Root Mean Square Error Percentage Research Articles

DeepInvest: Stock Market Predictions with a Sequence-Oriented BiLSTM Stacked Model – A Dataset Case Study of AMZN

Intelligent forecasters are now being considered in the stock market, providing essential insights and strategic guidance to investors and traders by presenting analytical tools and predictive models, thus enabling informed decision-making and mitigating financial risks in this dynamic market. The importance of intelligent analyzers in stock trading routines is considered in this work, where DeepInvest, a multimodal deep learning model tailored for stock price prediction, is introduced. Employing a Sequence-Oriented, Long-Term Dependent (SoLTD) architecture featuring Bidirectional Long Short-Term Memory (BiLSTM) networks, DeepInvest is applied to essential features of the Amazon Corp. (AMZN) market dataset, gathered from Yahoo Finance, including Closing, Opening, High, Low, Volume, and Adj Close prices. Exceptional performance in forecasting Closing, Opening, High, Low, and Adj Close prices is demonstrated, with minimal Mean Absolute Percentage Error (MAPE) and Root Mean Squared Percentage Error (RMSPE) scores, coupled with high R-squared (R2) values, manifesting a robust fit to the data, as well as computational complexity, and Rates Per Second (RPS) metrics in comparison to other models of KNN, LSTM, RNN, CNN, and ANN. Finally, challenges in the accurate prediction of trading volumes are identified, highlighting an area for future enhancement.

Bone age estimation with HS-optimized Resnet and Yolo for child growth disorder

Bone Age Assessment (BAA) plays a significant part in assessing skeletal maturity for preventing different types of skeletal disorders in the human body. Predicting BAA helps clinicians make decisions about the patient’s biological maturity. An increase or decrease in bone age (BA) causes serious issues for pediatricians who need advanced equipment for BA consistency with the chronological age. Recent techniques fail to accurately estimate the region over interest (ROI) of hand-wrist images to predict BA under low time complexity accurately. Hence, this article brings novel deep learning (DL) techniques to assess the BA to overcome child growth disorders accurately. The proposed methodology undergoes three major stages: pre-processing, ROI detection, and BA detection. In the pre-processing stage, contrast-limited fuzzy adaptive histogram equalization (CLFAHE) is done to maintain the originality of the hand image. The size of the image is also increased by using data augmentation. After pre-processing, the required hand bone region with ROI detection approaches you only live once version 3 (YOLOv3). At the final stage, the BA is predicted using the residual network-152 (ResNet152) technique. Finally, the weight updation takes place in the ResNet152 model using harmony search (HS) optimization techniques. The dataset used in this research is collected from the RSNA bone age dataset. The performance measures such as mean absolute error (MAE), root mean square error (RMSE), and root mean squared percentage error (RMSPE) are analyzed and compared with other existing measures. In an experimental scenario, the MAE of 2.59, RMSE of 2.97, and RMSPE of 0.35 for the male gender is obtained.

Enhancing capacity estimation of retired electric vehicle lithium-ion batteries through transfer learning from electrochemical impedance spectroscopy

The low economic feasibility caused by inefficient testing and inaccurate performance estimation is one of the main bottlenecks in the echelon utilization of large-scale retired batteries. This study proposes a fast and accurate capacity estimation method for retired batteries based on electrochemical impedance spectroscopy (EIS). Firstly, the EIS of the batteries that experience multi-condition aging in the laboratory are collected. EIS characteristic parameter sequences highly related to battery performance, including real part and magnitude, are directly extracted to establish a base bi-directional long short-term memory model. Secondly, a transfer learning method based on feature matching is designed, which applies a linear transformation layer to map the features between the source and target domains. The proposed transfer learning method has been effectively validated on laboratory battery data measured at different temperatures and retired battery datasets of different material types. The improvements are especially notable for retired batteries. The detection time has been reduced, with each cell requiring only 1.67 min. And using only a small amount of data as input for transfer learning can achieve an accuracy improvement of over 90 %, indicating an effective transfer channel from the base model established on laboratory small-capacity battery aging data to large-capacity retired battery data is successfully established for the first time. For retired nickel-cobalt-manganese batteries, the mean absolute percentage error (MAPE) and the root mean square percentage error (RMSPE) are 2.33 % and 2.75 %, respectively, while for retired lithium-iron-phosphate batteries, the MAPE and RMSPE reached 4.12 % and 5.04 %, respectively. The results demonstrate the proposed method reduces the cost of repeated testing, modeling, and training for specific retired batteries while maintaining the accuracy of capacity estimation. This advancement helps to improve the efficiency of large-scale retired battery grading, and injects new momentum into facilitating more effective decision-making processes.

Chlorophyll content estimation and ripeness detection in tomato fruit based on NDVI from dual wavelength LiDAR point cloud data

Tomato (Solanum lycopersicum L.) is a climacteric fruit exhibiting the ripening pattern with change in colour from green to red. Normalized difference vegetation index (NDVI) derived from spectral-optical analysis has been previously employed in ripeness analysis of various fruits based on its strong correlation to ripening-related chlorophyll content. In this study, light detection and ranging (LiDAR) laser scanner was applied to gain 3D tomato point clouds aimed at estimating the spatially resolved chlorophyll content and ripeness class of tomato fruit. Freshly harvested tomatoes, capturing six ripeness classes (mature green, breaker, turning, pink, light red, and red) according to USDA colour standard, were analysed using a linear conveyor mounted sensor system consisting of two LiDAR units measuring position and return signal strength intensity at 660 nm and 905 nm. Fruit point clouds were pre-processed including geometric correction considering the curvature of each tomato sample, removing highest intensity areas at the specular highlighted spots, calibration of intensity values using standard black and white colour coated boards, and calibration on attenuation of the opaque samples. Particularly, for gaining the attenuation coefficient (μ), 9 opaque synthetic models with known attenuation were used. Obtained μ660 and μ905 were merged to gain NDVILiDAR. Chemically analysed chlorophyll content of tomato samples was correlated to μ660 and NDVILiDAR, whereas low correlation appeared for μ905 with a coefficient of determination (R2) of 0.58, 0.60, and 0.06, respectively. Regression models were used to estimate the total tomato chlorophyll content and the lowest root mean square error (RMSE) was found for μ660 (RMSE = 4.97 mg (100 g dry mass)−1) followed by NDVILiDAR (RMSE = 5.22 mg (100 g dry mass)−1), and μ905 (RMSE = 53.50 mg (100 g dry mass)−1). Histograms of NDVILiDAR and μ660 were extracted from each tomato point cloud and utilized to construct PLS-DA model for tomato ripening class prediction considering the spatial chlorophyll distribution. The overall accuracies for NDVILiDAR and μ660 were 70 % and 68 %, respectively, in leave-one-out cross-validation for visually defined colour classes. The LiDAR-based approach could support selective detection of ripe fruit in robots for harvesting and postharvest handling.

River stream flow prediction through advanced machine learning models for enhanced accuracy

The Narmada River basin, a significant water resource in central India, plays a crucial role in supporting agricultural, industrial, and domestic water supply. Effective management of this basin requires accurate streamflow forecasting, which has become increasingly important. This study delves into streamflow forecasting using historical data from five major river stations, covering the upper reaches in the East and middle sections. The dataset spans from 1978 to 2020 and undergoes rigorous screening and preparation, including normalization using StandardScaler. The research adopts a comprehensive approach, developing models for training on 70 % of historical data, validation on the most current 15 %, and testing against future targets with another 15 % of the data. To achieve precise predictions, a suite of machine learning models is employed, including CatBoost, LGBM (Light Gradient Boosting Machine), Random Forest, and XGBoost. Performance evaluation of these models relies on key indices such as mean squared error (MSE), mean absolute error (MAE), root mean square error (RMSE), root mean square percent error (RMSPE), normalized root mean squared error (NRMSE), and R-squared (R2). Notably, among these models, Random Forest emerges as the most robust for streamflow prediction, showcasing its effectiveness in handling the complexities of hydrological forecasting. This research contributes significantly to the field of streamflow forecasting in the Narmada River basin by providing insights into the performance of various machine learning models. The findings not only enhance our understanding of watershed dynamics but also highlight the pivotal role that machine learning can play in improving hydrological forecasting for sustainable watershed management.

One-dimensional model of manifold microchannels for embedded cooling: Prediction of thermal performance and flow non-uniformity

A one-dimensional model has been developed to accurately predict the thermal performance and flow non-uniformity of the manifold microchannels (MMC) for embedded liquid cooling. The model consists of one-dimensional governing equations derived from the integral relations of momentum and energy over appropriately-defined two separate control volumes. To validate the model, a series of 3-D numerical simulation is conducted over the wide ranges of the Reynolds number (Rem,in) at the manifold inlet from 560 to 3190, the dimensionless hydraulic flow length (x+) from 0.012 to 0.123, and the dimensionless thermal flow length (x∗) from 0.002 to 0.023. It is shown that the model provides accurate predictions of the thermal performance and flow non-uniformity (CV) of MMC heat sinks within the root mean square percentage error (RMSPE) of 6% and 26% for 50 data points, respectively. The significant improvement of the prediction accuracy is made over the earlier models with an error reduction of 82%. Finally, a design guideline for the uniform flow distribution is suggested for the first time based on a newly proposed explicit correlation for predicting the flow non-uniformity: the dynamic pressure at the manifold inlet should be kept smaller than the pressure drop across the microchannels.

Performance of IRI 2016 model in predicting total electron content (TEC) compared with GPS-TEC over East Africa during 2019–2021

This study evaluated the applicability of IRI-2016 model in predicting GPS TEC using the monthly means of the five (5) quiet days for equinoxes and solstices months. GPS-derived TEC data were obtained from the IGS network of ground based dual frequency GPS receivers from three stations [(KYN3 0.53° S, 38.53° E; Geom. Lat. 3.91.63° S), (MBAR 0.60° S, 30.74° E; Geom. Lat. 2.76° S) and HOID 1.45° S, 31.34° E; Geom. Lat. 3.71° S]. All the three options for topside Ne of IRI-2016 model and ABT-2009 for bottomside thickness have been used to compute the IRI TEC. The results were compared with the GPS TEC measurements. Correlation Coefficients between the two sets of data, the Root-Mean Square Errors of the IRI-TEC from the GPS-TEC, and the percentage RMSE of the IRI-TEC from the GPS-TEC have been computed. In general, the IRI-2016 model underestimated GPS-TEC during the nighttime, whereas the model overestimated GPS-TEC values during the daytime. At most of the stations and during all seasons where data were available, correlation coefficient was above 0.9, which is quite strong. The variation of O/N2 ratio may potentially be the cause of the IRI TEC deviation from the GPS TEC. This variation arises from lower thermosphere plasma drift that moves upward.

Generating Wall-to-Wall Canopy Height Information from Discrete Data Provided by Spaceborne LiDAR System

Provision of multi-temporal wall-to-wall canopy height information is one of the initiatives to combat deforestation and is necessary in strategizing forest conversion and reforestation initiatives. This study generated wall-to-wall canopy height information of the subtropical forest of Lishan, Taiwan, using discrete data provided by spaceborne LiDARs, wall-to-wall passive and active remote sensing imageries, topographic data, and machine learning (ML) regression models such as gradient boosting (GB), k-nearest neighbor (k-NN), and random forest (RF). ICESat-2- and GEDI-based canopy height data were used as training data, and medium-resolution passive satellite image (Sentinel-2) data, active remote sensing data such as synthetic aperture radar (SAR), and topographic data were used as regressors. The ALS-based canopy height was used to validate the models’ performance using root mean square error (RMSE) and percentage RMSE (PRMSE) as validation criteria. Notably, GB displayed the highest accuracy among the regression models, followed by k-NN and then RF. Using the GEDI-based canopy height as training data, the GB model can achieve optimum accuracy with an RMSE/PRMSE of 8.00 m/31.59%, k-NN can achieve an RMSE/PRMSE of as low as 8.05 m/31.78%, and RF can achieve optimum RMSE/PRMSE of 8.16 m/32.24%. If using ICESat-2 data, GB can have an optimum RMSE/PRMSE of 13.89 m/54.86%; k-NN can have an optimum RMSE/PRMSE of 14.32 m/56.56%, while RF can achieve an RMSE/PRMSE of 14.72 m/58.14%. Additionally, integrating Sentinel-1 with Sentinel-2 data improves the accuracy of canopy height modeling. Finally, the study underlined the crucial relevance of correct canopy height estimation for sustainable forest management, as well as the potential ramifications of poor-quality projections on a variety of biological and environmental factors.

Automated volumetry of meningiomas in contrast-enhanced T1-Weighted MRI using deep learning

Background: Meningiomas are among the most common intracranial tumors. In these tumors, volumetric assessment is not only important for planning therapeutic intervention but also for follow-up examination.However, a highly accurate automated volumetric method for meningiomas using single-modality magnetic resonance imaging (MRI) has not yet been reported. Here, we aimed to develop a deep learning-based automated volumetry method for meningiomas in MRI and investigate its accuracy and potential clinical applications. Methods: For deep learning, we used MRI images of patients with meningioma who were referred to Osaka University Hospital between January 2007 and October 2020. Imaging data of eligible patients were divided into three non-overlapping groups: training, validation, and testing. The model was trained and tested using the leave-oneout cross-validation method. Dice index (DI) and root mean squared percentage error (RMSPE) were measured to evaluate the model accuracy. Result: A total of 178 patients (64.6 ± 12.3 years [standard deviation]; 147 women) were evaluated. Comparison of the deep learning model and manual segmentation revealed a mean DI of 0.923 ± 0.051 for tumor lesions. For total tumor volume, RMSPE was 9.5 ± 1.2%, and Mann–Whitney U test did not show a significant difference between manual and algorithm-based measurement of the tumor volume (p = 0.96). Conclusion: The automatic tumor volumetry algorithm developed in this study provides a potential volume-based imaging biomarker for tumor evaluation in the field of neuroradiological imaging, which will contribute to the optimization and personalization of treatment for central nervous system tumors in the near future

Bayesian optimization algorithm‐based Gaussian process regression for in situ state of health prediction of minorly deformed lithium‐ion battery

AbstractAccurate on‐board state‐of‐health (SOH) prediction is crucial for lithium‐ion battery applications. This study presents an in situ prediction technique for minorly deformed battery SOH, utilizing a Gaussian process regression (GPR) model tuned by a Bayesian optimization algorithm. Unlike previous methods that interpret voltage–time data as incremental capacitance curves, our approach directly operates on raw voltage–time data. We apply gray relational analysis to select feature variables as inputs and train the Bayesian Gaussian process regression (BGPR) model using experimental data from batteries under different working conditions. To demonstrate the performance of the BGPR model, we compare it with stepwise linear regression, neural network, and Bayesian support vector machine (BSVM) models. The performance of these four models is evaluated using different performance indicators: mean absolute percentage error (MAPE), root‐mean‐squared percentage error (RMSPE), and coefficient of determination (R²). The results demonstrate that the BGPR model exhibits superior prediction performance with the lowest MAPE (0.11%), RMSPE (0.12%), and the highest R² (0.9915) for minorly deformed batteries. Furthermore, the BGPR model exhibits excellent robustness for SOH prediction of normal batteries under different conditions. This study provides an effective and robust method for accurate on‐board SOH prediction in lithium‐ion battery applications.

How Not to Make the Joint Extended Kalman Filter Fail with Unstructured Mechanistic Models.

The unstructured mechanistic model (UMM) allows for modeling the macro-scale of a phenomenon without known mechanisms. This is extremely useful in biomanufacturing because using the UMM for the joint estimation of states and parameters with an extended Kalman filter (JEKF) can enable the real-time monitoring of bioprocesses with unknown mechanisms. However, the UMM commonly used in biomanufacturing contains ordinary differential equations (ODEs) with unshared parameters, weak variables, and weak terms. When such a UMM is coupled with an initial state error covariance matrix P(t=0) and a process error covariance matrix Q with uncorrelated elements, along with just one measured state variable, the joint extended Kalman filter (JEKF) fails to estimate the unshared parameters and state simultaneously. This is because the Kalman gain corresponding to the unshared parameter remains constant and equal to zero. In this work, we formally describe this failure case, present the proof of JEKF failure, and propose an approach called SANTO to side-step this failure case. The SANTO approach consists of adding a quantity to the state error covariance between the measured state variable and unshared parameter in the initial P(t = 0) of the matrix Ricatti differential equation to compute the predicted error covariance matrix of the state and prevent the Kalman gain from being zero. Our empirical evaluations using synthetic and real datasets reveal significant improvements: SANTO achieved a reduction in root-mean-square percentage error (RMSPE) of up to approximately 17% compared to the classical JEKF, indicating a substantial enhancement in estimation accuracy.

An aviation accident report data-driven approach to scenario design for a centrifuge-based dynamic flight simulator

The use of flight simulators in investigating an aviation incident or accident related to human errors has been identified as an important part of a strategy to improve safety. This study aimed to replicate a real flight of the MiG-29 aircraft using a centrifuge-based dynamic flight simulator and to determine the simulator’s accuracy in recreating in-flight aircraft performance. A 60-second recording of the real flight of the MiG-29 aircraft, captured by the flight data recorder, was chosen for replication in the HTC-07 human training centrifuge simulator. To evaluate how accurately the simulator replicates the performance of the aircraft, the linear accelerations and angular velocities acting on a pilot during the real flight were compared with those during the replication of that flight in the simulator. The fit of these parameters was assessed using the root mean square percentage error (RMSPE) and the correlation coefficient (r). The highest replication accuracy was achieved for the vertical component of the linear acceleration (RMSPE=2068; r=0.98), while the worst result was obtained for the longitudinal component (RMSPE=14205; r=0.31). Inaccuracies were much more pronounced for the angular velocity. The roll angular velocity had the lowest replication error (RMSPE=12640). However, its correlation with the recorded velocity during the real flight was very weak (r=-0.02). Despite some inaccuracies in replicating other components of the acceleration and angular velocity vectors, the HTC-07 simulator seems valuable for investigating aviation incidents or accidents related to human factors.

Assessing the carbon footprint in dairy cattle farms in the northern temperate region of Spain

The availability of models constitutes a key factor when selecting a decision support tool aimed at improving the production and environmental aspects of farms. There is a need for robust models that are user-friendly, facilitating the estimation of farm emissions and the analysis of their temporal fluctuations. The objectives of this study were i) to calculate both the partial (PCF) and total carbon (TCF) footprints of 212 dairy farms, distinguishing those with and without maize cultivation; ii) to identify critical variables related to feed, nutrition, productivity and environmental efficiency; and iii) to formulate and validate prediction equations based on available data from dairy farms. The database encompasses information from 212 dairy cattle farms situated in the temperate-humid zone of northern Spain, spanning the period from 2014 to 2018. Farm classification was based on the presence (CcMz) or absence (ScCMz) of maize cultivation for silage production, resulting in 96 farms in the CcMz category and 116 farms in the ScCMz category.Among the variables considered, the variable herd N-use efficiency (NUECR) for (PCF) showed the lowest root mean square error of prediction at 0.39% and the correspondingly lowest root men. The root mean squared percentage error (RMSPE): standard deviation ratio (RSR) at 0.52. In the case of total carbon footprint (TCF), herd N-use efficiency (NUECR) again showed the lowest root mean square error of prediction at 0.52%. Regarding TCF, herd feed efficiency (EACR) was the variable with the lowest both RMSPE and RSR, with 0.65 and 0.64, respectively. Consequently, the estimation of the PCF and TCF of 1 ​kg of milk from the temperate-humid zone of northern Spain at the farm gate can be feasibly accomplished utilizing NUECR and EACR, respectively.

Enhancement Methods for Energy Consumption Prediction in Smart House based on Machine Learning

Energy efficiency in modern homes has recently become a significant issue due to the emergence of smart home infrastructure. Numerous public structures, such as homes, hospitals, schools, and other institutions, use more energy. To come close to meeting the actual energy demand, it is crucial that we create as much energy as we can. Machine learning has various advantages for improving the effectiveness and efficiency of smart home systems and appliances, including managing and lowering energy use. Additionally, as a key component of the smart home idea, we explore the potential integration of machine learning-based on some algorithm methodologies ways to improve power energy consumption system and control. The models were used to identify patterns for smart home and variations in energy consumption. This study's conclusions were used to analyze case studies and forecast energy consumption. Detection Change (of used and generation) for all appliances, which excessive foresees energy use and stops a rise in usage. Predict Future Energy use by using meteorological data and maximizing the supply of energy to forecast future energy generation and use. Finally, using five machine learning algorithms, including the Linear Regression (LR), Gradient Boosting Regression (GBoostR), Decision Tree Regression (DTR), Stochastic Gradient Descent Regression (SGDR), and Bayesian Ridge Regression (BRR), we can measure the Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Absolute Error (RMAE), and Root Mean Squared Percentage Error (RMSPE), in order to determine how well models.

Data-driven capacity estimation for lithium-ion batteries with feature matching based transfer learning method

Accurate capacity estimation is essential in the management of lithium-ion batteries, as it guarantees the safety and dependability of battery-powered systems. However, direct measurement of battery capacity is challenging due to the unpredictable working conditions and intricate electrochemical characteristics, which complicates the identification of battery degradation. In this work, through in-depth analysis of battery aging data, an incremental slope (IS) aided feature extraction method is proposed to obtain universal multidimensional features that adapt to different working conditions. With the extracted features, a simple multilayer perceptron (MLP) is used to achieve high-precision capacity estimation. Furthermore, a feature matching based transfer learning (FM-TL) method is proposed to automatically adapt the capacity estimation across different types of batteries that are cycled under various working conditions. 158 batteries covering five material types and 15 working conditions are used to validate the proposed method. Results suggest that the MLP model can provide an accurate capacity estimation, where the overall mean absolute percentage error (MAPE) and root mean square percentage error (RMSPE) are limited to 1.22% and 1.61%, respectively. Furthermore, compared with the traditional fine-tuning method, the overall MAPE and RMSPE under various transfer learning application scenarios respectively decrease by up to 78.23% and 75.31%, indicating that the FM-TL method is promising to construct a reliable transfer learning path, which improves the accuracy and reliability of capacity estimation when applied to various target domains.

출생 성비와 명절 시기를 고려한 월별 한우 도축두수 예측모형 개발

The number of Korean native cattle, Hanwoo, has exceeded 3.5 million, the historical largest one, in May 2022. As a result, the Hanwoo industry are in a state of oversupply and are concerned about the risk of a price collapse. The main objective of this study is to develop medium- and long-term forecasting models for monthly Hanwoo supply as a way to reduce the risk of imbalance between supply and demand. To construct the forecasting models, the empirical histogram of the month-specific rate for Hanwoo is applied. The monthly data of Hanwoo in 2014-2022 are obtained from the Animal Product Traceability System in Korea. The results are as follows: A forecasting model for the number of monthly Hanwoo at 6-month-old with considering the birth sex ratio outperforms the other models. Also, adjusting the slaughter rate based on holiday season made the forecast more accurate. The forecasting models were evaluated based on multiple accuracy measures such as mean absolute percentage error(MAPE) and root mean square percentage error (RMSPE). The MAPE for the monthly slaughtered Hanwoo females and males were 9.37% and 8.98%, respectively. Therefore, the forecasting model developed in this study is expected to be useful as a tool to reduce the risk faced by those in the Hanwoo industry.

DeepResTrade: a peer-to-peer LSTM-decision tree-based price prediction and blockchain-enhanced trading system for renewable energy decentralized markets

Intelligent predictive models are fundamental in peer-to-peer (P2P) energy trading as they properly estimate supply and demand variations and optimize energy distribution, and the other featured values, for participants in decentralized energy marketplaces. Consequently, DeepResTrade is a research work that presents an advanced model for predicting prices in a given traditional energy market. This model includes numerous fundamental components, including the concept of P2P trading systems, long-term and short-term memory (LSTM) networks, decision trees (DT), and Blockchain. DeepResTrade utilized a dataset with 70,084 data points, which included maximum/minimum capacities, as well as renewable generation, and price utilized of the communities. The developed model obtains a significant predictive performance of 0.000636% Mean Absolute Percentage Error (MAPE) and 0.000975% Root Mean Square Percentage Error (RMSPE). DeepResTrade’s performance is demonstrated by its RMSE of 0.016079 and MAE of 0.009125, indicating its capacity to reduce the difference between anticipated and actual prices. The model performs admirably in describing actual price variations in, as shown by a considerable R2 score of 0.999998. Furthermore, F1/recall scores of [1, 1, 1] with a precision of 1, all imply its accuracy.

A comprehensive approach to simulation of cartridge filtration using CFD

Household water treatment (HWT) systems are widely used for the provision of potable water in many countries with their low-cost key to attaining universal and equitable access to safe and affordable drinking water, Sustainable Development Goal 6.1. Removal of suspended particles (turbidity) from water sources via cartridge filters is often the first step of a HWT system, with the primary treatment increasing the efficiency of a subsequent disinfection step.Whilst the performance of cartridge filters (removal efficiency and pressure drop) can be determined experimentally in long experiments with high volumes of water, numerical simulation adds fundamental insight to the influence of fluid dynamics on particle deposition and vice versa. In this study, a novel computational fluid dynamic (CFD) model was developed to simulate the fundamental mechanisms underpinning the removal of particles within the widely used 10 in. cartridge filter, informed by and complemented with laboratory validation. The Eulerian approach was used to simulate fluid flow with the Lagrangian approach adopted for particle tracking. Rosin-Rammler distribution was implemented with respect to the particle size distribution of the diatomaceous earth particles used in the experiments. Given particles were non-spherical (disk shape), Wadell’s sphericity was included to account for the effect of particle shape on drag force. A porous domain was implemented to simulate the filter element through addition of a source term to the momentum equations, with the likelihood of particle deposition, detachment and rebound also considered.Laboratory based validation studies confirmed the novel CFD model to accurately model removal of turbidity and predict the pressure drop across the filter with Root Mean Square Percentage Error (RMSPE) of less than 3%. The simulated location of particle deposition on the filter elements closely matched images taken at several stages during filtration experiments with the model aiding understanding of pattern of particle removal along and within the porous filter structure. This novel and comprehensive modelling methodology can be utilized to simulate the filtration process at the macro-scale, permitting evaluation of new filter designs and materials for advanced filtration systems; ultimately improving HWT system performance and reducing costs to users.

Robust tool condition monitoring in Ti6Al4V milling based on specific force coefficients and growing self-organizing maps

Tool condition monitoring (TCM) is a mean to optimize production systems trying to use cutting tool life at its best. Nevertheless, nowadays available TCM algorithms typically lack robustness in order to be consistently applied in industrial scenarios. In this paper, an unsupervised artificial intelligence technique, based on Growing Self-Organizing Maps (GSOM), is presented in synergy with real-time specific force coefficients (SFC) estimation through the regression of instantaneous cutting forces. The conceived approach allows robustly mapping the SFC, exploiting process parameters and similarity to manage the variability of their estimation due to unmodelled phenomena, like machine dynamics and tool run-out. The devised approach allowed detecting the tool end-of-life in cutting tests with variable lubrication, machine tool and cutting speed, through the adoption of a self-starting control chart running on real-time clustered data. The solution was validated through the comparison of the GSOM framework with respect to the optimized self-starting control chart applied without GSOM clustering. The GSOM reached a root mean squared percentage error (RMSPE) of 13.2% with respect to 56.1% obtained with the analogous control chart in a full-set optimization scenario. When optimised on tests for a unique machine tool and tested on another machine tool, GSOM scored an RMSPE of 34.5%, whereas the optimized control chart scored 64.5%.

State of health estimation for lithium-ion batteries using Gaussian process regression-based data reconstruction method during random charging process

State of health (SOH) estimation is a critical technology to guarantee the safe and reliable operation of battery energy systems. Data-driven methods have been widely studied in the field of lithium-ion battery SOH estimation. However, random charging in real operating scenarios will result in difficult extraction of health features, which in turn limits the online application of data-driven methods. In this paper, a SOH estimation method using a piece of random charging data is proposed. In random charging scenarios, the health features screened within the predefined charging range are easily lost. For this reason, the partial charging data is reconstructed by the Gaussian Process Regression (GPR) model to intelligently supplement the health features that are not sampled. In terms of health feature selection, this paper selects the charging time sequence within a voltage interval as the input of the Long-Short-Term Memory (LSTM) estimation model, so that the complicated feature extraction work can be avoided, such as the peak of incremental capacity (IC) curve, the slope of the charging voltage and so on. The experimental results show that the data reconstruction method based on the GPR model can reconstruct the missing charging data accurately under a large number of simulated random charging data, and the reconstruction error of the charging data under two types of aging paths is less than 4.30%. Finally, using most known health features and a few reconstructed health features within certain charging intervals, the proposed GPR-LSTM method achieves a root-mean-squared percentage error (RMSPE) of 2.51% in SOH estimation for the two well-known laboratory datasets, and compared to 3.89% for the IC peak feature-based method, a reduction by 1.47%. Meanwhile, for a real-world battery data, the RMSPE of the proposed method is below 3%, illustrating the usefulness of the proposed method under the on-road conditions.