Support Vector Machine Regression Method Research Articles

The combination mode of forest and SVM for power network disaster response failure identification

The damage to power transmission and distribution equipment in different regions is inseparable from the erosion of natural and artificial disasters, affecting the power grid's regular power supply and users' typical power consumption. Therefore, research on failure risk prediction technology for power transmission and distribution equipment under different disasters has increasingly become critical to power grid reliability. Based on the combination mode of forest and Support Vector Machine (SVM), this study proposes a disaster response failure prediction model that considers the time cumulative effect of disaster elements. Based on this model, the failure degree indicators of different disaster elements are constructed by analyzing the relationship between disaster frequency and time. Then, the historical disaster data is used to calculate the index and form a training set with the monitoring data. The binary decision graph transformation method is combined with the adjacent node priority method to obtain the minimum cut set, and the SVM regression method is used to train to get the prediction model. The experimental results show that the effectiveness and accuracy of the proposed method are verified through the analysis of numerical examples and comparison of various techniques. The model's accuracy is 8.63 % higher than that of the traditional disaster prediction model, and the error rate of disaster failure prediction analysis is not more than 0.01.

Intelligent soaring and path planning for solar-powered unmanned aerial vehicles

PurposeThis study aims to study the thermal identification issue by harvesting both solar energy and atmospheric thermal updraft for a solar-powered unmanned aerial vehicle (SUAV) and its electric energy performance under continuous soaring conditions.Design/methodology/approachThe authors develop a specific dynamic model for SUAVs in both soaring and cruise modes. The support vector machine regression (SVMR) is adopted to estimate the thermal position, and it is combined with feedback control to implement the SUAV soaring in the updraft. Then, the optimal path model is built based on the graph theory considering the existence of several thermals distributed in the environment. The procedure is proposed to estimate the electricity cost of SUAV during flight as well as soaring, and making use of dynamic programming to maximize electric energy.FindingsThe simulation results present the integrated control method could allow SUAV to soar with the updraft. In addition, the proposed approach allows the SUAV to fly to the destination using distributed thermals while reducing the electric energy use.Originality/valueTwo simplified dynamic models are constructed for simulation considering there are different flight mode. Besides, the data-driven-based SVMR method is proposed to support SUAV soaring. Furthermore, instead of using length, the energy cost coefficient in optimization problem is set as electric power, which is more suitable for SUAV because its advantage is to transfer the three-dimensional path planning problem into the two-dimensional.

Texture feature similarity-based roughness intelligent evaluation: a case study applied to milled surfaces

Surface roughness is of great significance in maintaining mechanical performance and improving the reliability of the equipment. However, fast surface roughness evaluations that are sufficiently stable and efficient for engineering in situ use have not yet been realized. To address this issue, an image-driven roughness intelligent method is proposed in this research. By evaluating the texture similarity intelligently between the testing image and the reference image, the surface roughness of the testing image can be acquired. Firstly, with a proposed adaptive texture extraction method, the texture feature of an image can be extracted even under a complex background. Secondly, by establishing the graph structure of the texture grayscale features, the similarity between different images is evaluated. Finally, by establishing a sparrow-optimized support vector machine regression method, the mapping relationship between the similarity and the surface roughness can be acquired. The experimental results indicate that the proposed method for intelligent evaluation of roughness has superior prediction performance (the average relative prediction error of Ra and Rz are 8.8156% and 8.0571%, respectively). Therefore, this work provides a useful tool for non-contact detection of workpiece surface roughness.

Quantitative determination of the carbonyl value in frying oils based on LF-NMR combined with chemometrics

Low-field nuclear magnetic resonance (LF-NMR) combined with chemometrics was used for the rapidly and nondestructively quantitative determination of carbonyl value (CV) of frying oils during frying process. The quality indices of frying oils during frying process were analyzed as well as the changes of its LF-NMR signal, followed by exploring the correlation between the quality indices of frying oils and the relaxation-characteristic indices and investigating the influences of modeling data and preprocessing methods on the modeling performance. Based on the echo decay curves data of LF-NMR and the relaxation-characteristic indices data, the prediction model of CV was established by partial least square (PLS), principal component regression (PCR), multiple linear regression (MLR), and support vector machine regression (SVR). Results showed that there was a significant correlation between CV and relaxation-characteristic indices (p < 0.01) with both modeling data and preprocessing methods affecting the model performance. The prediction performance of non-linear SVR model was significantly superior to that of linear modeling methods (PLS, PCR, and MLR). The best prediction model of CV was built using the SVR method based on echo decay curves data. The root mean square error of prediction and coefficient of determination were 2.599 mmol/kg and 0.9873, respectively.

Анализ и прогнозирование трафика современных телекоммуникационных систем на основе методов искусственного интеллекта

In recent years, artificial intelligence technologies have demonstrated significant success in solving the problem of traffic analysis and forecasting in various telecommunication systems. Forecasting allows the telecom operator to know about the future behavior of the network, take timely necessary measures to improve the quality of customer service, and decide on the need to install or upgrade equipment. Using data collected from IoT mobile devices as an example, this article provides an overview and analysis of various time series forecasting models describing the traffic behavior of telecommunication systems. Forecasting models such as the exponential smoothing method, linear regression, the autoregressive integrated moving average (ARIMA) method, the support vector machine regression method, the N-BEATS method, which uses fully connected layers of a neural network for forecasting a one-dimensional time series, are discussed; the features of some of them are briefly outlined. For a specific data array, data preparation operations are described: removing unused columns, replacing missing data on transaction durations with their median values, and describing the main statistical characteristics of the data array. A preliminary data analysis is presented, which consists of using smoothing methods: moving average and exponential smoothing. The process of training models and a comparative analysis of the quality of their training are described. For this data set, it was concluded that for the UDP protocol the ARIMA model has the best learning quality, for the TCP protocol - linear regression and the Theta model, for the HTTPS protocol – linear regression, ARIMA and N-BEATS.

Water Quality Prediction of Small-Micro Water Body Based on the Intelligent-Algorithm-Optimized Support Vector Machine Regression Method and Unmanned Aerial Vehicles Multispectral Data

Accurate prediction of spatial variation in water quality in small microwaters remains a challenging task due to the complexity and inherent limitations of the optical properties of small microwaters. In this paper, based on unmanned aerial vehicles (UAV) multispectral images and a small amount of measured water quality data, the performance of seven intelligent algorithm-optimized SVR models in predicting the concentration of chlorophyll (Chla), total phosphorus (TP), ammonia nitrogen (NH3-N), and turbidity (TUB) in small and micro water bodies were compared and analyzed. The results show that the Gray Wolf optimized SVR model (GWO-SVR) has the highest comprehensive performance, with R2 of 0.915, 0.827, 0.838, and 0.800, respectively. In addition, even when dealing with limited training samples and different data in different periods, the GWO-SVR model also shows remarkable stability and portability. Finally, according to the forecast results, the influencing factors of water pollution were discussed. This method has practical significance in improving the intelligence level of small and micro water body monitoring.

Insights into the Effects of Study Area Size and Soil Sampling Density in the Prediction of Soil Organic Carbon by Vis-NIR Diffuse Reflectance Spectroscopy in Two Forest Areas

Sustainable forest land management requires measuring and monitoring soil organic carbon. Visible and near-infrared diffuse reflectance spectroscopy (Vis-NIR, 350–2500 nm), although it has become an important method for predicting soil organic carbon (SOC), requires further studies and methods of analysis to realize its full potential. This study aimed to determine if the size of the study area and soil sampling density may affect the performance of Vis-NIR diffuse reflectance spectroscopy in the prediction of soil organic carbon. Two forest sites in the Calabria region (southern Italy), which differ in terms of area and soil sampling density, were used. The first one was Bonis catchment area (139 ha) with a cover consisting mainly of Calabrian pine, while the second was Mongiana forest area (33.2 ha) within the “Marchesale” Biogenetic Nature Reserve, which is covered by beech. The two study areas are relatively homogeneous regarding parent material and soil type, while they have very different soil sampling density. In particular, Bonis catchment has a lower sampling density (135 samples out of 139 ha) than Mongiana area (231 samples out of 33.2 ha). Three multivariate calibration methods (principal component regression (PCR), partial least square regression (PLSR), and support vector machine regression (SVMR)) were combined with different pretreatment techniques of diffuse reflectance spectra (absorbance, ABS, standard normal variate, SNV, and Savitzky–Golay filtering with first derivative (SG 1st D). All soil samples (0–20 cm) were analyzed in the laboratory for SOC concentration and for measurements of diffuse reflectance spectra in the Vis-NIR region. The set of samples from each study area was randomly divided into a calibration set (70%) and a validation set (30%). The assessment of the goodness for the different calibration models and the following SOC predictions using the validation sets was based on three parameters: the coefficient of determination (R2), the root mean square error (RMSE), and the interquartile range (RPIQ). The results showed that for the two study areas, different levels of goodness of the prediction models depended both on the type of pretreatment and the multivariate method used. Overall, the prediction models obtained with PLSR and SVMR performed better than those of PCR. The best performance was obtained with the SVMR method combined with ABS + SNV + SG 1st D pretreatment (R2 ≥ 0.77 and RPIQ &gt; 2.30). However, there is no result that can absolutely provide definitive indications of either the effects of the study area size and soil sampling density in the prediction of SOC by vis-NIR spectroscopy, but this study fostered the need for future investigations in areas and datasets of different sizes from those in this study and including also different soil landscapes.

Development of a screening method for adulteration detection in coconut milk via mid-infrared spectroscopy: A study of linear and nonlinear regression method

In the present study, we developed a screening method for detection of adulteration in coconut milk via mid-infrared spectroscopy. Linear and nonlinear regression methods (principal component regression (PCR), partial least squares regression (PLSR), and support vector machine regression (SVMR)) were employed and compared to achieve an optimal screening method. Spectral data were scanned using the FTIR benchtop with a wavelength range of 4000–16702 nm. The calibration models of the linear and nonlinear regression methods were developed using the leave-one-out cross-validation method before testing using predictive data that had been prepared. Furthermore, five spectral data treatment techniques were employed to improve the accuracy of the proposed calibration model. The results obtained show that the SVMR method is better than PCR and PLSR for the detection of adulteration in coconut milk by mid-infrared data spectroscopy. The coefficient of determination for calibration (R2c) and prediction (R2p), the root mean square error of calibration (RMSEC) and prediction (RMSEP) and the ratio of prediction to deviation (RPD) using the SVMR method were 100%, 0.81, 98.40%, 0.87 and 7.86, respectively. Furthermore, based on RPD analysis, it is known that the SVMR model can be used to perform excellent quality control of water-adulterated coconut milk.

Acid Number Prediction Model of Lubricating Oil Based on Mid-Infrared Spectroscopy

The monitoring and replacement of lubricating oil has an important impact on mechanical equipment. In this study, based on the infrared spectroscopy monitoring method, an acid value index prediction model is established. The support vector machine regression method is used to quantitatively analyze the acid number of the oil sample, which verifies the stability and predictive ability of the quantitative prediction model, and we provide a theoretical basis and practical examples for the online monitoring of oil indicators. In addition, a support vector machine regression model is established by observing the changing law of the spectral absorption peak and selecting the dominant wavelength, and it is compared against the prediction algorithm of the long- and short-term memory network. By comparing the deviation relationship between the predicted value and the real chemical value, the feasibility of the infrared spectroscopy prediction model is verified. The experimental results show that the correlation coefficient between the predicted value of the model and the actual measured value reaches 0.98. This proves that the prediction effect of the prediction model based on the infrared spectrum data and the support vector machine regression method is better than that of the long- and short-term memory network trend prediction model, and the predicted results are reliable.

Suitability of different multivariate analysis methods for monitoring leaf N accumulation in winter wheat using in situ hyperspectral data

Leaf nitrogen accumulation (LNA) is an important indicator of crop growth, and the real-time and nondestructive estimation of LNA by hyperspectral remote sensing is crucial for crop N management. In this study, five field experiments were conducted over four successive years at three ecological sites (Xinyang, Zhengzhou and Zhoukou) in Henan, China, incorporating different N amounts and wheat varieties. The canopy spectral data were obtained by a ground-based spectrometer, and the original spectral reflectance (R) was transformed using continuum removal (CR), first-derivative reflectance (FDR) and inverse-log reflectance (log (1/R)) methods. The effective bands of the above four transformed reflectance data were selected by the variable importance in projection (VIP) scores of the partial least square regression method (PLSR). Then, four multivariate regression methods (PLSR, stepwise multiple linear regression (SMLR), extreme learning machine (ELM) and support vector machine (SVM)) were used to develop an optimal forecasting model for LNA in winter wheat. The results show that FDR spectral data had the best comprehensive performance and were a better method for spectral data preprocessing. In the comparison of different nonparametric algorithms, the SVM model performed the best using the full wavelengths, in which R2cal/val and RMSEcal/val were 0.927–0.963/0.899–0.925 and 0.725–1.066/1.161–1.331 g m−2, respectively. The characteristic bands of FDR were mainly distributed in the visible light and red edge regions (445, 501, 546, 683, 747 and 771 nm, respectively), and these effective bands were consistent under different field conditions. The SVM regression method using the effective wavelengths of FDR exhibited excellent performance for LNA prediction with R2val = 0.875 and RMSEval = 1.294. In addition, the FDR-PLSR-SVM model produced satisfactory inversion results under different field conditions. The accuracy of the filling period was better than that of the reviving-anthesis growth period, which might be due to diminished N dilution and stable canopy structure during the reproductive stage. R2val and RMSEval of the FDR-PLSR-SVM model during the filling period were 0.914 and 1.156 g m−2 in Xinyang, 0.825 and 1.542 g m−2 in Zhengzhou and 0.796 and 1.902 g m−2 in Zhoukou, respectively. Considering the advantages of simple and efficient computer calculations, the FDR-PLSR-SVM algorithm is strongly recommended to predict the LNA in winter wheat. This study provides a reference basis for the selection of effective wavelengths for N nutrition diagnosis by unmanned aerial vehicles and satellite remote sensing in the future.

A novel multi-innovation gradient support vector machine regression method

For the regression problem of support vector machine, the solution processes of the most existing methods use offline datasets, which cannot be realized online. For this problem, this paper presents a new online approach to identify these unknown parameters contained in the support vector machine. A new cost function is constructed by substituting the error term into the standard cost function, which is different from the standard support vector machine, and the gradient descent approach is then used to minimize the newly created loss function, thus proposing a stochastic gradient support vector machine algorithm to estimate the unknown parameters based on the recursive identification methods. Furthermore, to advance the property of the stochastic gradient support vector machine algorithm, a moving data window is used to widen the scalar information into a fixed-length innovation vector, thereby increasing the amount of information used in the parameter estimation based on the multi-innovation identification theory. In addition, the forgetting factor is brought into the proposed algorithms, and the corresponding forgetting factor recursive algorithms are derived. These methods are recursive identification methods, which may be implemented online and are more efficient in terms of computing. Finally, utilizing the MatLab platform, the validity and usefulness of the explored methodologies are proven using several numerical simulation examples.

Convergence rates of support vector machines regression for functional data

Support vector machines regression (SVMR) is an important part of statistical learning theory. The main difference between SVMR and the classical least squares regression (LSR) is that SVMR uses the ϵ-insensitive loss rather than quadratic loss to measure the empirical error. In this paper, we consider SVMR method in the field of functional data analysis under the framework of reproducing kernel Hilbert spaces. The main tool used in our theoretical analysis is the concentration inequalities for suprema of some appropriate empirical processes. As a result, we establish explicit convergence rates of the prediction risk for SVMR, which coincide with the minimax lower bound obtained recently in literature for LSR.

Ability of near infrared spectroscopy and chemometrics to measure the phytic acid content in maize flour

Phytic acid is one of the important biochemical components in maize as in many plant species. Near infrared spectroscopy has a potential for determination of the phytic acid content in the maize grain. However, there are a limited number of studies on the determination of phytic acid in maize. Also, the effect of chemometric methods on the success of near infrared spectroscopy calibration models for phytic acid content has not been investigated sufficiently yet. To fill these gaps, we create a total of 360 different prediction models and evaluate the effect of chemometric methods on prediction robustness. To develop calibration models, 4 derivatives, 5 pretreatments, 9 wavelength selection methods were used, and partial least squares regression and support vector machines regression methods were applied. Model reliability was evaluated by external validation. Results revealed that spectral pretreatment and wavelength selection methods improve model prediction results. In general, support vector machines yielded more successful results than partial least squares models in detecting phytic acid. The best model was the combination of first derivative + standard normal variate + interval partial least squares combined with support vector regression. While creating calibration models for phytic acid detection, it was concluded that the use of appropriate chemometric methods increases the success of the model.

Prediction of Insulator Pollution Flashover Voltage Based on Data Mining Technology

The pollution flashover of insulators is of great significance to the safe and stable operation of power grids. Therefore, studying the pollution flashover voltage of insulators has very important guiding value for the external insulation prevention of power grids. For this reason, this article has obtained the flashover voltage under different salt densities and gray densities through manual pollution tests. Data mining technology is now widely used. In order to better apply data mining to the insulation flashover voltage prediction,this paper adopts the support vector machine regression method in data mining technology to establish a prediction model to predict the insulator pollution flashover voltage. Through this method, the flashover voltage of the insulator can be predicted in time, which is of great significance for maintaining the safe and stable operation of the power system.

Crack Prediction Based on Wavelet Correlation Analysis Least Squares Support Vector Machine for Stone Cultural Relics

Preventive protection of cultural relics is to make use of all the science and technology beneficial to the research and protection of archaeological heritage to predict the disease of cultural relics. The existing preventive cultural relics protection system has made some achievements in environmental monitoring, but the analysis and utilization of large data of cultural relics are still insufficient. In this paper, under the idea of multisource information fusion, a least squares support vector machine regression method based on multivariate time series wavelet correlation analysis is proposed to achieve accurate crack prediction of stone cultural relics. Firstly, the correlation of multivariate time series of stone cultural relics are quantitatively analyzed and the validity of characteristic variables of the crack is discriminated by wavelet correlation analysis; then, a least squares support vector machine prediction model is constructed based on the correlation obtained from the analysis; finally, the good performance of the method is verified by using the environmental monitoring data of the rock mass fracture in the North Qianfo Cliff of Dafo Temple in Binzhou City of Shaanxi Province. The experimental results show that the proposed method is more effective than the traditional backpropagation neural network, support vector machine, and relevance vector machine regression methods. This method is universal and easy to implement for multisource data prediction of nonmovable cultural relics diseases. It provides a scientific theoretical reference for the preventive protection of cultural relics.

NIMG-22. INTEGRATION OF A RADIOMIC SIGNATURE, CLINICAL VARIABLES AND PLASMA CELL-FREE DNA IN ADULT PATIENTS WITH NEWLY DIAGNOSED GLIOBLASTOMA PREDICTS PATIENT SURVIVAL AND IMPROVES DISEASE STRATIFICATION

Abstract PURPOSE We have previously demonstrated the potential role of liquid biopsy, specifically plasma cell-free DNA (cfDNA), as a non-invasive biomarker for prognostication in patients with glioblastoma. In separate prior studies, we have also developed MRI-based radiomic signatures to predict survival outcomes in glioblastoma. In this study, for the first time, we evaluated the potential of combining radiomic signatures, epidemiological and clinical variables, and plasma cfDNA quantification for upfront prediction of overall survival (OS) in patients with newly diagnosed glioblastoma. METHODS Quantitative radiomic features were extracted from multiparametric MRI (T1, T1Gd, T2, T2-FLAIR) scans of a discovery cohort of 505 and an independent replication cohort of 50 IDH-wildtype glioblastoma patients. For the independent replication cohort, pre-surgical plasma cfDNA was extracted and quantified. In the first stage, a radiomic signature was created for stratification of patients into categories of short (OS ≤ 6 months) and long (OS ≥ 18 months) survivors using a cross-validated XGBoost method based on the discovery cohort, which was tested independently on the replication cohort. In the second stage, the radiomic signature and clinical variables were integrated to build a second-stage signature using a cross-validated support vector machine (SVM) classifier to stratify the patients into short and long survivor categories. In the third stage, the value of the second-stage signature integrated with cfDNA concentration was assessed through a cross-validated SVM regression method. RESULTS The combination of radiomic, clinical, and cfDNA variables resulted in the best overall predictive accuracy, with Pearson’s correlation coefficient of 0.59 (p&amp;lt; 0.0001) between actual and predicted OS. CONCLUSION In this study, we evaluated the value of combining plasma cfDNA, radiomic, and clinical variables for predicting OS, and showed that it could act as an effective non-invasive prognostic and patient stratification tool in patients with newly diagnosed glioblastoma.

Destek Vektör Makineleri ve Türkiye’deki Enerji Santrallerinde Doğal Gaz Tüketimi Üzerine Bir Uygulama

Bu çalışmanın amacı, destek vektör makinelerinde regresyon yöntemini kullanarak Türkiye'deki enerji santrallerinin doğalgaz tüketimi üzerine ön kestirim yapmaktır. Bu amaçla, veri seti 2013-2018 yılları arasında Türkiye Enerji Piyasası Düzenleme Kurumu ve Enerji İşleri Genel Müdürlüğünden elde edilmiştir. Bu çalışmada ilk olarak, Türkiye'de doğalgazın enerji piyasasındaki yeri, birincil enerji kaynakları içindeki payı, üretim, tüketim, ithalat ve ihracat değerleri incelenmiştir. Bu değerlerin ölçümündeki farklılıklardan dolayı, ilgili veri seti istatistiksel analizden önce standartlaştırılmıştır. Enerji santralleri tüketimi bağımlı değişken iken; sanayi tüketimi, şehir tüketimi, üretim, ithalat ve ihracat değerleri bağımsız değişken olarak belirlenmiştir. Destek Vektör Regresyonda kullanılan tüm kernel fonksiyonları (Doğrusal, Polinomiyal, Radyal Tabanlı Fonksiyon (RTF) ve Sigmoid) test edilmiştir. En küçük Hata Kareler Ortalaması (HKO)'na sahip olan RTF tahmin kernel fonksiyonu olarak seçilmiştir. Daha sonra, destek vektörler, ağırlıklar ve karar sabiti belirlenmiştir. Ağırlıklar ve destek vektörler çarpılıp sapma eklenerek, son model elde edilmiştir. Son model yardımıyla da Mayıs-Aralık 2018 için Türkiye'deki enerji santrallerinin doğalgaz tüketimlerine ait önkestirimler yapılmıştır. Sonuçlar ilgili tablolarda gösterilmiştir.

A Comparative Study of PLSR and SVM-R with Various Preprocessing Techniques for the Quantitative Determination of Soluble Solids Content of Hardy Kiwi Fruit by a Portable Vis/NIR Spectrometer.

Linear partial least square and non-linear support vector machine regression analysis with various preprocessing techniques and their combinations were used to determine the soluble solids content of hardy kiwi fruits by a handheld, portable near-infrared spectroscopy. Fruits of four species, namely Autumn sense (A), Chungsan (C), Daesung (D), and Green ball (Gb) were collected from five different areas of Gwangyang (G), Muju (M), Suwon (S), Wonju (Q), and Yeongwol (Y) in South Korea. The dataset for calibration and prediction was prepared based on each area, species, and in combination. Half of the dataset of each area, species, and combined dataset was used as calibrated data and the rest was used for model validation. The best prediction correlation coefficient ranges between 0.67 and 0.75, 0.61 and 0.77, and 0.68 for the area, species, combined dataset, respectively using partial least square regression (PLSR) method with different preprocessing techniques. On the other hand, the best correlation coefficient of predictions using the support vector machine regression (SVM-R) algorithm was 0.68 and 0.80, 0.62 and 0.79, and 0.74 for the area, species, and combined dataset, respectively. In most cases, the SVM-R algorithm produced better results with Autoscale preprocessing except G area and species Gb, whereas the PLS algorithm shows a significant difference in calibration and prediction models for different preprocessing techniques. Therefore, the SVM-R method was superior to the PLSR method in predicting soluble solids content of hardy kiwi fruits and non-linear models may be a better alternative to monitor soluble solids content of fruits. The finding of this research can be used as a reference for the prediction of hardy kiwi fruits soluble solids content as well as harvesting time with better prediction models.

Multivariate Analysis Models Based on Full Spectra Range and Effective Wavelengths Using Different Transformation Techniques for Rapid Estimation of Leaf Nitrogen Concentration in Winter Wheat.

To develop a stable estimation model and identify effective wavelengths that could explain the variations in leaf nitrogen (N) concentration with different N supplies, growing seasons, ecological locations, growth stages, and wheat cultivars. Four field experiments were performed during two consecutive years (2017–2019) at three sites (Yuanyang, Hebi, and Wenxian) in Henan, China. In situ canopy spectral reflectance data under the aforementioned N supply conditions were obtained over a range of 400–950 nm (visible and near-infrared region). On the basis of the canopy raw spectral reflectance data and their subsequent transformation by two different techniques, first-derivative reflectance (FDR) and continuum removal (CR), four multivariate regression methods were comparatively analyzed and used to develop predictive models for estimating leaf N concentration: multiple linear regression (MLR), principal component regression (PCR), partial least square (PLS), and support vector machine (SVM). Results showed that leaf N concentration and canopy reflectance significantly varied with the levels of N fertilization, and a good correlation was observed for all the spectral techniques. Seven wavelengths with relatively higher r values than the bands of the raw spectra centered at 508, 525, 572, 709, 780, 876, and 925 nm were specified using the FDR technique. Based on the full wavelengths, the FDR-SVM model exhibited a good performance for leaf N concentration estimation, with coefficients of determination (r2val) for the validation datasets and corresponding relative percent deviations (RPDval) values of 0.842 and 2.383, respectively. However, the FDR-PLS yielded a more accurate assessment of the leaf N concentration than did the other methods, with r2val and RPDval values of 0.857 and 2.535, respectively. The variable importance in projection (VIP) scores from the FDR-PLS with the all canopy spectral region were used to screen the effective wavelengths of the spectral data. Therefore, six effective wavelengths centered at 525, 573, 710, 780, 875, and 924 nm were identified for leaf N concentration estimation. The SVM regression method with the effective wavelengths showed excellent performance for leaf N concentration estimation with r2val = 0.823 and RPDval = 2.280. These results demonstrated that the in situ canopy spectral technique is promising for the estimation of leaf N concentration in winter wheat based on the FDR-PLS regression model and the effective wavelengths identified.

Gas face seal status estimation based on acoustic emission monitoring and support vector machine regression

The difficulty of knowing the real-time status of gas face seals is the main cause of common problems, including sudden failure, ineffective diagnosis, and unpredictability of service life. This study analyzed the acoustic emission signals generated from experiments, uncovering their features in terms of the frequency distribution, periodic fluctuations, and the behaviors during different operation phases. A new vectorization procedure was designed according to the knowledge of informative acoustic emission features. Based on the vectorization procedure, a support vector machine regression method was applied to develop models predicting the eccentric load on the stator of the seal. Cross-validation was conducted to evaluate the regression performance and search for a proper kernel scale. This study found the informative features of acoustic emissions at different timescales and during different seal operation phases, and particularly the great informative potential of certain segments of the starting and stopping phases. The vectorization and support vector machine regression were shown to be effective in estimating the loads in experiments with cross-validation. Thus, a method for estimating the status of gas face seals based on acoustic emission monitoring was established.