Search

• Five Machine Learning (ML) models compared for wave height time series prediction. • Complex ML models do not outperform simple AR in wave height time series prediction. • Comment to related papers: signal decomposition in test set series is WRONG. Significant Wave Height (SWH) is crucial in many aspect of ocean engineering. The accurate prediction of SWH has therefore been of immense practical value. Recently, Artificial Intelligence (AI) time series prediction methods have been widely used for single-point short-term SWH time-series forecasting, resulting in many AI-based models claiming to achieve good results. However, the extent to which these complex AI models can outperform traditional methods has largely been overlooked. This study compared five different models - AutoRegressive (AR), eXtreme Gradient Boosting (XGB), Artificial Neural Network (ANN), Long Short-Term Memory (LSTM), and WaveNet - for their performance on SWH time series prediction at 16 buoy locations. Surprisingly, the results suggest that the differences of performance among different models are negligible, indicating that all these AI models have only "learned" the linear auto-regression from the data. Additionally, we noticed that many recent studies used signal decomposition method for such time series prediction, and most of them decomposed the test sets, which is WRONG. [ABSTRACT FROM AUTHOR]

Background: MicroRNAs (miRNAs) emerge in various organisms, ranging from viruses to humans, and play crucial regulatory roles within cells, participating in a variety of biological processes. In numerous prediction methods for miRNA-disease associations, the issue of over-dependence on both similarity measurement data and the association matrix still hasn't been improved. In this paper, a miRNA-Disease association prediction model (called TP-MDA) based on tree path global feature extraction and fully connected artificial neural network (FANN) with multi-head self-attention mechanism is proposed. The TP-MDA model utilizes an association tree structure to represent the data relationships, multi-head self-attention mechanism for extracting feature vectors, and fully connected artificial neural network with 5-fold cross-validation for model training. Results: The experimental results indicate that the TP-MDA model outperforms the other comparative models, AUC is 0.9714. In the case studies of miRNAs associated with colorectal cancer and lung cancer, among the top 15 miRNAs predicted by the model, 12 in colorectal cancer and 15 in lung cancer were validated respectively, the accuracy is as high as 0.9227. Conclusions: The model proposed in this paper can accurately predict the miRNA-disease association, and can serve as a valuable reference for data mining and association prediction in the fields of life sciences, biology, and disease genetics, among others. [ABSTRACT FROM AUTHOR]

This paper focuses on building a ground motion model (GMM) using artificial neural network technique driven by a synthetically generated database in a data poor region. A hybrid simulated method considering nonlinear site response is used to generate potential ground motions in Tianjin area. The synthetic database containing a total of 9267 ground motion records from 93 simulated earthquake events are utilized to develop the model. A back propagation neural network developed by genetic algorithm (GA-BPNN) is used to obtain the optimal weights and biases that most fit the target dataset without overfitting. The structure of GA-BPNN model consists of 322 unknowns including weights and biases that connect input and output parameters. The model is developed to predict PGA and 5% damped spectral acceleration (periods from 0.01 to 10 s). The input parameters consist of moment magnitude (Mw), strike (φ), dip (δ), rake angle (λ), focal depth (h), epicenter distance (R) and azimuth (θ). The model performance is observed to be within the appropriate error limits. It is found that the model has the ability to capture the parametric variation relationship. Additionally, the synthetically based model is compared with empirically based GMM relations derived from NGA-West2, finding it has similar performance and behavior to leading GMMs. The model application in Tianjin area demonstrates that this paper provides a scheme to utilize synthetic database and build GMMs in data poor regions. [ABSTRACT FROM AUTHOR]

In recent years, Model Predictive Torque Control (MPTC) has gained popularity as a powerful method of controlling Induction Motors (IMs). There are, however, some factors such as the weighted coefficient that must be taken into account to attain satisfactory performance at different operation points for stator flux. As of now, an empirical procedure has been used for tuning the weighting factor in MPTC, which is not as effective. Model Predictive Flux Control (MPFC) is proposed in this paper, which eliminates complicated process of predicting the stator current by using the flux vectors as control variables. In order to get over these obstacles, a novel Model Predictive Control (MPC) method utilizing a Cascaded Artificial Neural Network (ANN) is proposed in this research. As a result, the control complexity is diminished and weighting factor in standard MPTC is eliminated. In-depth evaluations and comparisons of MPTC and MPFC are conducted, covering low-speed function, dynamic response, and steady-state performance. In contract to traditional approaches, the preferred method has much lower computational cost, virtuous dynamic response, and superior steady-state performance. The overall proposed work is simulated using MATLAB/SIMULINK platform and experimental tests performed on 3Φ VSI fed IM drives are used to verify the efficacy of proposed approach. [ABSTRACT FROM AUTHOR]

The credit risk assessment acts as an important part in daily affairs for financial institutions. But in the era of big data, the growing business volume makes it an urgent demand to develop digital ways of credit risk assessment. Currently, the machine learning is universally employed to establish various data-driven models for this purpose. However, machine learning models generally suffer from limited ability of feature representation and robustness, and cannot deal with more complex financial security scenarios. To deal with this issue, this work introduces ensemble learning to construct a stronger credit risk prediction model via integration of several basic machine learning models. Thus, an ensemble learning-enhanced smart prediction model for financial credit risk is proposed in this paper. Three classification-based machine learning models (support vector machine, artificial neural network and radial basis function) are selected as the basic classifiers, and "voting" strategy is utilized to integrate them into a novel strong classifier. A real-world financial credit dataset released by a Chinese commercial bank was selected as the experimental scenario. The obtained results show that the proposal has better prediction accuracy compared with basic machine learning models without ensemble learning. [ABSTRACT FROM AUTHOR]

In recent studies, artificial intelligence models have been developed for the prediction of nitrogen oxide emissions from thermal power plants. This research utilizes an artificial intelligence prediction model with more coal input features than boiler features, including volatility, ash content, sulfur content, fixed carbon, total moisture, calorific value, grinding rate, fuel ratio, coal feeding rate, boiler efficiency, total air volume, and excess air volume. The paper delves into the importance analysis of input features for artificial intelligence models. Moreover, feature importance analysis is not only a prerequisite for predicting nitrogen oxide emissions but also a study providing insights into model performance. An artificial neural networks (ANN) regression model is employed to predict nitrogen oxide emissions, and the results demonstrate that the number of feature importance significantly impacts model performance. The best model performance is achieved with eight specific input features. Finally, after the training and validation processes, the ANN model yields the optimal coefficient of determination (R2) value. • Development of an artificial intelligence prediction model for nitrogen oxide emissions from thermal power plants. • A comprehensive set of input features is utilized for emissions prediction. • Importance analysis of input features significantly impacts model performance. • Optimal model performance was achieved with eight specific input features. • The robust coefficient of determination (R2) value demonstrates the accuracy of the ANN model. [ABSTRACT FROM AUTHOR]

To achieve successful prevention of fire incidents originating from human activities, it is imperative to possess a thorough understanding. This paper introduces a machine learning approach, specifically utilizing deep neural networks (DNN), to develop predictive models for fire occurrence in Keelung City, Taiwan. It investigates ten factors across demographic, architectural, and economic domains through spatial analysis and thematic maps generated from geographic information system data. These factors are then integrated as inputs for the DNN model. Through 50 iterations, performance indices including the coefficient of determination (R2), root mean square error (RMSE), variance accounted for (VAF), prediction interval (PI), mean absolute error (MAE), weighted index (WI), weighted mean absolute percentage error (WMAPE), Nash–Sutcliffe efficiency (NS), and the ratio of performance to deviation (RPD) are computed, with average values of 0.89, 7.30 × 10−2, 89.21, 1.63, 4.90 × 10−2, 0.97, 2.92 × 10−1, 0.88, and 4.84, respectively. The model's predictions, compared with historical data, demonstrate its efficacy. Additionally, this study explores the impact of various urban renewal strategies using the DNN model, highlighting the significant influence of economic factors on fire incidence. This underscores the importance of economic factors in mitigating fire incidents and emphasizes their consideration in urban renewal planning. [ABSTRACT FROM AUTHOR]

The reliability of thimble tubes plays a critical role for maintaining the safety of a nuclear power plant. The defect depth needs to be quantified and predicted to support the operational decision-making. This paper presents a method to quantify the defects on thimble tube wall based on the analyzation of eddy current testing (ECT) data. Then, a method using artificial neural network (ANN) to predict the detect depth is studied. The tubes are divided into 2 shapes and four regions according to their positions and the data of each region and each shape is expanded by mean interpolation. A prediction model based on ANN is constructed for each shape in each region. The experimental results show that the model can predict the signal of the next year according to the signal of the previous three years with mean absolute percentage error less than 16%. [ABSTRACT FROM AUTHOR]

Banking and stock markets consider gold to be an important component of their economic and financial status. There are various factors that influence the gold price trend and its fluctuations. Accurate and reliable prediction of the gold price is an essential part of financial and portfolio management. Moreover, it could provide insights about potential buy and sell points in order to prevent financial damages and reduce the risk of investment. In this paper, different architectures of deep neural network (DNN) have been proposed based on long short-term memory (LSTM) and convolutional-based neural networks (CNN) as a hybrid model, along with automatic parameter tuning to increase the accuracy, coefficient of determination, of the forecasting results. An illustrative dataset from the closing gold prices for 44 years, from 1978 to 2021, is provided to demonstrate the effectiveness and feasibility of this method. The grid search technique finds the optimal set of DNNs' parameters. Furthermore, to assess the efficiency of DNN models, three statistical indices of RMSE, RMAE, and coefficient of determination (R2), were calculated for the test set. Results indicate that the proposed hybrid model (CNN-Bi-LSTM) outperforms other models in total bias, capturing extreme values and obtaining promising results. In this model, CNN is used to extract features of input dataset. Furthermore, Bi-LSTM uses CNN's outputs to predict the daily closing gold price. [ABSTRACT FROM AUTHOR]

The reusability of by-products in the food industry is consistent with sustainable and greener production; therefore, the aim of this paper was to evaluate the applicability of multiple linear regression (MLR), piecewise linear regression (PLR) and artificial neural network models (ANN) to the prediction of grape-skin compost's physicochemical properties (moisture, dry matter, organic matter, ash content, carbon content, nitrogen content, C/N ratio, total colour change of compost samples, pH, conductivity, total dissolved solids and total colour change of compost extract samples) during in-vessel composting based on the initial composting conditions (air-flow rate, moisture content and day of sampling). Based on the coefficient of determination for prediction, the adjusted coefficient of determination for calibration, the root-mean-square error of prediction (RMSEP), the standard error of prediction (SEP), the ratio of prediction to deviation (RPD) and the ratio of the error range (RER), it can be concluded that all developed MLR and PLR models are acceptable for process screening. Furthermore, the ANN model developed for predicting moisture and dry-matter content can be used for quality control (RER >11). The obtained results show the great potential of multivariate modelling for analysis of the physicochemical properties of compost during composting, confirming the high applicability of modelling in greener production processes. [ABSTRACT FROM AUTHOR]

Modern data mining techniques using machine learning (ML) and deep learning (DL) algorithms have been shown to excel in the regression-based task of materials property prediction using various materials representations. In an attempt to improve the predictive performance of the deep neural network model, researchers have tried to add more layers as well as develop new architectural components to create sophisticated and deep neural network models that can aid in the training process and improve the predictive ability of the final model. However, usually, these modifications require a lot of computational resources, thereby further increasing the already large model training time, which is often not feasible, thereby limiting usage for most researchers. In this paper, we study and propose a deep neural network framework for regression-based problems comprising of fully connected layers that can work with any numerical vector-based materials representations as model input. We present a novel deep regression neural network, iBRNet, with branched skip connections and multiple schedulers, which can reduce the number of parameters used to construct the model, improve the accuracy, and decrease the training time of the predictive model. We perform the model training using composition-based numerical vectors representing the elemental fractions of the respective materials and compare their performance against other traditional ML and several known DL architectures. Using multiple datasets with varying data sizes for training and testing, We show that the proposed iBRNet models outperform the state-of-the-art ML and DL models for all data sizes. We also show that the branched structure and usage of multiple schedulers lead to fewer parameters and faster model training time with better convergence than other neural networks. Scientific contribution: The combination of multiple callback functions in deep neural networks minimizes training time and maximizes accuracy in a controlled computational environment with parametric constraints for the task of materials property prediction. [ABSTRACT FROM AUTHOR]

The widespread risks of leakages in the hydrogen industry chain require a method that can quickly predict the consequences of accidents, especially in the hydrogen refueling station (HRS). This paper presents a surrogate model based on physics-informed neural network (PINN) that can predict the distribution of hydrogen concentration after a leakage. The proposed Physics-informed Convolutional Long Short-Term Memory Network (PI-ConvLSTM) model improves the concentration prediction results at the gas cloud boundary by adding a physical constraint term to the loss function of the ConvLSTM model. The concentration distributions after hydrogen leakage at HRS simulated by FLACS are used as the training samples, and the concentration data are converted into grayscale maps for training. The hydrogen concentration prediction method with the proposed surrogate model as the core achieves fast prediction of the gas cloud concentration distribution with acceptable accuracy. It is observed that the method can greatly reduce the prediction time of the consequences of hydrogen leak accidents with the surrogate model already trained. It can provide real-time risk warning and consequence prediction for hydrogen refueling station leakage accidents. • The hydrogen leak consequences data of HRS is generated using CFD. • ConvLSTM-based surrogate model can predict hydrogen concentration in HRS. • Physics-informed loss function improves accuracy of cloud boundary prediction. • Proposed surrogate model greatly reduces the prediction time compared to CFD. [ABSTRACT FROM AUTHOR]

BACKGROUND: Milling operations of laminae in spinal surgery generate high temperatures, which can lead to thermal injury and osteonecrosis and affect the biomechanical effects of implants, ultimately leading to surgical failure. OBJECTIVE: In this paper, a backpropagation artificial neural network (Bp-ANN) temperature prediction model was developed based on full factorial experimental data of laminae milling to optimize the milling motion parameters and to improve the safety of robot-assisted spine surgery. METHODS: A full factorial experiment design were used to analyze the parameters affecting the milling temperature of laminae. The experimental matrixes were established by collecting the corresponding cutter temperature T c and bone surface temperature T b for the milling depth, feed speed and different bone densities. The Bp-ANN lamina milling temperature prediction model was constructed from experiment data. RESULTS: Increasing milling depth increases bone surface and cutter temperature. Increasing feed speed had little effect on cutter temperature, but decreased bone surface temperature. Increasing bone density of laminae increased cutter temperature. The Bp-ANN temperature prediction model had best training results in the 10th epoch, and there is no overfitting (training set R = 0.99661, validation set R = 0.85003, testing set R = 0.90421, all temperature data set R = 0.93807). The goodness of fit R of Bp-ANN was close to 1, indicating that the predicted temperature was in good agreement with the experiment measurements. CONCLUSION: This study can help spinal surgery-assisted robot to select appropriate motion parameters at different density bones to improve lamina milling safety. [ABSTRACT FROM AUTHOR]

A time-varying, nonlinear soil-plant system contains many unknown elements that can be quantified based on analytical methodologies. Artificial neural networks (ANNs) are a widely used mathematical computing, modeling, and predicting methods that estimate unknown values of variables from known values of others. This paper aims to simulate the relationship between soil moisture, bulk density, porosity ratio, depth, and penetration resistance and to estimate soil penetration resistance with the help of ANNs. For this aim, the generalized regression neural network (GRNN) and radial basis function (RBF) models were developed and compared for the estimation of soil penetration resistance values in MATLAB. A dataset of 153 samples was collected from experimental field. From the 153 data, 102 data (33%) were selected for training and the remaining 51 data (67%) were used for testing. The estimation process implemented 10 replications using randomly selected testing and training data. mean squared error (MSE), root mean square error (RMSE), and mean absolute error (MAE) were used to evaluate estimation accuracy on the developed ANN methods. Based on MSE, RMSE, MAE and standard deviation, statistical results showed that the GRNN modeling presented better results than the RBF model in predicting soil penetration resistance success. [ABSTRACT FROM AUTHOR]

Background: Detecting breast cancer in its early stages remains a significant challenge in the present context and is a leading cause of death among women, primarily due to delayed identification. This paper presents a practical and accurate approach based on deep learning to identify breast cancer in cytology images. Method: The analytical approach leverages knowledge from a related problem through a technique known as transfer learning. Convolutional neural networks (CNNs) are employed due to their remarkable performance on large datasets. Image classification architectures such as Google network (GoogleNet), Visual geographical group network (VGGNet), residual network (ResNet), and dense convolution network (DenseNet) are utilized in this approach. By applying transfer learning, the images are classified into two categories: those containing cancer cells and those without them. The performance of the proposed ensemble method is evaluated using a breast cytology image dataset. Results: The results of our proposed ensemble framework outperform conventional CNN models in terms of precision, recall, and F1 measures, achieving an impressive 86% prediction accuracy. Visual representations of validation graphs for each classifier demonstrate that the ensemble framework surpasses the performance of pre-trained CNN architectures. Conclusion: Combining the outcomes of conventional CNN architectures into an ensemble framework enhances early breast cancer detection, leading to a reduction in mortality through timely medical interventions. [ABSTRACT FROM AUTHOR]

Traditional risk analysis models often struggle to provide effective insights into real-time and objective risk assessment of ship navigation. Therefore, this paper introduces a ship navigation risk online prediction model (NROPM) based on the fusion of multi-source data. Firstly, this paper extracts risk feature information from multiple sensors, such as AIS data, wind, and water currents. These data are integrated through feature fusion to create the research dataset. Subsequently, the entropy weight-cloud model is constructed to calculate the risk and generate a set of test samples. An emerging informer network is introduced to train the model and predict the risk of ship navigation. Finally, this paper compares the proposed model with traditional neural network models through comparative experimental. The research was conducted using a real dataset from the Pearl River area over a period of three months. Model performance was assessed in terms of mean squared error and training time consumption. The experimental results indicate that the proposed NROPM outperforms traditional models like the BP (backpropagation) neural network and transformer in both accuracy and computational efficiency for real-time ship navigation risk prediction. • This paper proposes a ship navigation risk online prediction model using multi-source data. • An improved informer network is used to achieve the real-time risk prediction. • A quantification method for ship navigation risk is introduced, and the accuracy of the evaluation results is verified using entropy-weighted cloud models. • Experimental results show this model's superior accuracy over traditional methods, enabling objective, dynamic ship navigation risk assessment in busy waters. [ABSTRACT FROM AUTHOR]

Deep neural networks have revolutionized many fields, including image processing, inverse problems, text mining and more recently, give very promising results in systems and control. Neural networks with hidden layers have a strong potential as an approximation framework of predictive control laws as they usually yield better approximation quality and smaller memory requirements than existing explicit (multi-parametric) approaches. In this paper, we first show that neural networks with HardTanh activation functions can exactly represent predictive control laws of linear time-invariant systems. We derive theoretical bounds on the minimum number of hidden layers and neurons that a HardTanh neural network should have to exactly represent a given predictive control law. The choice of HardTanh deep neural networks is particularly suited for linear predictive control laws as they usually require less hidden layers and neurons than deep neural networks with ReLU units for representing exactly continuous piecewise affine (or equivalently min–max) maps. In the second part of the paper we bring the physics of the model and standard optimization techniques into the architecture design, in order to eliminate the disadvantages of the black-box HardTanh learning. More specifically, we design trainable unfolded HardTanh deep architectures for learning linear predictive control laws based on two standard iterative optimization algorithms, i.e., projected gradient descent and accelerated projected gradient descent. We also study the performance of the proposed HardTanh type deep neural networks on a linear model predictive control application. [ABSTRACT FROM AUTHOR]

• Developing optimized neural network-based models for trapdoor stability prediction. • New FELA solutions for 3D rectangular trapdoors in anisotropic clays. • Developing hybrid machine learning models based on ANNs with various optimization algorithms. • Examining several performance metrics, regression plots, and rank analysis. The problem of trapdoor stability is a crucial problem in geotechnical engineering. This study is the first to introduce novel neural network-based metaheuristic models for the stability prediction of 3D rectangular trapdoors in anisotropic and nonhomogeneous clays. However, no researcher has considered such trapdoor problems in the past. In this study, the dataset is obtained by using finite element limit analysis (FELA). The proposed hybrid machine learning models based on artificial neural networks (ANNs) and various types of optimization algorithms (OAs), namely, ant colony optimization (ACO), artificial lion optimization (ALO), the imperialist competition algorithm (ICA), and shuffled complex evolution (SCE), are also proposed in this paper by undergoing rigorous optimization to ensure accuracy and efficiency in capturing the intricate dynamics of the stability investigations from the models. The performance of the proposed ANN-based models is assessed using several performance metrics, regression plots, and rank analysis. The proposed ANN-SCE model outperforms the other proposed models in predicting trapdoor stability, where the ANN-SCE model achieved the highest rank, with a score of 58, followed by the ANN-ALO (47), ANN-ICA (33), and ANN-ACO (22) models. The proposed neural network-based metaheuristic models deliver precise and effective forecasts of trapdoor stability to make informed decisions concerning road design and mitigation tactics, ultimately improving the robustness of infrastructure facing geotechnical challenges. [ABSTRACT FROM AUTHOR]

• A LDNE-stacking integration fusion spatio-temporal feature multi-step prediction model is proposed. It shows excellent predictive performance in spatio-temporal prediction. • Moran index is used to measure the clustering relationship and significant level between solar irradiance and spatial feature variables. It is a novel way to model spatial correlation. • Embedding the structured LSTM deep network into the Stacking model. • Comparing ten different models, including LSTM, CNN-LSTM and traditional integrated models. Solar irradiance prediction of multi-station is influenced by complex spatial characteristics and temporal fluctuations, so how to obtain accurate short-term irradiance prediction by using photovoltaic cluster data becomes a challenge. In order to address the challenge, a structural depth network embedding stacking model based on Moran's I and seasonal trend for short-term solar irradiance prediction is proposed. Firstly, the deep spatial characteristics of photovoltaic clusters are analyzed by using bivariate Moran index to capture strong correlations between stations. Secondly, seasonal trend decomposition is used to smooth the seasonal factors in time series, and Complete Ensemble Empirical Mode Decomposition with Adaptive Noise is used to decompose the residual components to reduce the fluctuation of signals and the complexity of data feature mapping. Finally, based on the Stacking framework, Long Short-term Memory Network, Least Absolute Selection and Shrinkage Operator, Gradient Boosted Decision Tree, and Artificial Neural Network are superimposed. And the aggregated sub-modal components are introduced into the robust integration model to predict the solar irradiance in multiple steps. Through the case study of the National Renewable Energy Laboratory open-source dataset in the United States, the conclusions are as follows: (1) The spatio-temporal method in this paper can effectively extract the deep spatial characteristics of photovoltaic stations. (2) The fusion strategy of embedded structured deep network effectively improves the stacking effect. (3) Comparing with many classical forecasting models, it shows that this method can effectively improve the forecasting effect. [ABSTRACT FROM AUTHOR]

• Introduction of a generalizable Bayesian recurrent neural network framework for uncertainty quantification in traffic flow prediction. • Utilization of spectral normalization in hidden layers of neural network to enhance generalizability and improve uncertainty estimates. • Demonstration of how normalization alters the training process, controlling model complexity and reducing overfitting. • Spectral normalization improves generalization performance of deep learning models on out-of-distribution datasets, which is crucial for traffic management applications. • Ability of spectral normalization to localize latent feature space under data perturbations, leading to improved generalizability. • Significant outperformance of spectral normalization over layer normalization and models without normalization in both single and multistep prediction horizons. Deep-learning models for traffic data prediction can have superior performance in modeling complex functions using a multi-layer architecture. However, a major drawback of these approaches is that most of these approaches do not offer forecasts with uncertainty estimates, which are essential for traffic operations and control. Without uncertainty estimates, it is difficult to place any level of confidence in the model predictions, and operational strategies relying on overconfident predictions can lead to worsening traffic conditions. In this study, we propose a Bayesian recurrent neural network framework for uncertainty quantification in traffic prediction with higher generalizability by introducing spectral normalization to its hidden layers. In our paper, we have shown that normalization alters the training process of deep neural networks by controlling the model's complexity and reducing the risk of overfitting the training data. This, in turn, helps improve the generalization performance of the model on out-of-distribution datasets. Results demonstrate that spectral normalization improves uncertainty estimates and significantly outperforms both the layer normalization and model without normalization in both single and multistep prediction horizons. This improved performance can be attributed to the ability of spectral normalization to better localize the latent feature space of the data under perturbations. Our findings are especially relevant to traffic management applications, where predicting traffic conditions across multiple locations is the goal, but the availability of training data from multiple locations is limited. Spectral normalization, therefore, provides a more generalizable approach that can effectively capture the underlying patterns in traffic data without requiring location-specific models. [ABSTRACT FROM AUTHOR]

Performance prediction plays a key role in product quality control and quality optimization. However, due to the error transmission in the refrigeration compressor assembly process and the imbalance of data collected during the refrigeration compressor assembly process, existing performance prediction methods cannot accurately extract key features of the assembly process resulting in low prediction accuracy. In this paper, a knowledge graph neural network (AKGNN) model with predictive value calibration is proposed for refrigeration compressor performance prediction in two stages. First, a sequence graph attention feature extraction module is designed to extract key features of the assembly process data to cope with the assembly error propagation. Second, to deal with the imbalanced distribution of compressor data, a predictive value calibration (PC) module is designed to calibrate the output of the first stage. Experimental results demonstrate the proposed method outperformed all compared cutting-edge methods in four evaluated indicators. [ABSTRACT FROM AUTHOR]

The use of machine learning models for productivity prediction in complex manufacturing systems has garnered significant attention. However, the implementation of conventional black-box machine learning models is challenging due to the complexity and limited data availability in aircraft final assembly lines. These challenges stem from a lack of explainability, leading to compromised accuracy and diminished trustworthiness. To address this issue, this paper proposes an explainable neural network model, the Attention-SP-LSTM-FIG, specifically designed for productivity prediction in aircraft final assembly lines. First, the serial-parallel LSTM is utilized to map the process sequence within each assembly station, and a customized attention mechanism maps the requirements of workers and materials. This process results in independent sub-models corresponding to each station. Subsequently, the outputs from each sub-model are combined using a specially-designed final integrated gate to produce the final output. Finally, a post-hoc analysis of the weight data in the model is performed following the backpropagation order to identify the various production bottlenecks in the assembly line. The performance of the proposed model has been evaluated through an industrial case study. In terms of accuracy, the Attention-SP-LSTM-FIG surpasses other benchmark neural network models in terms of error, correlation, and precision metrics. Regarding explainability, the Attention-SP-LSTM-FIG can precisely identify bottlenecks for practitioners, thus facilitating a better understanding of the rationale behind the model's outputs. [ABSTRACT FROM AUTHOR]

Video anomaly detection aims to identify a set of abnormal events in videos. Deep reconstruction and prediction-based models have been employed to detect anomalies. Deep reconstruction models sometimes recreate the abnormal events along with the normal ones. However, the prediction-based approaches have demonstrated encouraging results. This paper presents a video vision transformer (ViViT) based generative adversarial network (GAN), TransGANomaly, a novel approach for detecting anomalies. The proposed framework is a video frame predictor and trained only on normal video data adversarially. The generator of the GAN is a ViViT network that receives 3D input tokens from the video snippets. The generator aims to predict the future frame based on past sequences. After that, the predicted and original frames are given to the model's discriminator for binary classification. Extensive experiments have been performed on UCSD Pedestrian, CUHK Avenue, and ShanghiaTech datasets to validate the efficacy of the proposed method. [ABSTRACT FROM AUTHOR]

This paper aims to develop a unique artificial neural network (ANN)-based equation as well as MATLAB- and Python-based graphical user interfaces (GUIs) using the most comprehensive and up-to-date database for predicting the behaviour of axially loaded concrete-filled double skin tubular (CFDST) short and slender columns with normal- and high-strength materials. Two machine learning (ML) methods, which are ANN and extreme gradient boosting (XGBoost), are trained and tested using 1721 sets of data, with 129 of them collected from experimental studies and 1592 generated by finite element (FE) simulations. The accuracy of the developed ML models is assessed through comparing their predictions with the experimental and FE results. To demonstrate the effect of each parameter on the predicted results, the SHapley Additive exPlanations (SHAP) method is used. The developed ML models are also used to conduct parametric studies to examine the effect of geometric and material parameters on the predicted results. The accuracy of the ML models and the proposed ANN-based equation in predicting the ultimate axial capacity of CFDST columns is compared with that of six design methods including two design code provisions and four design equations proposed by researchers. A numerical example is presented to illustrate the design procedure of the CFDST column using the proposed ANN-based equation. The results indicate that the ANN model performs better on unseen data than the XGBoost model with lower root mean square error for the test set. The results also show that the ML models and the proposed ANN-based equation are superior to the other design models in prediction accuracy. • 1721 data samples of CFDST columns are used to train and test the ML models. • Optimized ANN and XGBoost models are used to predict the behaviour of CFDST columns. • ANN-based equation, MATLAB- and Python-based GUIs are developed. • Parametric studies are conducted using the trained ML models. • The superiority of the proposed design models over six design equations is shown. [ABSTRACT FROM AUTHOR]

The failure of concrete for hydraulic structures due to the excessive abrasion depth has caused great economic losses. Despite the tremendous work that has been done, the durability design of hydraulic concrete structures concerning abrasion damage is still a difficult task, especially with a long service life aiming at economic benefit and environmental sustainability. In this paper, machine learning methodologies including random forests (RFs) and artificial neural networks (ANNs), are used to establish prediction models based on concrete mixture proportions, curing age, and hydraulic conditions. Furthermore, RF models are coupled with multiple objective particle swarm optimization (MOPSO) algorithms to perform computational design optimization of concrete mixtures. The optimal solutions are generated based on 5000 randomly generated mixture proportions, among which three optimal designs of concrete mixtures are selected to guide field applications on account of concrete durability requirements and economic benefits. [ABSTRACT FROM AUTHOR]

Viscoelastic characterization of asphalt binders is critical for selecting suitable binders for a given climatic region and for improving long-term performance prediction using mechanistic-empirical pavement analysis and design methods, such as the Mechanistic-Empirical-Pavement-Design-Guide (MEPDG). A-VTS information (viscosity data) was used in the past and existing databases have this type of information. Nowadays, Dynamic Shear Modulus (|G*|) and Phase Angle (δ) determination using the Dynamic Shear Rheometer (DSR) is common practice. This resulted in the existence of old and new databases in highway agencies. The goal is to find a way to combine these separate data sets and merge them into a unified national database based on |G*| and δ. The work discussed in this paper is critical to make this merger possible and to use this in future binder to mixture predictive model developments and in MEPDG. In this study, |G*| and δ prediction models were developed. A machine learning technique named Artificial Neural Network (ANN) was employed in this process. 8940 data points from 41 binders were utilized in the model development process. After robust training, validation, and testing process, a |G*| prediction model with seven neurons and a δ prediction model with three neurons were reported. The models were compared with previous Bari-Witczak and Onifade-Birgisson models. ANN-based models performed better in every statistical parameter. Two equations were also extracted from the ANN models, which can be used reliably to predict |G*| and δ from A-VTS data. These two models with closed-form equations can be incorporated into the MEPDG for |G*| and δ predictions. • New models for Binder dynamic shear modulus and phase angle using binder viscosity data with temperature and frequency. • Artificial Neural Network (ANN) is utilized in the model development process. • An equation is extracted from each ANN-based model. • The prediction performance of new models is better than existing models in every statistical parameter. [ABSTRACT FROM AUTHOR]

Ground motion prediction is an important and complex research subject in earthquake engineering and traditional approaches based on statistical regression have much room for improvement in prediction accuracy. Utilizing the 67,164 ground motion records from KiK-net and K-Net for 777 shallow crustal earthquakes between 1997 and 2019 in Japan, this paper proposes a ground motion prediction model XGBoost-SC based on the machine learning algorithm of eXtreme Gradient Boosting (XGBoost) for Japan. Magnitude, focal depth, hypo-central distance, V s30 , site altitude, and focal mechanism were used as the feature parameters, and XGBoost, Random Forest, and Deep Neural Networks (DNN) algorithms were selected for model training while Bayesian optimization was used to search for optimized hyperparameters to improve the prediction accuracy. XGBoost algorithm was selected for further study based on the comparison of results from the three algorithms. Residual change with magnitude and hypo-central distance, the probability distribution of residuals, residual standard deviation (σ), residual mean squared error (MSE), and Pearson correlation coefficient (R) were used as the evaluation parameters, and a comparison study was performed against the ground motion prediction equation based on traditional approaches. Actual earthquake events were selected to compare the prediction results against the observation records. To further validate and explain the proposed model, SHapley Additive exPlanations (SHAP) analysis was performed to explain the impact of selected feature parameters on the proposed model. The results demonstrate that the proposed XGBoost-SC model has good prediction stability for all periods, and its residual errors are smaller than those of other models. Therefore, the proposed model can better reflect the ground motion attenuation for shallow crustal earthquakes in Japan and can serve as a better model for ground motion prediction in future aseismic design and earthquake disaster mitigation efforts. • A Ground Motion Prediction Model was proposed for shallow crustal earthquakes. • The proposed model was more precise than traditional approaches. • Explained the effect of feature parameters on the prediction results. [ABSTRACT FROM AUTHOR]