving classification problems using a DNN. First, the configuration of the DNN is represented as an optimization problem so that it can be used to find a specific FCNN configuration that is suitable for a specific problem. Second, the approach is a global optimization approach as it tunes the DNN hyperparameter (configuration) as well as the DNN parameter (connection weight). Therefore, it is able to find the best combination of DNN configuration as well as its connection weight. However, there is a need to develop a strategy to balance the hyperparameter tuning and the parameter tuning. Inappropriate balance could lead to a high computational cost. Future research can be directed to balance the hyperparameter and parameter tuning during the solution search. [ABSTRACT FROM AUTHOR]

Deep transfer learning is frequently employed to address the challenges arising from limited or hard-to-obtain training data in the target domain, but its application in axial compressors has been scarcely explored thus far. In this paper, a multi-objective optimization framework of a transonic rotor is established using deep transfer learning. This framework first pre-trains deep neural networks based on the peak efficiency condition of 100% design speed and then fine-tunes the networks to predict the performance of off-design conditions based on the small training dataset. Finally, the design optimization of the transonic rotor is carried out through non-dominated sorting genetic algorithm II. Compared to neural networks that are trained directly, transfer learning models can achieve higher prediction accuracy, particularly in scenarios with small training datasets. This is because the pre-trained weights can offer a better initial state for transfer learning models. Moreover, transfer learning models can use fewer samples to obtain an approximate Pareto front, making the optimized rotor increase the isentropic efficiency at both peak efficiency and high loading conditions. The efficiency improvement of the optimized rotor is attributed to the reduction of the loss associated with the tip leakage flow by adjusting the tip loading distribution. Overall, this study fully demonstrates the effectiveness of transfer learning in predicting compressor performance, which provides a promising approach to solving high-cost compressor design problems. [ABSTRACT FROM AUTHOR]

Many dynamic random-access memory (DRAM) manufacturing companies encounter significant resistance value deviations during the PVD sputtering process for manufacturing Ti thin films. These resistance values are influenced by the thickness of the thin films. Current mitigation strategies focus on adjusting film thickness to reduce resistance deviations, but this approach affects product structure profile and performance. Additionally, varying Ti thin film thicknesses across different product structures increase manufacturing complexity. This study aims to minimize resistance value deviations across multiple film thicknesses with minimal resource utilization. To achieve this goal, we propose the TSDTM-ANN-GA framework, which integrates the two-stage dynamic Taguchi method (TSDTM), artificial neural networks (ANN), and genetic algorithms (GA). The proposed framework requires significantly fewer experimental resources than traditional full factorial design and grid search method, making it suitable for resource-constrained and low-power computing environments. Our TSDTM-ANN-GA framework successfully identified an optimal production condition configuration for five different Ti thin film thicknesses: Degas temperature = 245 °C, Ar flow = 55 sccm, DC power = 5911 W, and DC power ramp rate = 4009 W/s. The results indicate that the deviation between the resistance values and their design values for the five Ti thin film thicknesses decreased by 86.8%, 94.1%, 95.9%, 98.2%, and 98.8%, respectively. The proposed method effectively reduced resistance deviations for the five Ti thin film thicknesses and simplified manufacturing management, allowing the required design values to be achieved under the same manufacturing conditions. This framework can efficiently operate on resource-limited and low-power computers, achieving the goal of real-time dynamic production parameter adjustments and enabling DRAM manufacturing companies to improve product quality promptly. [ABSTRACT FROM AUTHOR]

In this research, the optimization of the geometric parameters of a 3-layer composite plate with an elliptical cutout in the center was investigated utilizing an artificial neural network and genetic algorithm for the first time to obtain the minimum ratio of linear buckling load to weight and the ratio of the diameters of the ellipse. After completing 294 simulations in Abaqus finite element software and forming an artificial neural network with two hidden layers, the function obtained in the artificial neural network was used as the objective function in the genetic algorithm, and the optimal values for the angle of the ellipse with the horizon axis, the diameter ratio of ellipses as well as the fiber angle of the middle layer of the composite were obtained. using artificial neural networks, three algorithms, Levenberg-Marquardt, Bayesian regularization; and scaled conjugate gradient backpropagation, were compared and the number of neurons in each algorithm was compared. The results showed that Levenberg-Marquardt's algorithm is more accurate compared to other algorithms. In the end, the optimal values for the angle of the ellipse with the horizontal axis; as the fiber angle of the middle layer of the composite and, the ratio of the oval diameters as α = 12.3001 degree , θ = 5.9481 degree and D = 1.5690. [ABSTRACT FROM AUTHOR]

Purpose: The purpose of this study is to compare the prediction models for half-cell potential (HCP) of RCC slabs cathodically protected using pure magnesium anodes and subjected to chloride ingress.The models for HCP using 1,134 data set values based on experimentation are developed and compared using ANFIS, artificial neural network (ANN) and integrated ANN-GA algorithms. Design/methodology/approach: In this study, RCC slabs, 1000 mm × 1000 mm × 100 mm were cast. Five slabs were cast with 3.5% NaCl by weight of cement, and five more were cast without NaCl. The distance of the point under consideration from the anode in the x- and y-axes, temperature, relative humidity and age of the slab in days were the input parameters, while the HCP values with reference to the Standard Calomel Electrode were the output. Experimental values consisting of 80 HCP values per slab per day were collected for 270 days and were averaged for both cases to generate the prediction model. Findings: In this study, the premise and consequent parameters are trained, validated and tested using ANFIS, ANN and by using ANN as fitness function of GA. The MAPE, RMSE and MAE of the ANFIS model were 24.57, 1702.601 and 871.762, respectively. Amongst the ANN algorithms, Levenberg−Marquardt (LM) algorithm outperforms the other methods, with an overall R-value of 0.983. GA with ANN as the objective function proves to be the best means for the development of prediction model. Originality/value: Based on the original experimental values, the performance of ANFIS, ANN and GA with ANN as objective function provides excellent results. [ABSTRACT FROM AUTHOR]

In this study, an Artificial Neural Network-Genetic Algorithm(ANN-GA) approachwas successfully applied to optimise the physicochemical factors influencing the synthesis of unsaturated fatty acids (UFAs) in the microalgae P. kessleri UCM 001. The optimized model recommended specific cultivation conditions, including glucose at 29 g/L, NaNO3 at 2.4 g/L, K2HPO4 at 0.4 g/L, red LED light, an intensity of 1000 lx, and an 8:16-h light-dark cycle. Through ANN-GA optimisation, a remarkable 66.79% increase in UFAs production in P. kessleri UCM 001 was achieved, compared to previous studies. This underscores the potential of this technology for enhancing valuable lipid production. Sequential variations in the application of physicochemical factors during microalgae culture under mixotrophic conditions, as optimized by ANN-GA, induced alterations in UFAs production and composition in P. kessleri UCM 001. This suggests the feasibility of tailoring the lipid profile of microalgae to obtain specific lipids for diverse industrial applications. Themicroalgaewere isolated froma high-mountain lake in Colombia, highlighting their adaptation to extreme conditions. This underscores their potential for sustainable lipid and biomaterial production. This study demonstrates the effectiveness of using ANN-GA technology to optimise UFAs production in microalgae, offering a promising avenue for obtaining valuable lipids. Themicroalgae's unique origin in a high-mountain environment in Colombia emphasises the importance of exploring and harnessing microbial resources in distinctive geographical regions for biotechnological applications. [ABSTRACT FROM AUTHOR]

Reversible plows are one of the important and efficient tools in primary tillage, which are affected by many dynamic loads. These tools are damaged in different working conditions, and they are damaged from the stem area. For this reason, in this research, a method based on the finite element method and genetic algorithm was presented to optimize the reversible plow shaft. In this research, two parameters of cross-sectional area and stem curvature were investigated as independent variables. A total of 24 different models for the plow shaft were designed in SolidWorks software, FEM software and used Iron Cow Stem. Then, the different designs of the stem in the environment of the abacus were loaded and stress free occurred in them and were eliminated. Then, using an artificial neural network, a model was presented to estimate the von Mises tension based on the information related to the cross-sectional area and stem curvature, and this model was able to estimate the maximum von Mises tension with an accuracy of 99%. Then the mentioned model was linked with the genetic algorithm and it was used to optimize the plow shaft. After selecting the optimized model through the genetic algorithm, the plow shaft was designed again and the tireless stress that occurred in it under the same loading conditions as the previous conditions was eliminated. The results showed that the amount of maximum stress in this model decreased by 6% compared to the previous models and the best stem designs is (Stress (MPa) VonMises=295.2). [ABSTRACT FROM AUTHOR]

Hellenic Open University has been developing Onlabs, a virtual biology laboratory simulating its on-site laboratory, for its students to be trained before the on-site learning activities. The evaluation of user performance in Onlabs is based on a scoring algorithm, which admits some optimization by means of Genetic Algorithms and Artificial Neural Networks. Moreover, for a particular experimental procedure (microscoping), we have experimented with incorporating into it some background knowledge about the procedure, which allows one to break it down in a series of conceptually linked steps in a hierarchical fashion. In this work, we review the flat and hierarchical modes used for the calibration of the automated assessment mechanism and offer an experimental comparison of both approaches with the aim of devising automated scoring schemes which are fit for training in an at-a-distance learning context. Overall, the genetic algorithm fails to deliver good convergence results in the non-hierarchical setting but performs better in the hierarchical one. On the other hand, the neural network most of the time converges, with the non-hierarchical network achieving a slightly better convergence than the hierarchical one, with the latter, however, delivering a smoother and more realistic assessment mechanism. [ABSTRACT FROM AUTHOR]

The Stochastic Reduced-Order Models (SROMs) are a non-embedded uncertainty analysis method that has the advantages of high computational efficiency, easy implementation, and no dimensional disasters. Recently, it has been widely used in the field of EMC simulation. In the process of optimizing electromagnetic protection design, the worst-case estimation value is an extremely important uncertainty quantification simulation result. However, the SROMs have a large error in providing this result, which limits its application in the field of EMC simulation prediction. An improved SROM based on the Radial Basis Function (RBF) neural network algorithm is proposed in this paper, which improves the fitness function in the genetic algorithm center clustering process and constructs an RBF neural network model to obtain accurate worst-case estimation results. The accuracy improvement effect of the algorithm proposed in this paper in worst-case estimation is quantitatively verified by using a parallel cable crosstalk prediction example from published literature. [ABSTRACT FROM AUTHOR]

Recent advancements in wind farm (WF) construction have prompted designers to focus on optimizing WF performance and minimizing the negative effect of wake interference on power output. One proven technique to reduce wake influence and improve power generation is wake steering, which involves adjusting the yaw angles of wind turbines (WTs). The objective of this study is to evaluate the optimization of WF using machine learning (ML), specifically an artificial neural network (ANN) approach. In the first phase of the study, an ANN power model is selected to standardize wake loss and optimize power. In the next phase, a genetic algorithm (GA) will be employed to determine the best yaw angles for the wind direction based on data collected from the ANN power model. The ANN power model might not account for maintenance tasks around the WF. The results of the optimization demonstrate that energy is maximized between wind directions of 166° to 178°. This research achieved a.97 optimized overall power ratio, compared to non-optimized scenarios, using an optimal yaw angle technique in all directions. The findings of the study demonstrate that ANN-based optimization, combined with standardized wake degradation and suitable yaw angle direction, is an efficient method for WF optimization. [ABSTRACT FROM AUTHOR]

In this paper, an innovative hybrid technique is proposed for the efficient training of artificial neural networks, which are used both in class learning problems and in data fitting problems. This hybrid technique combines the well-tested technique of Genetic Algorithms with an innovative variant of Simulated Annealing, in order to achieve high learning rates for the neural networks. This variant was applied periodically to randomly selected chromosomes from the population of the Genetic Algorithm in order to reduce the training error associated with these chromosomes. The proposed method was tested on a wide series of classification and data fitting problems from the relevant literature and the results were compared against other methods. The comparison with other neural network training techniques as well as the statistical comparison revealed that the proposed method is significantly superior, as it managed to significantly reduce the neural network training error in the majority of the used datasets. [ABSTRACT FROM AUTHOR]

Accurate prediction of NOx emissions under actual working conditions is crucial for managing regional pollutant emissions. Therefore, this study focuses on agricultural tractors and employs a Portable Emission Measurement System (PEMS) to gather NOx emission under real operating conditions. By conducting correlation analyses of factors affecting NOx emissions, the main factors affecting NOx emissions during the actual working conditions of tractors were determined, and the NOx emission prediction model was established by applying these factors. In the process of establishing a NOx emission prediction model for agricultural tractors, the research utilizes the Back Propagation (BP) neural network, and the Long Short-Term Memory (LSTM) neural network, and optimizes both the BP and LSTM neural networks using Genetic Algorithm (GA) for comparison and evaluation of their prediction performance. The results demonstrate that among the established models, the optimized GA-BP neural network model excels in predicting NOx emissions. This model outperforms other neural network models in various evaluation metrics, including Root Mean Square Error(RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and the coefficient of determination (R²), with values of 5.679 x 10-3, 4.057 x 10-3, 3.751% and 0.991 5, respectively. Therefore, it is feasible to use the GA-BP neural network model to predict NOx emissions from agricultural tractors. [ABSTRACT FROM AUTHOR]

Abstract In this study, four Artificial intelligence (AI) - based machine learning models were developed to estimate the Residual compressive strength (RCS) value of concrete supported with nano additives of Nanocarbon tubes (NCTs) and Nano alumina (NAl), after exposure to elevated temperatures ranging from 200 to 800 degrees. These models were developed via adapting meta- heuristic models including the Water cycle algorithm (WCA), Genetic algorithm (GA), and classical AI models of Artificial neural networks (ANNs), Fuzzy logic models (FLM), in addition to the statistical method of Multiple linear regression (MLR). 156 post heating experimental results available as a literature data (represents four input parameters of temperature change, heat exposure duration, nanomaterial type, and replacement proportion) are used to achieve the study’s objective. Results of the developed models demonstrated that ANN and FLM have strong potential in predicting RCS. However, it is often infeasible to generate practical equations that relate input and output variables from these models. Upon analysing the results of the WCA and GA, it was found that WCA yielded the most accurate predictions based on all performance indicators. Furthermore, RCS prediction equations with superior accuracy were derived utilizing the meta-heuristic AI models of WCA and GA, with Mean absolute errors (MAEs) of 3.09 kg/cm² and 3.53 kg/cm² for the training, 1.91 kg/cm² and 2.72 kg/cm² for the validation, and 1.91 kg/cm² and 2.72 kg/cm² for the testing data sets, respectively. Additionally, sensitivity analysis via neural networks weights and SHAP investigation were performed to reveals the impact and relationship of the input variables with the output variables. Both techniques reveal that temperature degree and time of exposure had the highest positive impact on RCS value, followed by NAl and NCTs, in order.

In this paper, an innovative hybrid technique is proposed for the efficient training of artificial neural networks, which are used both in class learning problems and in data fitting problems. This hybrid technique combines the well-tested technique of Genetic Algorithms with an innovative variant of Simulated Annealing, in order to achieve high learning rates for the neural networks. This variant was applied periodically to randomly selected chromosomes from the population of the Genetic Algorithm in order to reduce the training error associated with these chromosomes. The proposed method was tested on a wide series of classification and data fitting problems from the relevant literature and the results were compared against other methods. The comparison with other neural network training techniques as well as the statistical comparison revealed that the proposed method is significantly superior, as it managed to significantly reduce the neural network training error in the majority of the used datasets.

The purpose of this research is to develop a new method for automatically optimizing the seawall cross-section with composite slopes and a berm, considering both overtopping discharge and construction cost. Minimizing these competing multi-objectives is highly challenging due to the intricate geometry of seawalls. In this study, the surrogate model optimization algorithm DYCORS (Dynamic COordinate search using Response Surface models) is employed to search for the optimal seawall geometry, coupled with the ANN (Artificial Neural Network) model for determining the overtopping discharge. A total of 20 trials have been run to evaluate the performance of our methodology. Even the worst-performing Trial 7 among these 20 trials shows a satisfactory performance, with a reduction of 17.67% in overtopping discharge and a 12.1% decrease in cost compared to the original solution. Furthermore, compared to other optimization schemes using GAs (Genetic Algorithms) with the same decision vectors, constraints, and multi-objective functions, the methodology has been proven to be more effective and robust. Additionally, when facing different combinations of wave conditions and water levels, there was a 27.8% reduction in objective function value compared to the original solution. The optimal results indicate that this method can still be effectively applied for optimizing the seawall cross-section as it is a general method. [ABSTRACT FROM AUTHOR]

This paper aims to study an optimization method for the lightweight design of bus superstructure. According to the requirements of ECE R66, the bus rollover finite element model has been developed, and the bus rollover process has been simulated and validated by experimental tests. The maximum error between test results and simulation results was 6.8%, which indicated that the simulation of bus rollover had good accuracy. A multi-objective optimization method for rollover safety has been proposed by combining a back propagation (BP) neural network model with a non-dominated sorting genetic algorithm (NSGA-II). The neural network model considered a different joint scheme of closed loop, wall thickness of side frame longitudinal beam, section of side frame longitudinal beam, wall thickness of side frame beam and section of side frame beam as inputs. It took the minimum distance of the survival space from the side column and the mass of the superstructure as outputs. The results show that the coupling optimization of the BP neural network and NSGA-II can reduce the total mass of the bus by 7.7%, which verifies the feasibility of applying the intelligent algorithm to the lightweight design of the bus. [ABSTRACT FROM AUTHOR]

Recent studies have demonstrated that finely tuned deep neural networks (DNNs) are susceptible to adversarial attacks. Conventional physical attacks employ stickers as perturbations, achieving robust adversarial effects but compromising stealthiness. Recent innovations utilise light beams, such as lasers and projectors, for perturbation generation, allowing for stealthy physical attacks at the expense of robustness. In pursuit of implementing both stealthy and robust physical attacks, the authors present an adversarial catoptric light (AdvCL). This method leverages the natural phenomenon of catoptric light to generate perturbations that are both natural and stealthy. AdvCL first formalises the physical parameters of catoptric light and then optimises these parameters using a genetic algorithm to derive the most adversarial perturbation. Finally, the perturbations are deployed in the physical scene to execute stealthy and robust attacks. The proposed method is evaluated across three dimensions: effectiveness, stealthiness, and robustness. Quantitative results obtained in simulated environments demonstrate the efficacy of the proposed method, achieving an attack success rate of 83.5%, surpassing the baseline. The authors utilise common catoptric light as a perturbation to enhance the method's stealthiness, rendering physical samples more natural in appearance. Robustness is affirmed by successfully attacking advanced DNNs with a success rate exceeding 80% in all cases. Additionally, the authors discuss defence strategies against AdvCL and introduce some light‐based physical attacks. [ABSTRACT FROM AUTHOR]

Existing designs of edge computing models are mostly targeted to improve the performance of accuracy. Yet, besides accuracy, robustness and inference efficiency are also crucial attributes to the performance. To achieve satisfied performance in edge-cloud computing frameworks, each distributed model is required to be both robust to perturbations and feasible for information uploading in wireless environments with limited bandwidth. In other words, feature encoders should be more robust and have faster inference time while maintaining accuracy at a competitive level. Therefore, to design accurate, robust and efficient models for bandwidth limited edge computing, we propose a systematic approach to autonomously optimize parameters and architectures of arbitrary deep neural networks. This approach employs a genetic algorithm based bi-generative adversarial network, which is utilized to autonomously develop and select the number of filters (for convolutional layers) and the number of neurons (for fully connected layers) from a wide range of values. To demonstrate the performance, we test our approach on ImageNet and ModelNet databases, and compare it with the state-of-the-art 3D volumetric network and two exclusively GA-based methods. Our results show that the proposed method can significantly improve performance by simultaneously optimizing multiple neural network parameters, regardless of the depth of the network. [ABSTRACT FROM AUTHOR]

Deep neural networks (DNNs) have been applied in many pattern recognition or object detection applications. DNNs generally consist of millions or even billions of parameters. These demanding computational storage and requirements impede deployments of DNNs in resource-limited devices, such as mobile devices, micro-controllers. Simplification techniques such as pruning have commonly been used to slim DNN sizes. Pruning approaches generally quantify the importance of each component such as network weight. Weight values or weight gradients in training are commonly used as the importance metric. Small weights are pruned and large weights are kept. However, small weights are possible to be connected with significant weights which have impact to DNN outputs. DNN accuracy can be degraded significantly after the pruning process. This paper proposes a roulette wheel-like pruning algorithm, in order to simplify a trained DNN while keeping the DNN accuracy. The proposed algorithm generates a branch of pruned DNNs which are generated by a roulette wheel operator. Similar to the roulette wheel selection in genetic algorithms, small weights are more likely to be pruned but they can be kept; large weights are more likely to be kept but they can be pruned. The slimmest DNN with the best accuracy is selected from the branch. The performance of the proposed pruning algorithm is evaluated by two deterministic datasets and four non-deterministic datasets. Experimental results show that the proposed pruning algorithm generates simpler DNNs while DNN accuracy can be kept, compared to several existing pruning approaches. [ABSTRACT FROM AUTHOR]

This study introduces an integrated power quality (PQ) conditioner, referred to as UPQC, that is linked with photovoltaic (PV) and battery energy systems (BSS) in order to address and solve PQ issues. It is proposed to employ the Levenberg–Marquardt (LM) backpropagation (LMBP) trained artificial neural network control (ANNC) technique for generating reference signal for converters in UPQC. This approach eliminates the need for traditional abc to dq0 to αβ conversions. Additionally, the hybrid algorithm (FFHSA) in combination of harmony search algorithm (HSA), and firefly algorithm (FFA) is also implemented for the optimal selection of adaptive neuro-fuzzy interface system (ANFIS) parameters to maintain direct current link capacitor voltage (DLCV) constant. The prime goal of the developed hybrid ANNC-FFHSA is to stabilize the DLCV with low settling time during load and solar irradiation (G), Temperature (T) changes, minimization of distortions in the source current signal to diminish total harmonic distortion (THD) in turn boosting the power factor (PF), suppression of fluctuations like disturbances, swell, sag and unbalances in the supply voltage. The suggested method is validated by four test cases with several combinations of variable irradiation (G), temperature and loads. On the other hand, to reveal the superiority of the developed method, the comparison is carried out with the genetic algorithm (GA) and Ant colony algorithm (ACA) along with instantaneous power (p–q) and Synchronous reference frame (SRF) conventional methods. The proposed approach significantly diminishes the total harmonic distortion to values of 3.61%, 3.48%, 3.48%, and 4.51%, which are notably lower compared to the values reported in the existing literature and also improves the power factor to almost unity. The design and implementation of this method were carried out using MATLAB/Simulink software. [ABSTRACT FROM AUTHOR]

The objective of this study is to investigate the synergistic integration of machine learning and evolutionary algorithms for the discovery of equivalent morphologies exhibiting analogous behavior within the domain of composite materials. To pursue this objective, two comprehensive databases are meticulously constructed. The first database encompasses randomly positioned inclusions characterized by varying volume fractions and contrast levels. Conversely, the second database comprises microstructures of diverse shapes, such as elliptical, square, and triangular, while maintaining consistent volume fraction and contrast values across samples. Label assignment for both databases is conducted using a finite-element-method-based computational tool, ensuring a standardized approach. Machine learning techniques are then applied, employing distinct methodologies tailored to the complexity of each database. Specifically, an artificial neural network ANN model is deployed for the first database due to its intricate parameter configurations, while an eXtreme Gradient Boosting XGBoost model is employed for the second database. Subsequently, these developed models are seamlessly integrated with a genetic algorithm, which operates to identify equivalent morphologies with nuanced variations in geometry, volume fraction, and contrast. In summation, the findings of this investigation exhibit notable levels of adaptation within the discovered equivalent morphologies, underscoring the efficacy of the integrated machine learning and evolutionary algorithm framework in facilitating the optimization of composite material design for desired behavioral outcomes. [ABSTRACT FROM AUTHOR]

In the current era, the government consistently emphasizes the pursuit of high-quality development, as evidenced by the ongoing increase in the tertiary industry's GDP share. As a crucial component of the modern service sector, logistics plays a pivotal role in determining the operational efficiency and overall quality of the industrial economy. This study focuses on constructing a Chongqing logistics express demand prediction index system. It employs an improved BP neural network model to forecast the logistics express demand for Chongqing over the next five years. Given the limited express demand data sequence and the normalized characteristics of the data, the selected training method is the Bayesian regularization approach, with the LeCun Tanh function serving as the hidden layer activation function. Additionally, a genetic algorithm is designed to optimize the initial weights and thresholds of the BP neural network, thereby enhancing prediction accuracy and reducing the number of iterations. The experimental results of the improved GA-BP network are analyzed and compared, demonstrating that the improved BP neural network, utilizing GA optimization, can more reliably and accurately predict regional logistics express demand. According to the findings, the forecast indicates that the logistics express demand for Chongqing in 2026 will be 2,171,642,700 items. [ABSTRACT FROM AUTHOR]

Methaheuristics (MHs) are techniques widely used for solving complex optimization problems. In recent years, the interest in combining MH and machine learning (ML) has grown. This integration can occur mainly in two ways: ML-in-MH and MH-in-ML. In the present work, we combine the techniques in both ways—ML-in-MH-in-ML, providing an approach in which ML is considered to improve the performance of an evolutionary algorithm (EA), whose solutions encode parameters of an ML model—artificial neural network (ANN). Our approach called TS in EA in ANN employs a reinforcement learning neighborhood (RLN) mutation based on Thompson sampling (TS). TS is a parameterless reinforcement learning method, used here to boost the EA performance. In the experiments, every candidate ANN solves a regression problem known as protein structure prediction deviation. We consider two protein datasets, one with 16,382 and the other with 45,730 samples. The results show that TS in EA in ANN performs significantly better than a canonical genetic algorithm (GA in ANN) and the evolutionary algorithm without reinforcement learning (EA in ANN). Analyses of the parameter's frequency are also performed comparing the approaches. Finally, comparisons with the literature show that except for one particular case in the largest dataset, TS in EA in ANN outperforms other approaches considered the state of the art for the addressed datasets. [ABSTRACT FROM AUTHOR]