Neural Network Optimization Algorithm Research Articles

Prediction and optimization of heating and cooling loads in a residential building based on multi-layer perceptron neural network and different optimization algorithms

Since cooling and heating loads are regarded as significant parameters to examine the energy performance of buildings, the need to predict and analyze them for the residential buildings seems to be undeniable. Hence, the present paper wants to optimize the multi-layer perceptron neural network using several optimization methods to predict the heating and cooling of energy-efficient buildings. The dataset used in this study consists of eight independent factors: surface area, wall area, roof area, relative compactness, overall height, orientation, glazing area, and glazing area distribution. To prove the reliability and accuracy of the obtained results, test and training data are also considered. According to the statistical results, biogeography-based optimization has the highest value of R2 and the lowest values of RMSD, normalized RMSD, and MAE in both training data and test data for cooling and heating loads. Hence, the forecasting accuracy of the proposed MLP neural network optimized with the BBO optimization algorithm with the RMSD, R2, and MAE of 2.82, 0.920, 2.15 in the training phase of heating load and with the RMSD, R2, and MAE of 3.18, 0.887, 2.97 in the training phase of the cooling load is much better than those of the other models.

Adaptive modification of turbofan engine nonlinear model based on LSTM neural networks and hybrid optimization method

An accurate and reliable turbofan engine model which can describe its dynamic behavior within the full flight envelop and lifecycle plays a critical role in performance optimization, controller design and fault diagnosis. However, due to the performance differences caused by the tolerance of engine manufacturing and assembly, and performance degradation during continuously stringent environmental regulations, the model accuracy is severely reduced. In this paper, an adaptive modification method of turbofan engine nonlinear Component-Llevel Model (CLM) based on Long Short-Term Memory (LSTM) Neural Network (NN) and hybrid optimization algorithm is pro-posed. First, a dynamic compensator with a combined LSTM NN architecture is constructed to compensate for the initial error between the experimental data and CLM of a turbofan engine under health condition. Then, a sensitivity analysis approach based on the entropy coefficient and technique for order preference by similarity to an ideal solution integrated evaluation is developed to choose the unmeasurable health parameters to be adjusted. Finally, a parallel hybrid optimization algorithm is developed to complete the adaptive model modification when the performance degrades. The proposed method is verified on a military low-bypass twin-spool turbofan engine, and the experimental results show the effectiveness of the proposed method.

Grade estimation using a hybrid method of back-propagation artificial neural network and particle swarm optimization with integrated samples coordinate and local variability

Grade estimation is a critical issue in mineral resource evaluation, being extensively investigated by data mining techniques. In this paper, a hybrid method composed of back-propagation artificial neural network (BPANN) and particle swarm optimization (PSO) algorithms is proposed to solve the grade estimation problem. The PSO algorithm is implemented to optimize the BPANN parameters by reducing the effects of a local minimum problem, which is one of the critical drawbacks of BPANN. The proposed BPANN-PSO algorithm is validated for Al2O3 grade estimation in one of Iran's largest Bauxite deposits. The performance of BPANN-PSO algorithm for grade estimation is compared with BPANN and ordinary kriging. The experimental results indicate that the BPANN-PSO model is more appropriate for estimating Al2O3 grade with a reasonable error.

Breast Cancer Diagnosis by Convolutional Neural Network and Advanced Thermal Exchange Optimization Algorithm.

A common gynecological disease in the world is breast cancer that early diagnosis of this disease can be very effective in its treatment. The use of image processing methods and pattern recognition techniques in automatic breast detection from mammographic images decreases human errors and increments the rapidity of diagnosis. In this paper, mammographic images are analyzed using image processing techniques and a pipeline structure for the diagnosis of the cancerous masses. In the first stage, the quality of mammogram images and the contrast of abnormal areas in the image are improved by using image contrast improvement and a noise decline. A method based on color space is then used for image segmentation that is followed by mathematical morphology. Then, for feature image extraction, a combined gray-level cooccurrence matrix (GLCM) and discrete wavelet transform (DWT) method is used. At last, a new optimized version of convolutional neural network (CNN) and a new improved metaheuristic, called Advanced Thermal Exchange Optimizer, are used for the classification of the features. A comparison of the simulations of the proposed technique with three different techniques from the literature applied on the MIAS mammogram database is performed to show its superiority. Results show that the accuracy of diagnosing cancer cases for the proposed method and applied on the MIAS database is 93.79%, and sensitivity and specificity are obtained 96.89% and 67.7%, respectively.

A Neural Network Inverse Optimization Procedure for Constitutive Parameter Identification and Failure Mode Estimation of Laterally Loaded Unreinforced Masonry Walls

A new Neural Network Optimization (NNO) algorithm for constitutive material parameter identification based on inverse analysis of experimental tests of small-scale masonry prisms under compressive loads is presented. The Concrete Damaged Plasticity (CDP) constitutive model is used for the brick and mortar of the Unreinforced Masonry (URM) walls. By comparisons with experimental data taken from laboratory tests, it is demonstrated that the constitutive parameters calibrated by application of the proposed inverse optimization procedure on the small-scale (prism) experimental results are sufficiently accurate to allow for the prediction of the mechanical response of large-scale URM walls subject to compressive and lateral loads. This eliminates the need for large-scale URM wall experimental tests for the identification of their material properties, making the calibration process more economic. After verifying the accuracy of the calibrated constitutive parameters based on the above comparisons, a numerical parametric study is performed for the investigation of the effect of material behavior and geometrical aspect ratios on the failure mechanisms of large-scale URM walls.

Estimation of nuclear magnetic resonance (NMR) log permeability by integrating artificial neural network and imperialist competition optimization algorithm (ICA) in one of the oil fields of southwestern Iran

Estimation of nuclear magnetic resonance (NMR) log permeability by integrating artificial neural network and imperialist competition optimization algorithm (ICA) in one of the oil fields of southwestern Iran

Innovative Approach for Moment Capacity Estimation of Spirally Reinforced Concrete Columns Using Swarm Intelligence–Based Algorithms and Neural Network

The purpose of this paper is to present an innovative equation to predict the moment capacity of spirally reinforced concrete columns with high accuracy using a combination of neural network and metaheuristic optimization algorithms. To this end, a large experimental database has been gathered to train a neural network with seven independent parameters that deal with the dimensional properties of the columns, reinforcements, materials, and also the forces. Furthermore, the authors improved the process of training with consideration of two optimization techniques: particle swarm optimization (PSO) and Harris hawks optimization (HHO). Then, the best model was selected to a statistical methodology to extract an empirical equation to predict the target, which makes the proposed system of this article more applicable, especially for the practical usages. The results indicated that the neural network with the PSO algorithm had better results than the other model. Also, it has been found that the proposed formulation could predict the moment capacity of the considered element with high performance. The presented equation of this article has many applications in civil engineering, such as retrofitting and rehabilitation.

A PROPOSED RECOGNITION SYSTEM FOR ALZHEIMER’S DISEASE BASED ON DEEP LEARNING AND OPTIMIZATION ALGORITHMS

Alzheimer's disease (AD) is an advanced and incurable neurodegenerative disease that causes progressive impairment of memory and cognitive functions due to the deterioration of brain cells. Early diagnosis is substantial to avoid permanent memory loss and develop treatments that will be subtracted in the future. Deep learning (DL) is a vital technique for medical imaging systems for AD diagnostics. The problem is multi-class classification seeking high accuracy. DL models have shown strong performance accuracy for multi-class prediction. In this paper, a proposed DL architecture is created to classify magnetic resonance imaging (MRI) to predict different stages of AD-based pre-trained Convolutional Neural Network (CNN) models and optimization algorithms. The proposed model architecture attempts to find the optimal subset of features to improve classification accuracy and reduce classification time. The pre-trained DL models, ResNet-101 and DenseNet-201, are utilized to extract features based on the last layer, and the Rival Genetic algorithm (RGA) and Pbest-Guide Binary Particle Swarm Optimization (PBPSO) are applied to select the optimal features. Then, the DL features and selected features are passed separately through created classifier for classification. The results are compared and analyzed by accuracy, performance metrics, and execution time. Experimental results showed that the most efficient accuracies were obtained by PBPSO selected features which reached 87.3% and 94.8% accuracy with less time of 46.7 sec, 32.7 sec for features based ResNet-101 and DenseNet-201, receptively.

Optimization of Extraction Parameters of Ethanol Extracts of Propolis Samples Using Artificial Neural Network and Moth-Flame Optimization Algorithm

Abstract In this study, the optimum values of propolis ethanol extracts parameters were determined with the use of single and multi-objective optimization procedures. The euclidean distance approach was used in the multi-objective optimization process. Firstly, propolis was extracted using water with ethanol contents 40, 50, 60, 70 and 80% for 8, 10, 12, 16, 20 and 24 h. Then, total phenolic content (TPC) and ferric reducing antioxidant power (FRAP) activities of all extracts were determined. With the obtained data a prediction model was produced with the use of artificial neural networks (ANN), and optimization was performed using a moth-flame (MFO) algorithm. The best prediction models for the TPC and FRAP were observed in 2-5-1 and 2-5-1 network architecture with the mean absolute percentage error (MAPE) values, 5.126 and 2.451%, respectively. For maximum TPC, the extraction parameters were determined as ethanol content 57.5% and extraction time 13.56 h. To maximize FRAP, the optimized extraction parameters were ethanol content 72.03% and extraction time 18.04 h. The optimum extraction conditions for both maximum values of the studied assays were ethanol content 70.03% and extraction time 16.93 h. The study concluded that the integrated ANN and MFO algorithm system can be used in single and multi-objective optimization of extraction parameters. The established optimization model can save time, money, labor and energy.

Hydrogen Safety Prediction and Analysis of Hydrogen Refueling Station Leakage Accidents and Process Using Multi-Relevance Machine Learning

Hydrogen energy vehicles are being increasingly widely used. To ensure the safety of hydrogenation stations, research into the detection of hydrogen leaks is required. Offline analysis using data machine learning is achieved using Spark SQL and Spark MLlib technology. In this study, to determine the safety status of a hydrogen refueling station, we used multiple algorithm models to perform calculation and analysis: a multi-source data association prediction algorithm, a random gradient descent algorithm, a deep neural network optimization algorithm, and other algorithm models. We successfully analyzed the data, including the potential relationships, internal relationships, and operation laws between the data, to detect the safety statuses of hydrogen refueling stations.

Deep Learning-Based CT Images in Pulmonary Function Assessment of Patients Who Underwent Laparoscopic Surgery under Guidance of Electrical Impedance Tomography

This research aimed to study the application of CT images based on deep learning in pulmonary function assessment of patients who underwent laparoscopic surgery under the guidance of electrical impedance tomography (EIT). Sixty patients undergoing laparoscopic surgery were taken as the research subjects, who were randomly labelled as control group and experimental group. Based on deep learning, the empty convolution-combined fully convolutional neural network optimization algorithm (ECFCNN) was proposed, which was adopted to evaluate the pulmonary function of 60 patients and was compared with convolutional neural network (CNN) algorithm. The clarity of the edge contour of the image segmented by ECFCNN was better than that segmented by CNN. Average arterial pressure (MAP) and heart rate (HR) were recorded before induction (T1), 10 min before pneumoperitoneum (T2), 10 min after pneumoperitoneum (T3), 10 min before extubation (T4), and 10 min after extubation (T5), respectively. Oxygenation index (PaO2/FiO2), alveolary-arterial partial pressure of oxygen (A-ADO2), and respiratory index (RI) were recorded. The sharpness of the segmentation image edge contour of the algorithm model in this study was higher than that of the convolutional neural network. Compared with T1, T2-T4 MAP in 2 groups was decreased ( P < 0.05 ). Compared with T1, T2-T5 HR was significantly decreased ( P < 0.05 ). Compared with T2, T5 PaO2/FiO2 in control group was significantly decreased ( P < 0.05 ). Compared with the control group, T5 A-aDO2 was decreased ( P < 0.05 ). To sum up, EIT-guided lung protective ventilation can assess the pulmonary function of patients who underwent laparoscopic surgery, reduce the incidence of atelectasis, and improve postoperative lung oxygenation.

Applying hybrid genetic and artificial bee colony algorithms to simulate a bio-treatment of synthetic dye-polluted wastewater using a rhamnolipid biosurfactant

The present work aims at optimization and advanced simulation of removal efficiency of dye material from a synthetic wastewater using a locally generated rhamnolipid (RL) biosurfactant. For this purpose, bio-treatment of dye polluted synthetic wastewater was experimentally, kinetically, and statistically investigated by the ion flotation process in the presence of the RL. The removal rate of methylene blue (MB) as the dye material was assessed by the ultraviolet (UV)-visible absorbance measurements. The impact of operating variables including RL concentration (as a dye collector, 5–50 ppm), methyl isobutyl carbinol (MIBC) dosage (as a frother, 10–70 ppm), solution pH (2–12) and aeration rate (1–5 l/min) were assessed through one-way analysis of variance (ANOVA) and Anderson-Darling as the normality analysis strategy. The process was simulated using two artificial neural network (ANN) optimization algorithms, i.e., genetic algorithm (GA) and artificial bee colony (ABC) as a novel approach. The statistical results indicated that the dye removal process was significantly influenced by all operating variables (pvalue<0.05) while their relative intensity followed the order of aeration rate > solution pH > RL concentration > MIBC dosage. Anderson-Darling approach disclosed that the all factors were perfectly followed the normal trend with A2 less than unity and p-value of greater than 0.05 at 95% confidence level. Main effect plots revealed that except MIBC dosage with nonlinear trend, the rest of factors had an ascending influence on the removal efficiency. The process was optimized by interpreting the interaction effect among various variables to reach the maximum dye bioflotation. The maximum removal of 97 ± 0.13% was achieved at pH 12, airflow rate of 5 l/min, MIBC and rhamnolipid concentrations of 30 and 40 ppm, respectively with a flotation kinetic rate of 0.015 sec−1. Finally, the intelligent simulation results showed that the process could be modelled using an artificial bee colony algorithm of 4-7-1 structure with 99% and 98.8% accuracies in the training and testing steps, respectively. Further, we found that the artificial bee colony algorithm was superior to the genetic algorithm in terms of complexity analysis.

Analysis of radar technology identification model for potential geologic hazard based on convolutional neural network and Harris Hawks optimization algorithm

Analysis of radar technology identification model for potential geologic hazard based on convolutional neural network and Harris Hawks optimization algorithm

Convolutional Neural Network Optimization Algorithm-Based Magnetic Resonance Imaging in Analysis of Chronic Pain Caused by the Myofascial Trigger Point

This study was to explore the value of magnetic resonance imaging (MRI) technology processed by convolutional neural network (CNN) optimization algorithms in the clinical research of patients with chronic pain caused by myofascial trigger points (MTrPs). Firstly, referring to the traditional iterative algorithm, this study iterated the convolution network and data consistency layer as a whole for several times, which increased the fitting ability of the data consistency layer and network. When it was applied to magnetic resonance examination, it could be concluded that the effect of its reconstruction method was better than the traditional convolution neural network without the data consistency layer. The image edge was clear, and the restoration effect of details was better. 100 patients with chronic neck pain caused by MTrP were collected and divided into an ultrasound treatment group and a local anesthetic drug injection group, with 50 cases in each group. In addition, 50 healthy volunteers were selected. After clinical treatment, the results showed that, after 3 weeks of treatment, the visual analog score (VAS) and the pain rating index (PRI) of the injection group were 3.16 ± 1.14 points and 4.92 ± 1.26 points, respectively; the present pain intensity (PPI) score was 2.06 ± 0.85 points, and the number of pain days per month was 7.73 ± 1.15. After 1 month of treatment, the VSA and PRI of the injection group were 1.24 ± 0.89 and 1.31 ± 0.97, respectively; the PPI score was 1.34 ± 0.65, and the number of pain days per month was 5.34 ± 0.98. In addition, there were 38 cases reaching the level of clinical cure, accounting for 76%. Therefore, all indicators in the injection group were better than those in the ultrasound treatment group, and the differences were statistically significant ( P &lt; 0.05 ). The results of MRI examination showed that compared with the healthy control group, patients with chronic pain caused by the myofascial trigger point had reduced axial kurtosis (AK), mean kurtosis (MK), and radial kurtosis (RK) in multiple brain areas such as the right parahippocampal gyrus and the right medial prefrontal cortex. In short, chronic pain caused by the trigger point of the myofascial membrane would affect the microstructure of the gray matter of the patient’s brain. In clinical treatment, the efficacy of local anesthetic injection was better than ultrasound therapy.

Design of an IMCPID Optimized Neural Network for Stepless Flow Control of Reciprocating Mechinery

It is usually difficult to design a controller for a nonlinear multiple-input and multiple-output (MIMO) system. The methodological approach taken in this study is a mixed methodology based on a PID-type internal model control (IMC) method and neural network (NN) optimization algorithm. The NN controller is designed for adjusting the sole parameter in IMCPID and compensating the characteristic changes and non-linearity in stepless flow control. In this study, a simulation of a nonlinear MIMO system with strong coupling is carried out. The simulation results indicate that the proposed control method has a better performance in settle time, overshoot, robustness and set-point tracking accuracy compared with other considered methods.

Research on Machine Learning Optimization Algorithm of CNN for FPGA Architecture

In order to meet the application requirements of deep learning detection and recognition algorithm for Field Programmable Gate Array (FPGA) computing structure, this paper proposes a machine learning optimization algorithm of convolutional neural network (CNN) based on pruning and Int8 quantization. By analyzing the machine learning intelligent recognition network, the Yolo v3 target detection network is selected to verify the optimization algorithm. Aiming at the Yolo v3 network with balanced speed and precision, the optimization algorithm is studied in detail. Based on Amazon Cloud FPGA EC2 instance platform and ZCU104 FPGA hardware platform, machine learning optimization algorithms are used to accelerate the application of Yolo v3 network detection and recognition. Satisfactory results are obtained on both FPGA computing platforms. The experimental results show that the machine learning optimization technology can improve the inference speed of neural network based on both FPGA platforms. The network structure optimized by pruning and Int8 quantization algorithms can achieve high acceleration speed, with very limited accuracy loss.

Prophecy of Stock Price of Bharat Immunological &amp; Biological Corporation Ltd using Hybrid ANN and PSO Model

The prediction of the time series has always attracted much interest from investors and researchers to evaluate financial risk. Stock market movements are extremely complex and are influenced by different factors. Hence it is very important to find the most important factors for the stock market. But the high level of noise and complexity of the financial data makes this job very difficult. Many authors have already used artificial neural network for this kind of forecasting tasks, but hybridization model of artificial neural network is considered to be widely used and better performing forecasting model among others. The dormant high noises data mess up the performance, so to enhance the prediction accuracy. We considered a set of seven technical attribute of stock market to perform the hybrid model of Artificial Neural Network (ANN) and Particle Swarm Optimization algorithms. The efficiency of the proposed method is measured by the stock price of Bharat Immunological &amp; Biological Corporation Ltd with 3945 number of daily transactional data. Empirical prediction analysis shows that the proposed model enhances the performance in comparison to simple ANN model.

Neural Network Optimization Algorithms to Predict Wind Turbine Blade Fatigue Life under Variable Hygrothermal Conditions

Moisture and temperature are the most important environmental factors that affect the degradation of wind turbine blades, and their influence must be considered in the design process. They will first affect the resin matrix and then, possibly, the interface with the fibers. This work is the first to use a series of metaheuristic approaches to analyze the most recent experimental results database and to identify which resins are the most robust to moisture/temperature in terms of fatigue life. Four types of resin are compared, representing the most common types used for wind turbine blades manufacturing. Thermoset polymer resins, including polyesters and vinyl esters, were machined as coupons and tested for the fatigue in air temperatures of 20 °C and 50 °C under “dry” and “wet” conditions. The experimental fatigue data available from Sandia National Laboratories (SNL) for wind turbine-related materials have been used to build, train, and validate an artificial neural network (ANN) to predict fatigue life under different environmental conditions. The performances of three algorithms (Backpropagation BP, Particle Swarm Optimization PSO, and Cuckoo Search CS) are compared for adjusting the synaptic weights of the ANN and evaluating the efficiency in predicting the fatigue life of the materials studied, under the conditions mentioned above. For accuracy evaluation, the mean square error (MSE) is used as an objective function to be optimized by the three algorithms.

Thermal error prediction of ball screws based on PSO-LSTM

Thermal error of ball screws seriously affects the machining precision of computerized numerical control (CNC) machine tools especially in high speed and precision machining. Compensation technology is one of the most effective methods to address the thermal issue, and the effect of compensation depends on the accuracy and robustness of the thermal error model. Traditional modeling approaches have major challenges in time series thermal error prediction. In this paper, a novel thermal error model based on long short-term memory (LSTM) neural network and particle swarm optimization (PSO) algorithm is proposed. A data-driven model based on LSTM neural network is established according to the time series collected data. The hyperparameters of LSTM neural network are optimized by PSO, and then a PSO-LSTM model is established to precisely predict the thermal error of ball screws. In order to verify the effectiveness and robustness of the proposed model, two thermal characteristic experiments based on step and random speed are conducted on a self-designed test bench. The results show that the PSO-LSTM model has higher accuracy compared with the radial basis function (RBF) model and back propagation (BP) model with high robustness. The proposed method can be implemented to predict the thermal error of ball screws and provide a foundation for thermal error compensation.

Artificial Intelligence Analysis of EEG Amplitude in Intensive Heart Care.

This article first studied the morphological characteristics of the EEG for intensive cardiac care; that is, based on the analysis of the mechanism of disease diagnosis and treatment, a signal processing and machine learning model was constructed. Then, the methods of signal preprocessing, signal feature extraction, new neural network model structure, training mechanism, optimization algorithm, and efficiency are studied, and experimental verification is carried out for public data sets and clinical big data. Then, the principle of intensive cardiac monitoring, the mechanism of disease diagnosis, the types of arrhythmia, and the characteristics of the typical signal are studied, and the rhythm performance, individual variability, and neurophysiological basis of electrical signals in intensive cardiac monitoring are researched. Finally, the automatic signal recognition technology is studied. In order to improve the training speed and generalization ability, a multiclassification model based on Least Squares Twin Support Vector Machine (LS-TWIN-SVM) is proposed. The computational complexity of the classification model algorithm is compared, and intelligence is adopted. The optimization algorithm selects the parameters of the classifier and uses the EEG signal to simulate the model. Support Vector Machines and their improved algorithms have achieved the ultimum in shallow neural networks and have achieved good results in the classification and recognition of bioelectric signals. The LS-TWIN-SVM algorithm proposed in this paper has achieved good results in the classification and recognition of bioelectric signals. It can perform bioinformatics processing on intensive cardiac care EEG signals, systematically biometric information, diagnose diseases, the real-time detection, auxiliary diagnosis, and rehabilitation of patients.