Box-Cox Transformation Research Articles

Methods for estimating insulin resistance from untargeted metabolomics data

ContextInsulin resistance is associated with multiple complex diseases; however, precise measures of insulin resistance are invasive, expensive, and time-consuming.ObjectiveDevelop estimation models for measures of insulin resistance, including insulin sensitivity index (SI) and homeostatic model assessment of insulin resistance (HOMA-IR) from metabolomics data.DesignInsulin Resistance Atherosclerosis Family Study (IRASFS).SettingCommunity based.ParticipantsMexican Americans (MA) and African Americans (AA).Main outcomeEstimation models for measures of insulin resistance, i.e. SI and HOMA-IR.ResultsLeast Absolute Shrinkage and Selection Operator (LASSO) and Elastic Net regression were used to build insulin resistance estimation models from 1274 metabolites combined with clinical data, e.g. age, sex, body mass index (BMI). Metabolite data were transformed using three approaches, i.e. inverse normal transformation, standardization, and Box Cox transformation. The analysis was performed in one MA recruitment site (San Luis Valley, Colorado (SLV); N = 450) and tested in another MA recruitment site (San Antonio, Texas (SA); N = 473). In addition, the two MA recruitment sites were combined and estimation models tested in the AA recruitment sample (Los Angeles, California; N = 495). Estimated and empiric SI were correlated in the SA (r2 = 0.77) and AA (r2 = 0.74) testing datasets. Further, estimated and empiric SI were consistently associated with BMI, low-density lipoprotein cholesterol (LDL), and triglycerides. We applied similar approaches to estimate HOMA-IR with similar results.ConclusionsWe have developed a method for estimating insulin resistance with metabolomics data that has the potential for application to a wide range of biomedical studies and conditions.

BPNet: A multi-modal fusion neural network for blood pressure estimation using ECG and PPG

This paper aims to address the problem of estimating blood pressure (BP) based on specific physiological signals. However, several issues exist in current works. Firstly, some methods produce biased models by training them on a skewed distribution and failing to consider the assumptions underlying regression analysis. Secondly, data leakage can occur when overlapping samples are present in the dataset. Finally, some methods concatenate Electrocardiogram (ECG) and Photoplethysmogram (PPG) signals without fully capturing their relationship. To overcome these limitations, this paper applies the BoxCox transformation to correct for skewed label distributions and creates a non-overlapping dataset. Additionally, a novel end-to-end model, the blood pressure network (BPNet), is proposed, which can accurately estimate blood pressure. The Multi-parameter Intelligent Monitoring in Intensive Care database (MIMIC) is utilized to develop and verify the model, which includes ECG, PPG, and invasive BP data from patients in intensive care units (ICUs). The BPNet is evaluated on the MIMIC II (942 subjects) and MIMIC-III (833 subjects) datasets. In MIMIC-II, the estimation error for systolic blood pressure (SBP) and diastolic blood pressure (DBP) is −0.17 ± 4.62 mmHg and −0.24 ± 2.95 mmHg, respectively. In MIMIC-III, the estimation error for SBP and DBP is −0.30 ± 5.78 mmHg and −0.25 ± 3.80 mmHg, respectively. The experimental results exceed the performance of existing methods and meet the accuracy standards established by the Association for the Advancement of Medical Instrumentation (AAMI) and the British Hypertension Society (BHS). This demonstrates the effectiveness and accuracy of the proposed model, highlighting the potential for practical applications in clinical settings.

Prediction of low Zn concentrations in soil from mountainous areas of central Yunnan Province using a combination of continuous wavelet transform and Boruta algorithm

ABSTRACT Accurately predicting heavy metal concentration in soils is essential for agricultural production and food safety. The hyperspectral technique provides a feasible method for rapidly determining heavy metal concentration. In this study, we use spectral and Box-Cox transformations to pre-process the spectral and soil Zn concentration data to solve the problem of low prediction accuracy caused by the non-normal distribution of soil Zn concentration and the unobvious spectral characteristics. The characteristic wavelengths were selected using competitive adaptive reweighted sampling (CARS) and Boruta algorithms. Back propagation neural network (BPNN) was used to predict the low Zn concentration in the soil. The results showed that the accuracy of the models was mostly improved after Box-Cox transformation for low Zn concentration; both continuous wavelet transform (CWT) and fractional order differential (FOD) could explore the delicate features of the spectra to different degrees; the Boruta algorithm could select the wavelengths that were more relevant to Zn; the best performance could be obtained by combining CWT and Boruta algorithms, where the L5-Boruta model had the highest prediction accuracy with R2, RMSE and RPIQ of 0.656, 13.050 mg∙kg−1 and 2.351, respectively.

The distribution of the Anderson Darling statistic

The Anderson-Darling test for normality is attractive for heavy-tailed alternatives, and is easy to compute. Its null distribution is not known, and p values are currently computed using a simulation-based approximation with 17 parameters. An alternative using the Box-Cox transformation gives an approximation which uses only 5 parameters and is also more accurate, particularly in the relevant tail area.

Data-driven boundary regression for equivalent identification of external parameters in steady-state models of power systems

External Network Modeling (ENM) is an important method to simplify and accelerate power system control, protection, and optimization. Traditional model-based methods obtain accurate external parameters by analyzing the physical properties of the whole system. In the case of incomplete physical information, the use of data-driven models has become an effective means to analyze the operating state of power systems. In this paper, a data-driven approach is used to propose an external equivalent parameter identification framework for steady-state models of power systems to accurately estimate the operating states of power systems of interest. Linear correction coefficients are used to deal with the challenges brought by the nonconvexities of the power flow equation to the regression. Based on the linearly mapped power flow equation, we derive the Multivariate Linear Regression (MLR) model of the external power system. The Box-Cox transformation (BCT) and ridge regression algorithm (RRA) algorithms are designed to address data normality, collinearity, and overfitting. The proposed framework is tested on a series of IEEE standard cases, which include both meshed transmission grids and radial distribution grids, with both Monte Carlo simulated data and public testing data. The results show that the proposed method can obtain a higher calculation accuracy than model-based methods can. The results also demonstrate that the proposed data-driven equivalent framework can obtain accurate regression parameter matrices under the condition of non-normal data, and accurately reflect the physical parameters of the external power system.

Surface Water Quality Assessment through Remote Sensing Based on the Box–Cox Transformation and Linear Regression

A methodology to estimate surface water quality using remote sensing is presented based on Landsat satellite imagery and in situ measurements taken every six months at four separate sampling locations in a tropical reservoir from 2015 to 2019. The remote sensing methodology uses the Box–Cox transformation model to normalize data on three water quality parameters: total organic carbon (TOC), total dissolved solids (TDS), and chlorophyll a (Chl-a). After the Box–Cox transformation, a mathematical model was generated for every parameter using multiple linear regression to correlate normalized data and spectral reflectance from Landsat 8 imagery. Then, significant testing was conducted to discard spectral bands that did not show a statistically significant response (α = 0.05) from the different water quality models. The r2 values achieved for TOC, TDS, and Chl-a water quality models after the band discrimination process were found 0.926, 0.875, and 0.810, respectively, achieving a fair fitting to real water quality data measurements. Finally, a comparison between estimated and measured water quality values not previously used for model development was carried out to validate these models. In this validation process, a good fit of 98% and 93% was obtained for TDS and TOC, respectively, whereas an acceptable fit of 81% was obtained for Chl-a. This study proposes an interesting alternative for ordered and standardized steps applied to generate mathematical models for the estimation of TOC, TDS, and Chl-a based on water quality parameters measured in the field and using satellite images.

Optimization of wind off-grid system for remote area: Egyptian application

Due unexpected nature of renewable energy systems, the (Wind/Diesel/Battery) (W/D/B) off-grid system has initially been investigated at a South Sinai location in Egypt for home-scale consumption. Eight different systems, each of which consists of a small wind turbine, storage batteries, and diesel generator, are investigated in accordance with the varying needs of the power loads and seasonal weather data. The major goal is to investigate how adding wind power as an energy source will affect the price of electricity generated while taking into account the cost of reducing CO2 emissions as an external benefit of the wind turbine, which emits no pollutants during operation. In order to compare a Taguchi OA design to a two-level full factorial design to evaluate the systems at two separate sites (South Sinai and the Western Desert in Egypt), a design evaluation tool in DOE++ will be used. To pinpoint the crucial variables and analyze the impact of six different factors on eight different sets, Taguchi OA is used. The proportion of power shortfall is a production indicator, while the net present cost (NPC) and cost of energy (COE) are used as economic indicators. The simulation results demonstrate that W/D/B systems are economically viable for the hypothetical community site when using HOMER software, with electricity generated at a cost of about 0.285$/kWh without accounting for external benefits and 0.221$/kWh if CO2 emissions are competitive with diesel-only systems, where COE is 0.432$/kWh. As a new evaluation approach, the Box-Cox transformation calculated the best λ is about −2 at the two locations, indicating similar technique behaviors, and the fitted probability shows, meaning that the significant impact of system components are wind turbines. Regression model of CO2 emission is demonstrated to be successful for estimates at the Western Desert location than the South Sinai region

Early Prediction of Monkeypox Virus Outbreak Using Machine Learning

At the onset of an infectious disease, such as the monkeypox virus (MPXV), surveillance data is crucial in keeping track of the outbreak’s progression. The surveillance data for MPXV received considerable attention after multiple European countries recorded cases. Historical data obtained from May 9, 2022, to August 10, 2022, were used to model the cumulative case trajectories of MPXV in five countries. Our study employed autoregressive integrated moving averages (ARIMA), neural network autoregression (NNETAR), exponential smoothing (ETS), and seasonal naïve regression (SNAÏVE) for training and evaluation. The paper makes the following contributions: (1) enhanced model stability with the Box-Cox transformation as a preprocessing step, (2) experimentation with both linear and non-linear models, and (3) simulation of the top five countries during the impulsive rise in cases of MPXV. The results were evaluated using three metrics: root mean square error (RMSE), mean square error (MAE), and mean absolute percentage error (MAPE). The ARIMA (0,1,3) (1,0,0)[7] model yielded the lowest percentage error of 5.16 in the holdout set for MAPE in France observations. The ETS (A, A, A) model, the lowest percentage error in the holdout set for MAE was 7.35 in Germany. Regarding the NNETAR (1,1,2) [7] model, the lowest percentage error in the holdout observations for RMSE was 8.33 in Spain, 2.75 in the United Kingdom (UK), and 8.05 in the United States of America (USA) in that order. Based on these findings, we can conclude that while the transformation proved crucial for model performance, it was not necessary for all experiments, as ARIMA remained dominant in France and the ETS model in Germany. At the same time, NNETAR model outperformed in cumulative case counts in Spain, the UK, and the USA. Our experimentation allows for early identification and contributes to a better understanding of forecasting MPXV cases using combinations of both linear and nonlinear models.

Bayesian model updating utilizing scaled likelihood ratio and BCT-PCA with frequency response function

Bayesian inference has now been widely practiced for model updating. In this study, the discussion is conducted about two issues facing the implementation of Bayesian model updating when the Gaussian distribution is assumed. To begin with, the problem of the normal likelihood function used in Markov chain Monte Carlo (MCMC) is demonstrated and presented with an alternative approach by defining the scaled likelihood ratio (SLR). Then, a data preprocessing technique is proposed by introducing the Box-Cox transformation and principal component analysis (BCT-PCA) to address the obstacles in the practice of frequency response function (FRF)-based Bayesian model updating. The differential evolution adaptive metropolis (DREAM) is applied to explore the posterior distribution of the parameters. In addition, a novel procedure for FRF-based Bayesian model updating is presented. The effectiveness of the proposed methodology is verified in both numerical and experimental cases of dynamical model updating. In the numerical case, the results of eigenfrequency-based model updating show that the SLR is advantageous over the normal likelihood ratio (NLR) in estimating the posterior distribution of parameters. The results of FRF-based model updating show that the consistency and robustness of model updating performance as achieved through the BCT-PCA method are improved when different schemes of frequency selection are implemented. The results of the experimental case also show the superiority of the proposed method.

365-OR: A Machine-Learning Approach to Estimate Insulin Resistance from Untargeted Metabolomics Data

Insulin resistance (IR) is constantly invoked as a contributor to numerous common diseases, e.g. diabetes, cardiovascular disease and Alzheimer’s disease; however, with the exception of a narrow segment of diabetes-related studies, IR is infrequently measured and rarely measured well. Our goal was to develop IR estimation models for the dynamic measure of insulin sensitivity (SI) and the basal measure of homeostatic model assessment of IR (HOMAIR) from metabolomics data combined with minimal clinical data, e.g. age, sex, BMI. Least Absolute Shrinkage and Selection Operator (LASSO) and Elastic Net regression were used to build estimation models using 1274 metabolites, prioritized by trait associations and collected in the Insulin Resistance Atherosclerosis Family Study (IRASFS). Three metabolite transformations were implemented (inverse normal transformation, standardization, and Box Cox transformation) to account for distributional differences and tested to optimize estimation. The analysis was performed in one Mexican American (MA) recruitment site (San Luis Valley, CO (SLV), N=450) and tested in another MA recruitment site (San Antonio, TX (SA), N=473). In addition, the two MA recruitment sites were combined and estimation models tested in the African American recruitment site (AA; Los Angeles, California, N=495) to assess generalizability. Estimated SI was correlated with empiric SI in the SA (r2=0.77) and AA (r2=0.74) testing datasets. Further, we tested estimated SI for association with BMI, low-density lipoprotein cholesterol and triglycerides with results consistent with those observed for empiric SI. The same framework was used to estimate HOMAIR, and yielded similar results. In summary, we have developed methods for estimating multiple measures of IR from a single blood sample. This approach will greatly expand our ability to provide meaningful insight into the role of IR in disease, supporting the potential for application to a wide-range of biomedical studies. Disclosure N.Allred: None. D.W.Bowden: None. F.Hsu: None. S.Chen: None. M.Ng: None. M.O.Goodarzi: Advisory Panel; Nestlé Health Science, Other Relationship; Nestlé Health Science. J.I.Rotter: None. L.E.Wagenknecht: None. M.Bancks: None. R.N.Bergman: Consultant; Lilly, ReCor Medical, Inc., Research Support; AstraZeneca. Funding National Institutes of Health (DK118062)

A novel electricity load forecasting based on probabilistic least absolute shrinkage and selection operator-Quantile regression neural network

Predicting energy consumption is challenging since forecast accuracy is heavily impacted by several external elements, including the economy, community, environment, and sustainable energy, including wind, solar, and hydro energy. The most crucial aspect in successfully predicting electrical power demand is to gather crucial information about those external factors in the vast amount of data provided by the smart grid. The Least Absolute Shrinkage and Selection Operator-Quantile Regression Neural Network (LASSO-QRNN) is a strategy for probability density prediction proposed in this article. First, using LASSO regression, the primary features are determined from external factors impacting electrical power utilization. The LASSO-Quantile regression NN approach is then developed to forecast annual electrical power consumption. In the years that followed, various quantiles were employed to gauge electrical power demand, and the outcomes were assessed. The LASSO-QRNN approach also includes a kernel density evaluation that, in the event of a one-value forecast, will display a probability distribution. The experimental assessments from the Guangdong database in China and the California real-world databases in the US have evaluated the forecasting precision. The results of the numerical study demonstrate that, in comparison to Quantile Regression and Kernel Density, Box-Cox transformation, the provided model provides superior accuracy for the prediction of electrical power demand. Deep Generative Architecture, K-means Clustering, a Quantile Regression Neural Network, expert prediction, and fuzzy Bayesian theory. LASSO-Quantile regression neural network can handle the higher dimension information in electrical power usage prediction besides the more accurate outcomes.

On the Use of the Power Transformation Models to Improve the Temperature Time Series

The aim of this paper is to select an appropriate ARIMA model for the time series after transforming the original responses. Box-Cox and Yeo-Johnson power transformation models were used on the response variables of two time series datasets of average temperatures and then diagnosed and built the appropriate ARIMA models for each time-series. The authors treat the results of the model fitting as a package in an attempt to decide and choose the best model by diagnosing the effect of the data transformation on the response normality, significant of estimated model parameters, forecastability and the behavior of the residuals. The authors conclude that the Yeo-Johnson model was more flexible in smoothing the data and contributedto accessing a simple model with good forecastability.

Surface Roughness Modeling of Hard Turning 080A67 Steel

Surface roughness is an important parameter to evaluate the quality of a machining process in mechanical manufacturing. The construction of a surface roughness model of a machining process is the basis for predicting surface roughness corresponding to each certain case. This paper presents the construction of a surface roughness model in 080A67 steel turning. An experimental process was carried out with a total of 15 experiments, designed according to the Box-Behnken matrix. The cutting speed, feed rate, and cutting depth were changed in each experiment, and surface roughness values were measured to build a model that showed the mathematical relationship between surface roughness and the three cutting parameters. A second surface roughness model was also constructed using the Box-Cox transformation. The accuracy of these two models was compared through five coefficients: R2, R2(pred), R2(adj), Percentage Absolute Error (PAE), and Percentage Square Error (PSE). The results showed that all these coefficients of the model using the Box-Cox transformation were better than those of the first one. In detail, the values of R2, R2(pred), R2(Adj), PAE, and PSE of the first model were 94.55%, 12.79%, 84.74%, 8.79%, and 1.42%, while for the second model were 99.09%, 85.42%, 97.44%, 2.26%, and 0.18%, respectively, showing that the accuracy of the surface roughness model was improved by using the Box-Cox transformation.

Comparison of three indirect methods for verification and validation of reference intervals at eight medical laboratories: a European multicenter study

Abstract Objectives Indirect methods for the indirect estimation of reference intervals are increasingly being used, especially for validation of reference intervals, as they can be applied to routine patient data. In this study, we compare three statistically different indirect methods for the verification and validation of reference intervals in eight laboratories distributed throughout Europe. Methods The RefLim method is a fast and simple approach which calculates the reference intervals by extrapolating the theoretical 95 % of non-pathological values from the central linear part of a quantile-quantile plot. The Truncated Maximum Likelihood (TML) method estimates a smoothed kernel density function for the distribution of the mixed data, for which it is assumed that the ‘‘central’’ part of the distribution represents the healthy population. The refineR utilizes an inverse modelling approach. This algorithm identifies a model that best explains the observed data before transforming the data with the Box-Cox transformation. Results We show that the different indirect methods each have their advantages but can also lead to inaccurate or ambiguous results depending on the approximation of the mathematical model to real-world data. A combination of different methodologies can improve the informative value and thus the reliability of results. Conclusions Based on routine measurements of four enzymes alkaline phosphatase (ALP), total amylase (AMY), cholinesterase (CHE) and gamma-glutamyl transferase (GGT) in adult women and men, we demonstrate that some reference limits taken from the literature need to be adapted to the laboratory’s particular local and population characteristics.

Temperature Forecasting as a Means of Mitigating Climate Change and Its Effects: A Case Study of Mali

Temperature forecasts and trend analyzes were carried out for several locations in Mali as an important tool for warning of potentially threatening weather events such as severe heat waves, storms, droughts and floods, which could pose a great risk to humans and their environment. Five locations (Segou, Sikasso, Kayes, Gao and Taoudenni) across Mali (170 00’N – 40 00’W) were chosen for this research work. Satellite data of annual temperature obtained from the European Centre for Medium-Range Weather Forecast (ECMWF) database for 35 years (1985-2019) was used for this work. The Mann-Kendall trend test was carried out for various locations to observe and study the trend. Four Models including Auto Regressive and Integrated Moving Average (ARIMA), Exponential smoothening (ETS), TBATS (Trigonometric seasonality, Box-Cox transformation, ARMA errors, Trend and Seasonal components) and the linear model were employed to forecast average temperature for 10 years for all the locations. The model that produces the best forecast at the 95% confidence level is expected to have the lowest Root Mean Square Error (RMSE) value. The results showed that no significant trends were recorded at the considered locations. The linear model produced the best forecast for Segou, Kayes and Taoudenni, while the TBATS model produced the best forecast for Gao and the ARIMA model produced the best forecast for Sikasso.

Smoking is a Factor in Discordance Between QuantiFERONTB Gold Assay and Tuberculosis Etiology: Especially in Older Patients.

Exploring whether smoking is an influencing factor for the inconsistency between QuantiFERONTB Gold assay (QFT-GIT) and tuberculosis etiology. The clinical data of patients who were confirmed positive for Mycobacterium tuberculosis (MTB) after undergoing QFT-GIT testing from September 2017 to August 2021 were retrospectively analyzed. Chi-square and rank-sum tests were used to compare the differences in characteristics between smokers and non-smokers. Logistic regression was used to adjust for confounding factors affecting smoking. Propensity score matching (PSM) was used to verify the above conclusions again. Positive results of tuberculosis etiology were adopted as the standard, the incidence of inconsistent results between QFT-GIT and tuberculosis etiology was 8.90% (108/1213), of which the false negative rate was 6.27% (76/1213) and the indeterminate rate was 2.64% (32/1213). In the overall population, the smokers had a lower level of basal IFN-γ (Z=-2.079, P=0.038). Among 382 elderly (≥65 years old) patients, the smokers had lower levels of antigen-stimulated IFN-γ (Z=-2.838, P=0.005). After performing BOX-COX transformation on all non-normally distributed data, logistic stepwise regression was used to adjust confounding factors. The results showed that smoking was an influencing factor for the inconsistency between QFT-GIT and tuberculosis etiology results (OR=1.69, P=0.020). Using PSM for 1:2 matching, the results showed that smoking was still an independent risk factor for the inconsistent results of QFT-GIT and tuberculosis etiology (OR= 1.95, P=0.019). Age-stratified analysis showed that smoking was an independent risk factor in discordance between QFT-GIT and tuberculosis etiology in patients aged ≥65 years (OR=2.40, P=0.005), but not in patients aged <65 years (P > 0.05). Smoking can reduce the body's IFN-γ release ability, and smoking (especially the elderly) is an influencing factor for the inconsistency between QFT-GIT and tuberculosis etiological results.

Decision-level machinery fault prognosis using N-BEATS-based degradation feature prediction and reconstruction

Condition monitoring signals provide sufficient information about the health of machines and, therefore, are widely used for fault diagnosis, prognosis, and health management. Existing approaches generally extract one or more degradation features from original signals collected in a time interval and predict the remaining useful life of machinery based on a selected or fused feature under a pre-specified threshold. However, using a single feature is often inadequate in terms of the accuracy of fault prognosis due to the interval-based extraction procedure. To overcome the shortcoming, a new prognosis framework is proposed for machinery based on original condition monitoring signals in this paper. Technically, the Box-Cox transformation is first performed on the original signals point by point to construct a series of degradation features without losing information. Then, the neural basis expansion analysis for time series (N-BEATS) that has robust performance in time-series prediction is utilized to predict the future evolution for each feature with high volatility. By leveraging the similarity of multiple Box-Cox features, a parameter-based transfer learning method is proposed to reduce the computation complexity. Finally, we reconstruct the future original signals of machinery through the inverse Box-Cox transformation. Since the reconstructed original signals are noisy, a new failure criterion suitable for decision-level fault prognosis is defined from an industrial application perspective. An application on high speed train wheels and two follow-up simulations are used to illustrate the performance of our proposed framework in machinery fault prognosis.

Identification and maximum impact force modeling investigation for critical slugging in underwater compressed gas energy storage systems

Underwater compressed gas (air, natural gas, hydrogen, etc.) energy storage (UWCGES) is an emerging technology that is suitable for ocean energy storage. Liquid accumulation in gas transmission pipelines can be a significant obstacle in UWCGES systems. In this study, an experimental investigation is conducted to improve the understanding of the kinematics of liquid accumulation. A pressure signal is used for identifying the state of fluid flow in the pipe. Seven kinds of time-domain features are selected through monotonicity. An adaptive evaluation method for slug flow is proposed. The fused features are submitted as a Gaussian distribution through Box-Cox transformation (BCT), and ± 3 sigma criterion is set as the threshold range to adaptively identify slug flow. Beyond this, a maximum impact force model is established and the largest impact force produced during slugging is predicted. It is found that the inlet velocity, followed by the inclination angle of the pipe and the volume of liquid accumulation, most notably influences the impact force. The model and results have the potential to be further developed for remote monitoring of liquid accumulation in pipelines, which is important for UWCGES systems.

The Surface Roughness Model When Turning of 1066 Steel by TiC Insert

Surface roughness (SR) is a very important parameter to evaluate the quality of a machining process in mechanical production. The construction of the surface roughness model during the machining process will be the basis for predicting the surface roughness for each specific case. The construction of a surface roughness model when turning of 1066 steel was done in this study. An experimental process was performed with a total of fifteen experiments. These fifteen experiments have been designed according to a Box-Behnken matrix. In each experiment, the values of three parameters were changed, including cutting speed (V), feed rate (S), and depth of cut (t). Surface roughness values were also measured at each experiment and then a surface rouhness model was built. This model demonstrates the mathematical relationship between surface roughness and three cutting parameters. A second surface roughness model was also established using the Box-Cox transformation. The accuracy of these two models was compared through three parameters including coefficient R2, R2(pred), and R2(adj). The results show that all three parameters of the second model are better than the first model. In other words, the accuracy of the surface roughness model has been improved through the use of Box-Cox transformation to convert the data.

Remaining useful life prediction of Lithium-ion batteries by PSO-SVM based on multi-feature extraction

With the development of renewable energy, energy storage system can solve the problem of high volatility of wind and solar power generation, and lithium battery is widely used in energy storage field because of its small size, lightweight and strong storage capacity. But safety and reliability issues need to be considered during its use. Predicting the remaining life of the battery is convenient for managing and maintaining the energy storage system, which is an important step to ensure the safe operation of the equipment. In this paper, we adopt a data-driven approach to predict the remaining life of lithium batteries, comprehensively considering the whole charging and discharging process and the influence of temperature on the aging of lithium batteries, from which eight suitable features are extracted, followed by a feature-enhanced Box-Cox transformation to improve the correlation between the features and the aging state of batteries. Based on this, SVM is used to construct the prediction model and PSO is used to adaptively optimize the parameters of the original SVM model, thus further improving the prediction accuracy. The feature-enhanced PSO-SVM prediction can be performed on a single group of batteries, and also has high prediction accuracy for untrained batteries. The proposed features and the optimized model possess a certain reliability, and the model outperforms general algorithms in terms of accuracy.